AI-BASED ENERGY EDGE PLATFORM, SYSTEM, AND METHOD

Abstract

An AI-based energy edge platform has a wide range of capabilities, components, and capabilities for managing and improving legacy infrastructure and aligning with distributed systems to support critical use cases for a variety of enterprises. The platform may incorporate new technologies that enable efficiency, agility, engagement, and profitability for the ecosystem and individual energy edge nodes. Embodiments may forecast, plan, and manage energy demand and utilization in a more distributed environment. Embodiments may use technologies to more efficiently filter, process, and move data across AI, IoT, and communications networks. Platform embodiments may leverage energy market connectivity, communications, and trade enablement platforms. Embodiments may employ intelligent provisioning, data aggregation, and analytics.

Description

[Technical Field]

[0001] (CROSS-REFERENCE TO RELATED APPLICATIONS) This application claims the benefit of U.S. Provisional Patent Application No. 63 / 282,510, filed November 23, 2021, U.S. Provisional Patent Application No. 63 / 291,311, filed December 17, 2021, U.S. Provisional Patent Application No. 63 / 299,727, filed January 14, 2022, and U.S. Provisional Patent Application No. 63 / 302,016, filed January 21, 2022, the entire disclosures of which are incorporated by reference. [Background technology]

[0002] (background) Energy remains a critical component of the global economy and is undergoing an evolution and transformation that will involve changes in energy generation, storage, planning, demand management, consumption, and supply systems and processes. These changes are enabled by the development and convergence of numerous diverse technologies, including more decentralized, modular, mobile, and portable energy generation and storage technologies that will make energy markets more decentralized and localized, and a variety of technologies that will facilitate energy management in more decentralized systems, such as edge, Internet of Things, networking technologies, advanced computation and artificial intelligence technologies, and transaction-enabling technologies (e.g., blockchain, distributed ledgers, and smart contracts). The convergence of these networking, computation, and intelligence technologies with more decentralized energy technologies is referred to herein as the "energy edge."

[0003] Over the next few decades, the energy market is expected to evolve and transform from a highly centralized model dependent on fossil fuels and controlled grids to a more decentralized and distributed model that includes many more localized generation-storage-consumption systems. This transition will likely continue for many years, with hybrid systems in which traditional grids become more intelligent and distributed systems play a larger role. There is a need for a platform that can interface with distributed systems and make it easier to manage and improve legacy infrastructure. Summary of the Invention

[0004] (overview) Provided herein is an AI-based energy edge platform with a wide range of features, components, and capabilities for managing and improving legacy infrastructure and coordinating with distributed systems that support critical use cases for various enterprises. The platform can incorporate new technologies that enable efficiency, agility, engagement, and profitability for the ecosystem and individual energy edge nodes. Embodiments are guided by, and potentially integrated with, methodologies and systems used to forecast, plan, and manage energy demand and utilization in larger distributed environments. Embodiments can use AI and AI enablers such as IoT, which may be deployed in massively dense data environments (reflecting the proliferation of sensors in smart energy systems and IoT), and technologies that more efficiently filter, process, and move data over communication networks. Platform embodiments can leverage energy market connectivity, communications, and transaction enablement platforms. Embodiments can employ intelligent provisioning, data aggregation, and analytics. Among many use cases, the platform can enable improvements in optimizing energy generation, storage, supply, and / or corporate consumption in businesses (e.g., buildings, data centers, factories, etc.), integrating and using new power generation and energy storage technologies and assets (distributed energy resources, or "DERs"), optimizing energy utilization across existing networks, and digitalizing existing infrastructure and support systems.

[0005] The present disclosure will become more fully understood from the detailed description and the accompanying drawings. [Brief explanation of the drawings]

[0006] [Figure 1] FIG. 1 is a schematic diagram illustrating an introduction to the platform and key elements, according to some embodiments.

[0007] [Figure 2A] 2A and 2B are schematic diagrams illustrating an introduction to the major subsystems of the major ecosystem, according to some embodiments. [Figure 2B] 2A and 2B are schematic diagrams illustrating an introduction to the major subsystems of the major ecosystem, according to some embodiments.

[0008] [Figure 3] FIG. 3 is a schematic diagram illustrating more details of a distributed energy generation system according to some embodiments.

[0009] [Figure 4] FIG. 4 is a schematic diagram illustrating details of data resources, according to some embodiments.

[0010] [Figure 5] FIG. 5 is a schematic diagram showing more details of configured Energy Edge participants, according to some embodiments.

[0011] [Figure 6] FIG. 6 is a schematic diagram illustrating details of an intelligence enabling system, according to some embodiments.

[0012] [Figure 7] FIG. 7 is a schematic diagram illustrating more details of AI-based energy orchestration, according to some embodiments.

[0013] [Figure 8] FIG. 8 is a schematic diagram showing more details of configurable data and intelligence, according to some embodiments.

[0014] [Figure 9] FIG. 9 is a schematic diagram illustrating the dual-process learning functionality of a dual-process artificial neural network, according to some embodiments.

[0015] [Figure 10] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 11] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 12] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 13]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 14] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 15] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 16]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 17] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 18] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 19]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 20] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 21] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 22]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 23] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 24] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 25]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 26] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 27] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 28]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 29] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 30] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 31]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 32] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 33] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 34]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 35] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 36] 10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure. [Figure 37]10-37 are schematic diagrams of embodiments of neural net systems that may be connected to, integrated into, and accessible by a platform for enabling intelligent transactions, including expert systems, self-organization, machine learning, artificial intelligence, and neural net systems trained for pattern recognition, classification of one or more parameters, characteristics, or phenomena, support for autonomous control, and other purposes, in accordance with embodiments of the present disclosure.

[0016] [Figure 38] FIG. 38 is a schematic diagram of an example embodiment of a quantum computing service according to some embodiments of the present disclosure.

[0017] [Figure 39] FIG. 39 is a diagram illustrating quantum computing service request processing according to some embodiments of the present disclosure.

[0018] [Figure 40] FIG. 40 is a perspective view of the thalamic services according to the present disclosure and how they are coordinated within the module.

[0019] [Figure 41] FIG. 41 is another perspective view showing thalamic services according to the present disclosure and how they coordinate within modules. DETAILED DESCRIPTION OF THE INVENTION

[0020] Figure 1: Platform introduction and key elements In embodiments, provided herein is an AI-based energy edge platform 102, sometimes referred to herein for convenience simply as the platform 102, including a set of intelligent, sometimes autonomous or semi-autonomous, systems, subsystems, applications, processes, methods, modules, services, layers, devices, components, machines, products, subsystems, interfaces, connections, and other elements that work together to enable the orchestration and management of power and energy in various ecosystems and environments, including distributed entities (sometimes referred to herein as “distributed energy resources” or “DERs”) and other energy resources and systems that generate, store, consume, and / or transport energy, and including IoT, edge, and other devices and systems that process data related to DERs and other energy resources and can be used to inform, analyze, control, optimize, predict, and otherwise assist in the orchestration of distributed energy resources and other energy resources.

[0021] In an embodiment, the platform 102 enables a set of configured stakeholder energy edge solutions 108 having a wide range of functions, applications, capabilities, and uses that can be achieved, without limitation, by using or orchestrating a set of advanced energy resources and systems 104, including DERs and the like. The configured stakeholder energy edge solution 108 may integrate domain-specific stakeholder data, such as proprietary data sets generated in connection with an enterprise's operations, analytics, and / or strategy, real-time data from stakeholder assets (e.g., collected by IoT and edge devices located in proximity to the stakeholder's assets and operations), and real-time data from stakeholder-specific energy resources and systems 104 (e.g., available energy generation, storage, or distribution systems that may be deployed at stakeholder locations to augment or replace the electric grid) into a solution that meets the stakeholder's energy needs and capabilities, including baseline, period, and peak energy needs for conducting operations such as large-scale data processing, transportation, production of goods and materials, resource extraction and processing, heating and cooling, and many others.

[0022] In embodiments, the AI-based energy edge platform 102 (and / or its elements) and / or the set of configured stakeholder energy edge solutions 108 can obtain data from, provide data to, and / or exchange data with a set of data resources for the energy edge orchestration 110.

[0023] The AI-based energy edge platform 102 may include, integrate with, exchange data with, and / or link to a set of intelligence enablement systems 112, a set of AI-based energy orchestration, optimization, and automation systems 114, and a set of configurable data and intelligence modules and services 118.

[0024] The set of intelligence enabling systems 112 may include a set of intelligent data layers 130, a set of distributed ledger and smart contract systems 132, a set of adaptive energy digital twin systems 134, and / or a set of energy simulation systems 136.

[0025] The set of AI-based energy orchestration, optimization, and automation systems 114 may include a set of energy generation orchestration systems 138, a set of energy consumption orchestration systems 140, a set of energy market orchestration systems 146, a set of energy delivery orchestration systems 147, and a set of energy storage orchestration systems 142.

[0026] The set of configurable data and intelligence modules and services 118 may include a set of energy trading enablement systems 144, a set of stakeholder energy digital twins 148, and a set of data integration microservices 150 that may enable or contribute to the enablement of the configured set of stakeholder energy edge solutions 108.

[0027] The AI-based energy edge platform 102 may include, integrate with, link to, exchange data with, be governed by, take input from, and / or provide output to one or more artificial intelligence (AI) systems, which may include models, rule-based systems, expert systems, neural networks, deep learning systems, supervised learning systems, robotic process automation systems, natural language processing systems, intelligent agent systems, self-optimizing and self-organizing systems, and others, as described throughout this disclosure and in documents incorporated by reference herein. Unless the context indicates otherwise, reference to AI or to one or more examples of AI should be understood to encompass these various alternative methods and systems. For example, without limitation, AI systems described to enable any of the wide variety of functions, capabilities, and solutions described herein (optimization, autonomous operation, prediction, control, orchestration, etc.) should be understood to be implementable by operating on a model or rule set, by training on a training dataset of human tags, labels, etc., or by training on a training dataset of human interactions (e.g., as described below). g.,Training on training datasets of human interactions (e.g., human interactions with a software interface or a hardware system), training on resulting training datasets, training on AI-generated training datasets (e.g., the complete training dataset is generated by the AI ​​from a seed training dataset), supervised learning, semi-supervised learning, deep learning, etc.For any given function or capability described herein, various types of neural networks may be used, including any of the types described herein or in documents incorporated by reference, and in embodiments, a hybrid set of neural networks may be selected such that the more advantageous neural network type is implemented within the set to perform each element of a multi-function or multi-capability system or method. As one example among many, deep learning, or black box, systems may use gated recurrent neural networks for functions such as language translation for intelligent agents, where understanding the underlying mechanisms of AI operation is not necessary as long as the results are favorably perceived by the user, while more transparent models or systems and simpler neural networks may be used for systems of automated governance, where a deeper understanding of how inputs are transformed into outputs may be necessary to comply with regulations or policies. AI-based energy orchestration, optimization and automation system

[0028] In embodiments, the platform may employ demand forecasting, including automated forecasting using artificial intelligence or by ingesting data streams of forecast information from third parties. Demand forecasting can help inform site selection and intelligently planned network expansion, among other things. In embodiments, machine learning algorithms can generate multiple forecasts for factors such as weather, prices, solar generation, energy demand, and other factors to analyze how energy assets can best capture or generate value at different times and / or locations.

[0029] In embodiments, the AI-based energy orchestration, optimization, and automation system 114 can enable optimization of energy patterns by analyzing building or other operational energy usage and seeking to reshape the patterns for optimization (e.g., by modeling demand response to various stimuli).

[0030] The AI-based energy orchestration, optimization, and automation system 114 may be enabled by a set of intelligence enabling systems 112 that provide functions and capabilities to support a variety of applications and use cases. Intelligence Realization System Subsystems and Modules Intelligent Data Layer

[0031] The intelligence enablement system 112 may include a set of intelligent data layer 130, e.g., a set of services (including microservices), APIs, interfaces, modules, applications, programs, etc., that can consume any of the data entities and types described throughout this disclosure and undertake a wide range of processing functions, such as extraction, cleansing, normalization, calculation, transformation, loading, batching, streaming, filtering, routing, parsing, conversion, pattern recognition, content recognition, object recognition, etc. Through a set of interfaces, users of the platform 102 can configure the intelligent data layer 130 or its output to meet needs internal to the platform and / or enable further configuration, such as for a stakeholder's energy edge solution 108. The intelligent data layer 130, and more generally the intelligence enabling system 112, and / or the configurable data and intelligence modules and services 118, may access data from a variety of sources throughout the platform 102 and, in embodiments, may operate from a set of shared data resources 130, which may be included in a centralized database and / or a set of distributed databases, or may consist of a set of distributed or decentralized data sources, such as IoT or edge devices that generate energy-related event logs or streams.The intelligent data layer 130 may be configured for a wide range of energy-related tasks, such as forecasting / predicting energy consumption, generation, storage, or distribution parameters (e.g., at the individual device, sub-device, system, machine, or fleet level); optimizing energy generation, storage, distribution, or consumption (even at various levels of optimization); automatically discovering, configuring, and / or executing energy transactions (including micro-trades and / or larger transactions in spot and futures markets, as well as larger transactions in peer-to-peer group or single counterparty trading); monitoring and tracking energy consumption, generation, distribution, and / or storage parameters and attributes (e.g., baseline levels, volatility, cyclical patterns, episodic events, peak levels, etc.); monitoring and tracking energy-related parameters and attributes (e.g., pollution, carbon production, renewable energy credits, waste heat production, etc.); automatically generating energy-related alerts, recommendations, and other content (e.g., messaging to prompt or encourage preferred user behavior), and many others. Distributed Ledger and Smart Contract Systems

[0032] The energy edge intelligence enablement system 112 may include a smart contract system 132 for processing a set of smart contracts, each of which may optionally operate on a set of blockchain-based distributed ledgers. Each of the smart contracts may operate on data stored on a set of distributed ledgers or blockchains to record energy-related trading events such as energy purchases and sales (spot, forward, and peer-to-peer markets, as well as direct counterparty transactions), associated service charges, trading-related energy events such as consumption, generation, distribution, and / or storage events, and other trading-related events often related to energy, such as carbon production or reduction events, renewable energy credit events, pollution production or reduction events, etc. The set of smart contracts handled by smart contract system 132 can consume as a set of inputs any of the data types and entities described throughout this disclosure, undertake a set of computations (optionally configured with flows incorporating inputs from disparate systems in multi-step transactions), and provide a set of outputs that enable the transaction to be completed, reported (optionally recorded on a set of distributed ledgers), etc. The energy transaction enablement system 144 may be enabled or augmented by artificial intelligence, including autonomously discovering, configuring, and executing transactions according to a strategy and / or providing automation or semi-automation of transactions based on training and / or oversight by a set of transaction experts. In embodiments, smart contract system 132 may be used by energy transaction enablement system 144 (described elsewhere in this disclosure) to configure transaction solutions. Adaptive Energy Digital Twin System

[0033] Any entity, analysis result, artificial intelligence output, state, operating state, or other characteristic noted throughout this disclosure may, in embodiments, be presented in a digital twin, such as a broadly applicable adaptive energy digital twin 134 and / or a stakeholder energy digital twin 148 configured to the needs of a particular stakeholder or stakeholder solution. The adaptive energy digital twin 134 can provide, for example, visual or analytical indicators of energy consumption by a set of machines, a group of factories, a fleet of vehicles, etc., a subset of the same (e.g., to compare energy parameters by each of a set of similar machines to identify out-of-range operation), and many other aspects. The digital twin may be adaptive, such as filtering, highlighting, or otherwise adjusting the data presented based on real-time conditions, such as changing energy costs, changing driving behavior, etc. Energy Simulation System

[0034] In an embodiment, a set of energy simulation systems 136 is provided to develop and evaluate detailed simulations of energy generation, demand response, and charging management, including simulation environments that simulate the results of using various algorithms that may govern power generation across various generation assets, consumption by energy-demanding devices and systems, and energy storage. Data can be used to simulate the interaction of uncontrollable loads with optimized charging processes, among other use cases. The simulation environments can provide output to, integrate with, and share data with a set of advanced energy digital twin systems 134.

[0035] In embodiments, as more enterprises adopt hybrid infrastructure, uptime becomes more complex, requiring backup and failover strategies that span cloud, colocation, on-premises facilities, and edge infrastructure. This may include AI-based algorithms to automatically manage the energy of devices and systems within such equipment. For example, artificial intelligence may enable autonomous data center cooling and industrial controls. In embodiments, DERs 128 may be integrated with or be integrated with, for example, AI-driven computing infrastructure, smart PDUs, UPS systems, energy-aware airflow management systems, and HVAC systems. Introduction to the main subsystems and modules of an AI-based energy orchestration, optimization, and automation system

[0036] The set of AI-based energy orchestration, optimization, and automation systems 114 may include, among others, a set of energy generation orchestration systems 138, a set of energy consumption orchestration systems 140, a set of energy storage orchestration systems 142, a set of energy market orchestration systems 146, and a set of energy delivery orchestration systems 147. For example, the energy delivery orchestration system 147 may enable the orchestration of the delivery of energy to points of consumption, such as via fixed transmission lines, wireless energy transmission, delivery of fuel, delivery of stored energy (e.g., chemical or nuclear batteries), etc., and may include autonomously optimizing the combination of energy types among the aforementioned available resources based on various factors, such as location (e.g., based on distance from the grid), purpose or type of consumption (e.g., whether there is a need for very high peak energy supply, such as for a power-intensive production process), etc. Configurable Data and Intelligence Modules and Services

[0037] In an embodiment, the platform 102 may include a configurable set of data and intelligence modules and services 118. These may include an energy tradability enablement system 144, a stakeholder energy digital twin 148, energy-related data integration microservices 150, etc. Each module or service (optionally configured in a microservices architecture) may exchange data with various data resources 110 to provide relevant outputs, such as to support a set of internal functions or capabilities of the platform 102 and / or to support one or more functions or capabilities of configured stakeholder energy edge solutions 108. As one example among many, a service may be configured to obtain event data from IoT devices with cameras or sensors monitoring power generators and integrate it with weather data from public data resources 162 to provide a weather-related timeline of the power generator's energy generation data, which may in turn be consumed by the stakeholder energy edge solution 108, such as to help predict day-ahead energy generation by the power generator based on day-ahead weather forecasts. Such a wide range of configured data and intelligence modules and services 118 may be enabled by the platform 102 and represent, for example, various outputs consisting of the fusion or combination of the wide range of energy edge data sources handled by the platform, higher level analytical outputs resulting from expert analysis of the data, predictions and forecasts based on patterns in the data, automation and control outputs, and many other outputs. Energy trading support system

[0038] The configurable data and intelligence modules and services 118 may include an energy trade-enabling system 144. The trade-enabling system 144 may include a set of smart contracts that may operate on data stored in a set of distributed ledgers or blockchains, recording energy-related trade events such as, for example, energy purchases and sales (spot, forward, and peer-to-peer markets, as well as direct counterparty transactions) and associated service charges, trade-related energy events such as consumption, generation, distribution, and / or storage events, and other trade-related events often associated with energy, such as carbon production or reduction events, renewable energy credit events, pollution production or reduction events, etc. A set of smart contracts may consume any of the data types and entities described throughout this disclosure as a set of inputs, undertake a set of computations (optionally configured with flows that incorporate inputs from disparate systems in multi-stage transactions), and provide a set of outputs that enable the trade to be completed, reported (optionally recorded on a set of distributed ledgers), etc. The energy transaction enablement system 144 may be enabled or augmented with artificial intelligence, including autonomously discovering, structuring, and executing transactions according to a strategy and / or providing automated or semi-automated transactions based on training and / or oversight by a set of transaction experts. Autonomy and / or automation (supervised or semi-supervised) may be enabled by robotic process automation, such as by training a collection of intelligent agents in the interaction of transaction discovery, structuring, or execution with a group of transaction experts and a transaction enablement system (such as a software system used to configure and execute energy trading activity).

[0039] As energy becomes produced and consumed in local, decentralized markets, energy markets will likely follow the pattern of other peer-to-peer and shared economy markets, such as ride-sharing, apartment-sharing, and second-hand goods markets. Technology will enable businesses to bypass top-down, centralized energy supply and create platforms that allow operators to manage and monetize excess capacity through leasing and trading of assets and output.

[0040] As a more decentralized or peer-to-peer tradable energy market develops, platform 102 may include, link, integrate, or enable systems to other platforms that facilitate P2P trading, wholesale contracts, renewable energy certificate (REC) tracking, and broader decentralized energy supply, payment management, and other trading elements. In embodiments, the above may use a blockchain, distributed ledger, and / or smart contract system 132.

[0041] Specifically, increased transparency, choice and flexibility will enable consumers to actively participate in energy markets by generating, storing and selling electricity, rather than just consuming it.

[0042] In an embodiment, the transaction element may be configured by the energy transaction enablement system 144 to optimize energy generation, storage, or consumption, such as utility time-of-use rates. An IoT-based platform that can identify times of day when energy costs are lowest shifts energy demand away from high-price times. Stakeholders Energy Digital Twin

[0043] The configurable data and intelligence modules and services 118 may include one or more stakeholder energy digital twins 148, which, in embodiments, may include a set of digital twins configured to represent a set of energy-related stakeholder entities, including stakeholder-owned and stakeholder-operated energy generation resources, energy distribution resources, and / or energy distribution resources (including representing them by type, such as representing renewable energy systems, carbon generation systems, and others); stakeholder information technology and network infrastructure entities (e.g., edge and IoT devices and systems, networking systems, data centers, cloud data systems, on-premise information technology systems, etc.), energy-intensive stakeholder production facilities such as machinery and systems used in manufacturing, stakeholder transportation systems, market conditions (e.g., with respect to current and future market pricing for energy, stakeholder supply chains, stakeholder products and services, etc.), etc. The digital twin 148 can provide real-time information about the status, operating conditions, etc., particularly related to energy consumption, generation, storage, or distribution, such as provided sensor data, event logs, and other information streams from IoT and edge devices.

[0044] A stakeholder energy digital twin148 can provide a visual, real-time view of the energy impact of all aspects of an enterprise. Digital twins can also be role-based, providing visual and analytical metrics appropriate to the user's role, such as financial reporting information for a CFO, operating parameters information for a power plant manager, or energy market information for an energy trader. Data Integration Microservices

[0045] The configurable data and intelligence modules and services 118 may include configurable data integration microservices 150 organized in a service-oriented architecture such that various microservices may be grouped in series, in parallel, or in more complex flows to create higher-level, more complex services that each provide a defined set of outputs by processing a defined set of outputs, such as to enable a specific stakeholder solution 108 or to facilitate an AI-based orchestration, optimization, and / or automation system 114. The configurable data and intelligence modules and services 118 may be composed of, but not limited to, various functions and capabilities of the intelligent data layer 130, which operates on various data resources, such as internal event logs, outputs, data streams, etc. of the energy edge orchestration 110 and / or platform 102. Figure 2A-2B: Introduction of the main subsystems that make up the ecosystem Data Resources for Energy Edge Orchestration

[0046] Referring to FIG. 2A , data resources for energy edge orchestration 110 may include a set of edge and IoT networking systems 160, a set of public data resources 162, and / or a set of enterprise data resources 168, which, in embodiments, may use or be enabled by an adaptive energy data pipeline 164 that automatically handles the data processing, filtering, compression, storage, routing, transport, error correction, security, extraction, conversion, loading, normalization, cleansing, and / or other data processing capabilities involved in transporting data over a network or communication system. This may include, but is not limited to, based on data content (e.g., by packet inspection or other mechanisms for understanding the same), based on network conditions (e.g., congestion, delay / latency, packet loss, error rates, cost of transport, Quality of Service (QoS), etc.), based on the context of use (e.g., based on the user, system, use case, application, etc. (including based on similar prioritization), based on market factors (e.g., price or cost factors), based on user configuration, or other factors, as well as various combinations of the same. For example, among many other factors, the least cost route may be automatically selected for data related to managing low priority energy uses such as heating a swimming pool, while the fastest or highest QoS route may be selected for data supporting high priority uses or energy such as supporting critical medical infrastructure.

[0047] 2B, the platform 102 and orchestration may include, integrate, link, integrate, use, create, or otherwise handle a wide range of data resources for advanced energy resources and systems 104, composed stakeholder energy edge solutions 108, and / or energy edge orchestration 110. In embodiments, elements of the advanced energy resources and systems 104, composed stakeholder energy edge solutions 108, and / or energy edge orchestration 110 may be the same as, similar to, or different from corresponding elements shown in FIG. 1. The data resources 110 may include separate databases, distributed databases, and / or federated data resources, among many other things. Edge and IoT Networking Systems

[0048] A wide range of energy-related data may be collected and processed (including by artificial intelligence services and other functions) by a set of edge and IoT networking systems 160, and control instructions may be processed by devices, components, or systems integrated into them, located on IoT devices and systems, located on edge devices and systems, or located on or near energy-related entities, such as those involved in the generation, storage, delivery, or use of energy, including those used by consumers or businesses. These include any of the wide range of software, data, and network systems described herein. Public Data Resources

[0049] In embodiments, the platform 102 can track various public data resources 162, such as weather data. Weather conditions can affect energy use, particularly related to HVAC systems. By collecting, compiling, and analyzing weather data in conjunction with other building information, building managers can be proactive about HVAC energy consumption. The extensive public data resources 162 can include satellite data, demographic and psychographic data, population data, census data, market data, website data, e-commerce data, and many other types. Enterprise Data Resources

[0050] Enterprise data resources 168 include a wide range of enterprise resources, such as enterprise resource planning data, sales and marketing data, financial planning data, accounting data, tax data, customer relationship management data, demand planning data, supply chain data, procurement data, pricing data, customer data, product data, and sales data. Subsystems and modules of advanced energy resource systems

[0051] In an embodiment, advanced energy resources and systems 104 can include distributed energy resources 128, or "DERs" 128. More distributed energy resources mean that more individuals, networked groups, and energy communities will be able to generate and share their own energy and coordinate systems to achieve ultimate benefits. DERs 128 can be small- or medium-sized units of power generation and / or storage that operate locally, or they can be connected to a larger power grid at the distribution level. That is, DER systems 128 can be connected to a local power grid or can be isolated from the grid for standalone applications. Transforming Energy Infrastructure

[0052] The advanced energy resources and systems 104 orchestrated by the platform 102 may include a transformed energy infrastructure 120. The energy edge involves increasing digitization of generation, transmission, substation, and distribution assets, resulting in the operation, maintenance, and expansion of legacy grid infrastructure. In embodiments, a set of transformed energy infrastructure systems 120 may be integrated with or linked to the platform 102. The transition to an improved infrastructure may include a transition from SCADA systems and other existing control, automation, and monitoring systems to an IoT platform with advanced capabilities.

[0053] In an embodiment, new assets (e.g., DERs 128) that are added to or coordinated with the grid may be compatible with the existing infrastructure to maintain voltage, frequency, and phase synchronization.

[0054] Improvements to legacy grid assets, new grid-connected equipment, and support systems may, among other things, comply with regulatory standards from NERC, FERC, NIST, and other relevant authorities; positively impact grid reliability; make the grid less susceptible to cyberattacks and other security threats; increase the grid's ability to accommodate broader bidirectional flows of energy (e.g., DER penetration); and provide interoperability with technologies that improve grid efficiency (e.g., providing or facilitating demand response, reducing grid congestion).

[0055] Digitalization of legacy grid assets can relate to assets used for generating, transmitting, storing, and distributing electricity, such as power plants, substations, and transmission lines.

[0056] In an embodiment, to maintain and improve the existing energy infrastructure, the platform 102 may include a variety of capabilities, including fully integrated predictive maintenance across utility-owned assets (i.e., generation, transmission, substations, and distribution), smart (AI / ML-based) outage detection and response, and / or smart (AI / ML-based) load forecasting, including optional integration of DERs 128 with the existing grid.

[0057] In embodiments, proactive maintenance can be provided for power grids. Proactive maintenance allows utilities to accurately detect defects and reduce unplanned outages to better serve their customers. AI systems deployed with IoT and / or edge computing can help monitor energy assets and reduce maintenance costs. Digitized Resources

[0058] In embodiments, platform 102 can take advantage of the digital transformation of a wide range of digitized resources. Machines are becoming smarter, and software intelligence is being embedded into every aspect of business, helping to drive new levels of operational efficiency and innovation. Digital transformation is also underway, with the increasing presence of smart devices and systems capable of processing and communicating data, and near-ubiquitous sensors at the edge, IoT, and other devices generating large, dense data streams. All of this presents opportunities for increased intelligence, automation, optimization, and agility as information continuously flows between the physical and digital worlds. These devices and systems require large amounts of energy. For example, data centers consume large amounts of energy, and edge and IoT devices may be deployed in off-grid environments that require alternative forms of energy generation, storage, or movement. In embodiments, a set of digitized resources may be integrated, accessed, or used for energy optimization for compute, storage, and other resources in data centers and at the edge, among other locations. In an embodiment, as more devices incorporate sensors and controls, machines will "talk" to each other, enabling a continuous flow of information between the physical and digital worlds. Products can be tracked from source to customer or while in use, enabling rapid response to internal and external changes. Those responsible for managing and regulating such systems will be able to obtain detailed data from these machines to optimize the operation of the entire process. This trend will transform big data into smart data, enabling significant cost and process efficiencies.

[0059] In an embodiment, advances in digital technology have enabled a level of monitoring and operational performance that was previously impossible. Thanks to sensors and other smart assets, service providers can collect a wide range of data across multiple parameters and monitor in real time, 24 hours a day.

[0060] In embodiments, the DER128 is integrated into computing networks and infrastructure devices and systems to augment existing power grids, helping to reduce costs and improve reliability. Mobile Energy Resources

[0061] In embodiments, DERs may be integrated with mobile energy resources 124, such as electric vehicles (EVs) and their charging networks / infrastructure, thereby augmenting the existing power grid and helping to reduce costs and improve reliability. Given the rise of EVs (of all types), charging infrastructure and vehicle charging plans need to be optimized to match supply and demand. Furthermore, the growing demand for electricity and the development of EV infrastructure require optimization using related technologies, such as edge technology and IoT. Charging for electric vehicles may be integrated into distributed infrastructure, such as adding to the grid, such as bidirectional charging stations, or even used as DERs 128 by locally powering another system. Vehicle power electronics systems and batteries can benefit the power grid by providing system and grid services. Excess energy can be stored in the vehicle as needed and discharged when needed. This flexibility option not only avoids expensive load peaks during short-term high energy demand periods, but also increases the use of renewable energy.

[0062] In embodiments, integration with various other standardized communication protocols is required to universally integrate electric vehicles and charging infrastructure into the electric grid. The AI-based energy edge platform 102 may include, integrate, and / or link a set of communication protocols that enable management, provisioning, governance, control, etc. of energy edge devices and systems using such protocols. Energy Edge Solutions for Consistent Stakeholders

[0063] The set of configured stakeholder energy edge solutions 108 may include, among others, a set of mobility demand solutions 152, a set of enterprise optimization solutions 154, a set of energy supply and governance solutions 156, and / or a set of localized production solutions 158, which use various advanced energy resources and systems 104 and / or various configurable data and intelligence modules and services 118 to enable benefits to specific stakeholders, such as private companies, non-governmental organizations, independent service organizations, and government organizations. All such solutions may leverage edge intelligence, such as using data collected from on-board or integrated sensors, IoT systems, and edge devices located in proximity to entities that generate, store, supply, and / or use energy to feed models, expert systems, analytics systems, data services, intelligent agents, robotic process automation systems, and other artificial intelligence systems, to drive solutions to specific stakeholder needs. Enterprise Optimization Solutions

[0064] In embodiments, DER128 integrates with enterprise and shared resources to augment existing power grids, helping to reduce costs and improve reliability. Increasing levels of digitalization facilitate the integration of activities and new ways to optimize energy across buildings / operations, campuses, and enterprises. In embodiments, commercial enterprises can increase operational profits by leveraging big data and plug load analytics to efficiently manage buildings.

[0065] In embodiments, IoT sensors and building automation control systems may be configured to help optimize floor space, identify unused equipment, automate efficient energy consumption, improve safety, and reduce a building's environmental impact.

[0066] In embodiments, the platform 102 can manage the total energy consumption of systems and devices connected to an electrical network or set of DERs 128. Some systems may be running almost all the time, while other devices and machines may only be connected occasionally. By understanding both the total daily electrical consumption of a building and the role that individual devices play in the overall energy use of a particular system, the platform can, optionally through AI or algorithms, predict, provide, manage, and control the total consumption.

[0067] In embodiments, the platform 102 can track and leverage an understanding of occupant behavior. Resident activity levels, behavioral patterns, and comfort preferences can be considerations for energy efficiency measures. This includes tracking various cyclical or seasonal factors. Over time, a building's energy generation, storage, and / or consumption may follow predictable patterns that the IoT-based analytics platform can consider when generating proposed solutions.

[0068] In embodiments, the platform may enable or integrate with systems or platforms for autonomous operation. For example, industrial sites such as oil rigs and power plants require extensive monitoring for efficiency and safety, as liquid, steam, or oil leaks can lead to catastrophe, costs, and waste. AI and machine learning may provide autonomous capabilities to power plants, such as those provided by edge devices, IoT devices, and on-site cameras and sensors. Models may be deployed at the power plant edge or on DERs128 to use real-time inference and pattern detection to identify faults such as electrical leakage, shaking, and stress. Operators can use computer vision, deep learning, and intelligent video analytics (IVA) to monitor heavy equipment, detect potential hazards, and provide real-time alerts to protect the health and safety of workers, prevent accidents, and assign repair technicians for maintenance.

[0069] In embodiments, the platform may integrate with systems or platforms for pipeline optimization. For example, oil and gas companies may depend on finding the optimal route for transporting oil to refineries and ultimately fuel stations. Edge AI can calculate optimal oil flows to ensure production reliability and protect long-term pipeline health. In embodiments, companies can inspect pipelines for defects that could lead to dangerous failures and automatically alert pipeline operators. Energy Supply and Governance Solutions

[0070] Energy supply and governance solutions156 may include solutions for the governance of mining operations. Cobalt, nickel, and other metals are fundamental components of the batteries needed for the green electric vehicle revolution. Much of this mining occurs in areas with a long history of corruption, child labor, and violence, such as the Democratic Republic of the Congo. Companies are developing areas like Greenland for cobalt mining, in part because they can rely on reliable labor law enforcement, tax compliance, and other commitments. These commitments can be made with greater confidence locally and in other jurisdictions through one or more mining governance solutions542. Mining government solutions542 include mine-level IoT sensing of the mine environment, ground-penetrating sensing of unmined portions, mass spectrometry and computer vision-based sensing of mined materials, asset tagging of smart containers (e.g., detecting and recording opening and closing events to ensure material placed in a container is the same material released at the end point), wearable devices to detect the physiological state of miners, secure (e.g., blockchain and DLT-based) recording and resolution of transactions and transaction-related events, smart contracts for automatic revenue allocation (to tax authorities, labor, etc.), and automated systems for recording, reporting, and assessing compliance with contractual, regulatory, and legal policy requirements. All of the above, from base sensors to compliance reports, can optionally be represented in a digital twin representing each mine's owner or operated by the company.

[0071] The energy supply and governance solutions 156 may also include a set of carbon-aware energy solutions in which control of carbon-producing (or capturing) operators is managed by collecting data via edge and IoT devices about current carbon generation or emission conditions, and by automatically generating a set of recommendations and or control instructions to control the operators to meet policy, such as keeping operations within a range that can be offset by available carbon offset credits.

[0072] The various energy supply and governance solutions156 are detailed below. Local Production Solutions

[0073] In an embodiment, a set of localized production systems 158 may be integrated with, linked to, or managed by platform 102 to be able to meet localized production demands, particularly for goods that are very costly to transport (e.g., food) or services where the cost of energy distribution has a significant negative impact on the margins of the product or service (e.g., where intensive computing is required in locations where the electrical grid is non-existent, has insufficient capacity, is unreliable, or is too costly).

[0074] In embodiments, the power management system may converge with other systems, such as building management systems, operations management systems, production systems, service systems, data centers, etc., to enable enterprise-wide energy management. Figure 3: Details of a distributed energy generation system

[0075] 3, the distributed energy generation system 302 may include wind turbines, photovoltaic (PV), flexible and / or floating solar systems, fuel cells, modular nuclear reactors, nuclear batteries, modular hydroelectric systems, microturbines and turbine arrays, reciprocating engines, combustion turbines, and cogeneration plants, among others. The distributed energy storage system 304 may include battery-stored energy (including chemical batteries, etc.), molten salt energy storage, electrical thermal energy storage (ETES), gravity-based storage, compressed fluid energy storage, pumped hydro storage (PHES), and liquid air energy storage (LAES), among others. The DER system 128 may be managed by the platform 102. In an embodiment, the distributed energy storage system 304 may be portable so that units of energy can be transported to points of use, including points of use that are not connected to a conventional grid or where the conventional grid does not fully meet demand (e.g., where greater peak power, more reliable continuous power, or other capacity is required). Management also includes integrating, coordinating, and maximizing return on investment (ROI) of distributed energy resources (DERs).

[0076] In embodiments, DERs 128 may use a variety of distributed energy transmission methods and systems 308 having a variety of energy transmission capabilities, including power lines (e.g., conventional grid and building infrastructure), wireless energy transmission (including through coupling between high-Q resonators, resonant transfer, near-field energy transfer, and other methods), fluid transport, batteries, fuel cells, small nuclear systems, and the like).

[0077] Mobile energy resources 124 include a wide range of resources for the generation, storage, or delivery of energy at various scales, and accordingly, mobile energy resources 124 may constitute a subcategory of distributed energy resources 128 that have mobile attributes, such as when the mobile energy resource 124 is a vehicle 310 (e.g., an electric vehicle, a hybrid electric vehicle, a hydrogen fuel cell vehicle, etc., and in embodiments including a set of autonomous vehicles that may be unmanned autonomous vehicles (UAVs), drones, etc.), when the resource is integrated into or used by a mobile electronic device 312 or other mobile system, when the mobile energy resource 124 is a portable resource 314 (including when it is detachable and replaceable from a vehicle or other system), etc. As the capacity and availability of mobile energy resources 124 and their supporting infrastructure (e.g., charging stations) expand, orchestration of mobile energy resources 124 and other DERs 128, optionally in coordination with available grid resources, becomes increasingly important.

[0078] Resources involved in energy generation, storage, and transmission are increasingly digitized. These digitized resources 122 may include smart resources 318 (such as smart devices (e.g., thermostats), smart home devices (e.g., speakers), smart buildings, smart wearable devices, and many others enabled with processors, network connectivity, intelligent agents, and other on-board intelligence capabilities), where the intelligence capabilities of the smart resources 318 may be used for energy orchestration, optimization, autonomy, control, etc., and / or to provide data for artificial intelligence and analytics related to the foregoing. Digitization resources 122 can also include IoT and edge digitization resources 320, in which sensors or other data collectors (such as those monitoring event logs, network packets, network traffic patterns, location patterns of network devices, or other available data) provide additional energy-related intelligence related to energy generation, storage, transmission, or consumption by legacy infrastructure systems and devices, from large-scale generators and transformers to consumer or business devices, home appliances, and other systems in proximity to a set of IoT or edge devices that can monitor the same. Thus, IoT and edge devices can provide digital information about the energy status and flow of such devices and systems, regardless of whether the devices and systems have onboard intelligence capabilities. For example, among many others, IoT devices can deploy current sensors on power lines to home appliances to detect usage patterns, or edge networking devices can detect whether other devices or systems connected to the device are in use (and in what state) by monitoring network traffic from those other devices.Digitized resources 122 may also include cloud aggregation resources 322 related to energy generation, storage, transmission, or usage, such as by aggregating data across a fleet of similar resources owned or operated by an enterprise, used in connection with a defined workflow or activity, etc. Cloud aggregation resources 322 can consume data from various data resources 110, from crowdsourcing, from sensor data collection, from edge device data collection, and from many other sources.

[0079] In embodiments, digitized resources 122 may be used for a wide range of applications that involve or benefit from real-time information regarding the attributes, status, or flow of energy generation, storage, transmission, or consumption, such as to enable digital twins, such as adaptive energy digital twin systems 134 and / or stakeholder energy digital twins 148, and for variously configured stakeholder energy edge solutions 108.

[0080] Energy generation, storage, and consumption, particularly as it relates to green or renewable energy, have been the subject of intensive research and development in recent decades, resulting in higher peak power generation capacity, increased storage capacity, reduced size and weight, improved intelligence and autonomy, and many other achievements. Advanced energy resources and systems 104 may include a wide range of advanced energy infrastructure systems and devices resulting from a combination of features and capabilities. In embodiments, a set of flexible hybrid energy systems 324 may be provided that are adaptable to meet various energy consumption requirements, such as those capable of providing multiple types of energy (e.g., solar or wind power) to meet baseline requirements for off-grid operation, along with nuclear batteries to meet the much higher peak power requirements for temporary, resource-intensive activities, such as drilling operations in a mine or running periodic large factory machinery. A wide variety of such flexible hybrid energy systems 324 are contemplated herein, including those configured to modularly interconnect with various types of local production infrastructure, as described elsewhere herein. In embodiments, the advanced energy resources and systems 104 may include advanced energy generation systems that derive power from fluid flows, such as portable turbine arrays 328 that can be transported to consumption points proximate wind or water currents to replace or augment grid resources. The advanced energy resources and systems 104 may also include modular nuclear systems 330, including those configured to use nuclear batteries and those configured with mechanical, electrical, and data interfaces for interfacing with various consumption systems, including vehicles, localized production systems (as described elsewhere herein), smart buildings, and many others. The nuclear systems 330 may include SMRs and other nuclear reactor types.The advanced energy resources and systems 104 may include advanced batteries and advanced storage systems 332 including batteries with onboard intelligence for autonomous management, batteries with network connectivity for remote management, batteries with alternative chemistries (including green chemistries such as nickel-zinc), batteries made from alternative materials or structures (e.g., diamond batteries), batteries with built-in power generation capabilities (e.g., nuclear batteries), advanced fuel cells (e.g., cathode layer fuel cells, alkaline fuel cells, solid polymer electrolyte fuel cells, solid oxide fuel cells, and many others). Figure 4: Data resource details

[0081] 4, data resources for the energy edge orchestration 110 may include a wide range of public datasets, as well as private or proprietary datasets of an enterprise or individual. This may include datasets generated by or passing through the edge and IoT networking system 160, such as sensor data 402 (e.g., network data 404 (e.g., data about network traffic volume, latency, congestion, quality of service (QoS), packet loss, error rates, etc.); event data 408 (e.g., data from event logs of edge and IoT devices, data from event logs of enterprise operational assets, event logs of wearable devices, event data detected by inspection of traffic on application programming interfaces, event streams exposed by devices and systems, user interface interactions, etc.). such as, but not limited to, transactional events (e.g., captured by tracking clicks, eye tracking, etc.), user behavior events, transactional events (including financial transactions, database transactions, etc.), events in workflows (including directed flows, acyclic flows, iterative flows and / or looping flows, etc.); state data 410 (e.g., data indicative of the historical, current, or predicted / projected state of entities (e.g., machines, systems, devices, users, objects, individuals, and many others), including a wide range of attributes and parameters related to energy generation, storage, delivery, or utilization of such entities), and / or combinations of the foregoing (e.g., data indicative of the state of entities and the state of workflows involving entities).

[0082] In an embodiment, data resources may include, among other things, energy-related public data resources 162 such as energy grid data 422 (such as historical, current, and expected / projected maintenance status, operating status, energy production status, capacity, efficiency, or other attributes of energy grid assets involved in the generation, storage, or transmission of energy); energy market data 424 (such as historical, current, and forecast / projected pricing data for energy or energy-related entities, including spot market prices for energy based on location, type of consumption, type of generation, etc., day-ahead or other futures market prices for the same, cost of fuel, cost of raw materials involved (e.g., as follows)). g., costs of materials used in the manufacture of batteries), costs of energy-related activities such as mineral mining, and many others); location and movement data 428 (e.g., data indicating the historical, current, and / or expected / projected locations or movements of groups of individuals (e.g., crowds attending large events such as concerts, festivals, sporting events, conventions, etc.); data indicating the historical, current, and / or expected / projected locations or movements of vehicles (such as those used to transport people, goods, fuel, materials, etc.); data indicating the historical, current, and / or expected / projected locations or movements of resource production and / or demand points; and weather and climate data 430 (e.g., data indicating historical, current, and / or expected / projected locations or movements of energy-related weather patterns, including temperature data, precipitation data, cloud cover data, humidity data, wind speed data, wind direction data, storm data, barometric pressure data, etc.).

[0083] In embodiments, data resources for the energy edge orchestration 110 may include enterprise data resources 168, which may include, among other things, energy-related financial and transactional data 432 (e.g., data indicating historical, current, and / or expected / projected state, event, or workflow data, including financial entities, assets, etc., such as data related to prices and / or costs of energy and / or goods and services, data related to transactions, data related to asset valuations, balance sheet data, accounting data, data related to profits or losses, data related to investments, interest rate data, data related to debt and equity financing, capitalization data, and many others); operational data 434 (e.g., data related to the production and and other data describing the historical, current, and / or expected / projected state or flow of operating entities, such as those relating to the operation of assets and systems used in the performance of services, those relating to the movement of individuals, equipment, vehicles, machinery and systems, those relating to maintenance and repair work, and many others; human resources data 438 (such as those describing the historical, current, and / or expected / projected state, activity, location, or movement of a company's personnel); and sales and marketing data 440 (such as those describing the historical, current, and / or expected / projected state or activity of customers, advertising data, promotional data, loyalty program data, customer behavior data, demand planning data, pricing data, and many others); and others.

[0084] In embodiments, data resources for energy edge optimization 110 can be handled by an adaptive energy data pipeline 164, which can leverage the artificial intelligence capabilities of platform 102 to optimize the handling of various data resources. Increases in device processing power and storage capacity, combined with widespread deployment of edge and IoT devices, are creating a massive increase in the scale and granularity of available data of many of the types described herein. Thus, even more powerful networks like 5G and the anticipated 6G will likely struggle to transmit the volume of available data without issues of congestion, latency, errors, and reduced QoS. The adaptive energy edge data pipeline 164 can include a set of artificial intelligence capabilities to adapt the pipeline of data resources 110 to enable more effective orchestration of energy-related activities, such as by optimizing various elements of data transmission in coordination with energy orchestration needs. In embodiments, the adaptive energy data pipeline 164 may include self-organizing data storage 412 (e.g., edge, IoT, or other networking devices, cloud or data center systems, on-premise systems, etc.) based on patterns or attributes of the data (e.g., patterns of data volume over time or other metrics), the content of the data, the context of the data (e.g., whether the data is related to high-stakes business activities), etc.In an embodiment, the adaptive energy data pipeline 164 provides automated adaptive networking 414 (e.g., adaptive routing based on network path conditions (including packet loss, error rate, QoS, congestion, cost / pricing, etc.)), adaptive protocol selection (e.g., selecting among transport layer protocols (e.g., TCP or UDP), etc.), adaptive routing based on RF conditions (e.g., adaptively selecting between available RF networks (e.g., Bluetooth, Zigbee, NFC, etc.)), adaptive filtering of data (e.g., DSP-based filtering of data based on knowledge of whether a device is authorized to use RF capabilities), adaptive slicing of network bandwidth, adaptive use of cognitive and / or peer-to-peer network capacity, etc. In embodiments, the adaptive energy data pipeline 164 may adapt the data based on context (e.g., an enterprise operational context (e.g., distinguishing between mission-critical and less critical operations, distinguishing between time-sensitive and other operations, distinguishing between contexts required for policy or legal compliance, etc.), a transactional or financial context (e.g., based on, for example, whether the data is required based on a contractual requirement, whether the data is useful or necessary for real-time trading or financial gain (e.g., time-sensitive arbitrage opportunities or damage mitigation needs)), and many others). In embodiments, the adaptive energy data pipeline 164 may include market-based adaptation 420, such that storage, networking, or other adaptations are based on historical, current, and / or anticipated / projected market factors (e.g., the cost of storing, transmitting, and / or processing the data (including the cost of the energy used therein), the price, cost, and / or marginal profit of goods or services produced based on the data, and many others).

[0085] In an embodiment, the adaptive energy data pipeline 164 can adapt all aspects of data handling, including storage, routing, transmission, error correction, timing, security, extraction, conversion, loading, cleansing, normalization, filtering, compression, protocol selection (including physical layer, media access control layer, and application layer protocol selection), encoding, decoding, and more. Figure 5: Details of the configured Energy Edge stakeholder solutions Local production

[0086] Referring to FIG. 5 , the platform 102 can orchestrate the various services and functions described to form a set of configured stakeholder energy edge solutions 108, including mobility demand solutions 152, enterprise optimization solutions 154, localized production solutions 158, and energy supply and governance solutions 108.

[0087] A set of localized production solutions 158 could be any of a number of different applications, including data center operations (e.g., to support high-frequency trading operations that require low latency and benefit from proximity to market and exchange computing systems), operations using quantum computing, operations using very large neural networks or computationally intensive artificial intelligence solutions (e.g., encoding and decoding systems used in cryptography), operations involving complex optimization solutions (e.g., high-dimensional database operations, analytics, route optimization in computer networks, behavioral targeting in marketing, route optimization in transportation, etc.), operations supporting cryptocurrencies (e.g., mining operations in cryptocurrencies using proof-of-work or other computationally intensive approaches), operations where energy is supplied from local sources (e.g., hydroelectric dams, wind farms, etc.), and many others.

[0088] The set of localized production solutions 158 may include a set of transportation cost mitigation solutions 524, such as those where the cost of energy required to transport raw materials or finished goods to the point of sale or point of use is a significant component of the overall cost of the goods. The transportation cost mitigation solutions 524 may comprise a set of distributed energy resources 128 or other advanced energy resources 104 to provide energy supplementing or replacing traditional grid energy to enable localized production of goods traditionally produced in remote locations and transported to the point of sale or point of use via transportation and logistics networks (e.g., long-distance trucking). For example, moisture-rich crops may be produced locally, such as in containers equipped with lighting systems, hydration systems, etc., to shift the energy mix toward producing the crops rather than transporting the finished goods. The platform 102 may be used to optimize the combination of a set of localized modular energy generation or storage systems at the fleet level to support a set of localized production systems for heavy goods, such as by rotating energy generation or storage systems between the localized production systems to meet demand (e.g., seasonal demand, demand based on crop cycles, demand based on market cycles, etc.).

[0089] The set of local production solutions 158 can include a set of remote production operations solutions 528 for orchestrating distributed energy resources 128 or other advanced energy resources 104 to provide energy in a more optimal manner to remote operations, such as mineral mining operations, energy exploration operations, drilling operations, military operations, firefighting and other disaster response operations, forestry operations, and others where the local energy demand at a given time periodically exceeds what can be provided by the energy grid or where the energy grid is unavailable, including orchestrating the routing and provisioning of fleets of portable energy storage systems (e.g., vehicles, batteries), the routing and provisioning of fleets of portable renewable energy power generation systems (e.g., wind, solar, nuclear, hydroelectric), and the routing and provisioning of fuels (e.g., fuel cells).

[0090] A set of localized production solutions 158 may include a set of production assets (e.g., 3D printers, CNC machines, reactors, manufacturing systems, conveyors, and other components) configured to interface with a set of modular energy production systems to accept a combination of energy from the grid and energy from localized energy generation or storage sources, and the energy storage and generation systems configured to be modular, detachable, and portable between the production assets to provide grid augmentation or replacement at a fleet level without requiring dedicated energy assets for each production asset. Platform 102 may be used to configure and orchestrate the set of energy assets and the set of production assets to optimize localized production based on various factors noted herein, such as market conditions in the energy market and the market for the company's goods and services. Enterprise Optimization Solutions

[0091] The configured set of stakeholder energy edge solutions 108 may also include a set of enterprise optimization solutions 154, for example, providing companies with greater visibility into the role energy plays in their operations (e.g., enabling targeted strategic investments in energy-related assets); greater agility in structuring operations and transactions to meet operational and financial goals that are driven at least in part by energy availability, energy market prices, etc.; improved governance and control over energy-related factors such as carbon production, waste heat and pollution emissions; and improved efficiency in the use of energy at all scales of use, from electronic devices and smart buildings to factories and energy extraction activities. As used herein, the term "enterprise" includes, unless the context requires otherwise, private and public corporations, such as corporations, limited liability companies, partnerships, proprietorships, non-governmental organizations, for-profit organizations, non-profit organizations, public-private partnerships, military organizations, emergency response organizations (such as police, fire departments, and emergency medical services), private and public educational institutions (such as schools, colleges, and universities), governmental agencies (such as municipal, county, state, local, federal, national, and international), agencies (such as local, state, federal, national, international, cooperative, regulatory, environmental, energy, defense, civil rights, education, and many others), and others. Examples provided with respect to for-profit businesses should be understood to apply to other enterprises, and vice versa, unless the context excludes such application.

[0092] The enterprise optimization solutions 154 may include a set of smart building solutions 512, where the platform 102 may be used to orchestrate energy generation, transmission, storage and / or consumption across a set of buildings owned or operated by an enterprise, such as aggregating energy purchasing transactions across a fleet of smart buildings, providing a set of shared mobile or portable energy units across a fleet of smart buildings that are provided based on contextual factors such as usage requirements, weather, market prices, etc. in each building.

[0093] Enterprise optimization solutions 154 may include a set of smart energy delivery solutions 514, in which platform 102 may be used to orchestrate delivery or energy to operational points of use at advantageous costs and times. In embodiments, platform 102 may be used to time the routing of liquid fuels through pipeline elements by automatically controlling pipeline switching points based on contextual factors such as operational usage requirements, regulatory requirements, market prices, etc. In other embodiments, platform 102 may be used to orchestrate the routing of portable energy storage units or portable energy generation units to provide energy that augments or replaces grid energy capacity at the time and date of operational use. In embodiments, platform 102 may be used to orchestrate the routing and delivery of wireless power to provide energy at the time and date of use. Energy delivery optimization may be based on market prices (past, current, futures market, and / or forecast), operational conditions (current and forecast), policies (e.g., dictating the priority of certain uses), and many other factors.

[0094] Enterprise optimization solutions 154 may include a set of smart energy trading solutions 518, in which platform 102 can be used to orchestrate trading of energy or energy-related entities (e.g., renewable energy credits (RECs), pollution abatement credits, carbon reduction credits, etc.) across a fleet of enterprise assets and / or operations to optimize energy purchases and sales in coordination with energy-related operations at all scales of energy use. This may include, in embodiments, aggregating and timing current and futures market energy purchases across assets and operations, automatically configuring purchases of shared generation, storage, or delivery capacity for enterprise business uses, and the like. Platform 102 can leverage blockchain, smart contracts, and artificial intelligence capabilities trained as described throughout this disclosure to perform such activities based on the enterprise's operational needs, strategic objectives, and contextual factors, as well as external contextual factors such as market needs. Smart contracts can automatically secure future energy supply contracts to meet needs, either by purchasing grid-based energy from a provider or ordering portable energy storage units. Smart contracts may be configured with intelligence such as timing purchases based on market prices predicted by, for example, an intelligent agent, based on historical market prices and current contextual factors.

[0095] The enterprise optimization solutions 154 may include a set of enterprise energy digital twin solutions 520, where the platform 102 manages the assets of the enterprise involved in operations, the assets of external entities related to the enterprise's energy use or transactions (e.g., energy grid entities, pipelines, charging locations, etc.), and energy market entities (e.g., trading partners, smart contracts, blockchains, prices, etc.). A user of the set of enterprise energy digital twin solutions 520 may, for example, display a set of energy-consuming factories and be presented with a view showing the relative efficiency of each factory, individual machines within the factory, or machine components. In such an example, the digital twin may provide visual indicators of inefficient assets such as red flags, provide an ordered list of assets that would be most beneficial to replace, or provide recommendations that the user can accept (e.g., trigger a replacement order). Digital twins can be role-based, adaptive based on context or market conditions, personalized, augmented with artificial intelligence, and so on, in many ways described herein and in the documents incorporated by reference. Mobility Demand Solutions

[0096] With further reference to FIG. 5 , the set of configured stakeholder energy edge solutions 108 may include a set of mobility demand solutions 152 such that the platform 102 may be used to orchestrate energy generation, storage, delivery, or consumption by or for a set of mobile entities, such as a fleet of vehicles, a set of individuals, a set of mobile event production units, or a set of mobile factory units, among many others.

[0097] The set of mobility demand solutions 510 may include a set of transportation solutions 502, such as those used by the platform 102 to orchestrate energy generation, storage, delivery, or consumption by or for a set of vehicles, such as those used to transport goods, passengers, etc. The platform 102 can handle relevant operational and contextual data, such as those indicating transportation needs, priorities, etc., as well as relevant energy data, such as the energy costs used to transport entities using different transportation modes at different times, and can provide a set of transportation recommendations, or automated provisioning, to optimize transportation operations while fully considering energy costs and prices. For example, among other things, electric or hybrid passenger tour buses can be automatically routed to scenic spots in proximity to low-cost renewable energy charging stations so that tourists can charge the buses while experiencing the sites, meeting an energy-related objective (cost reduction) and an operational objective (customer satisfaction). The intelligent agent can be trained using the techniques described herein and in the documents incorporated by reference (such as by training the robotic process automation with a training set of expert interactions) to provide a set of recommendations for optimizing energy-related and other operational objectives.

[0098] The set of mobility demand solutions 510 may include a set of mobile user solutions 504 that the platform 102 may use to orchestrate energy generation, storage, delivery, and / or consumption by or for a set of mobile users, such as users of mobile devices. For example, in anticipation of a large, temporary surge in population in a location (such as a small city hosting a major sporting event), the platform 102 may provide a set of recommendations or automatically configure a set of orders for a set of portable charging units to support charging of consumer devices.

[0099] The set of mobility demand solutions 510 may include a set of mobile event production solutions 508 such that the platform 102 may be used to orchestrate energy generation, storage, delivery, or consumption by or for a set of mobile entities involved in the production of an event, such as a concert, sporting event, convention, circus, fair, revival, graduation, college reunion, festival, etc. This may include automatically configuring a set of energy generation, storage, or delivery units based on the operational configuration of the event (e.g., to meet the needs of lighting, food service, transportation, speakers and other audiovisual elements, machinery (e.g., 3D printers, video game consoles, etc.), vehicles, etc.), automatically configuring such operational configuration based on energy capabilities, configuring one or more of the energy or operational factors based on contextual factors (e.g., market prices, attendee demographics, etc.), etc.

[0100] The set of mobility demand solutions 510 can include a set of mobile factory solutions 510, for example, where the platform 102 can be used to orchestrate energy generation, storage, delivery, or consumption by or for a set of mobile factory entities. These can include container-based factories, such as those in which 3D printers, CNC machines, closed-environment agricultural systems, semiconductor fabricators, gene editing machines, biological or chemical reactors, furnaces, or other factory machines are integrated into or otherwise contained within shipping containers or other mobile factory housings, where the platform 102 can configure a set of recommendations or instructions for providing energy generation, storage, or delivery to meet the operational needs of the set of factory machines at a set of times and locations based on the operational needs of the set of factory machines. The configuration can be based on energy factors, operational factors, and / or contextual factors, such as market prices of commodities and energy, population needs (such as disaster recovery needs), and many other factors. Energy Supply and Governance Solutions

[0101] With further reference to FIG. 5 , the configured set of stakeholder energy edge solutions 108 includes a set of energy provisioning and governance solutions 156 that may be used by the platform 102 to orchestrate energy generation, storage, delivery, or consumption by or for a set of entities based on a set of policies, regulations, laws, etc., such as to facilitate compliance with corporate financial management policies, government or corporate policies regarding carbon reduction, and many other policies.

[0102] The set of energy supply and governance solutions 156 may include a set of carbon-aware energy edge solutions 532, such as a set of policies regarding carbon generation that may be discovered, configured, and implemented in the platform 102, such as requiring monitoring of energy generation by one or more assets or operations to track carbon generation or emissions, requiring offsets for such generation or emissions, etc. In embodiments, energy generation control instructions (e.g., for a machine or set of machines) may be configured with embedded policy instructions, such as requiring confirmation of available offsets before a machine is allowed to generate energy (and carbon) or before a machine can exceed a given amount of generation in a given period of time. In embodiments, the embedded policy instructions may include a set of override provisions (by a user or based on situational factors such as a declared emergency state) that allow policies to be overridden for mission-critical or emergency operations. Carbon generation, reduction, and offsets may be optimized across an enterprise's operations and assets, such as by intelligent agents trained in various ways as described elsewhere in this disclosure.

[0103] The set of energy supply and governance solutions 156 may include a set of automated energy policy deployment solutions 534, for example, where a user may interact with a user interface to design, develop, or configure a set of policies regarding energy generation, storage, supply, and / or utilization (e.g., by entering rules or parameters), which may be processed by the platform, for example, by presenting the policies to users interacting with entities that are subject to the policies (e.g., interfaces of such entities and / or digital twins of such entities, etc., and providing alerts regarding actions that are at risk of non-compliance, logging non-compliance events, recommending alternative compliance options, etc.); by embedding policies in the control systems of entities that generate, store, supply, or use energy (so that operations of such entities are controlled in a manner that complies with the policies); by embedding policies in smart contracts that enable energy-related transactions (so that transactions are automatically executed in compliance with the policies, so that warnings or alerts are provided in case of non-compliance, etc.); by setting policies that are automatically reconfigured based on contextual factors (such as operational and / or market factors), etc. In embodiments, intelligent agents may be trained, such as on training datasets of historical data, on feedback from outcomes, and / or on training datasets of human policy-setting interactions, to generate policies, configure or modify policies, and / or perform actions based on policies.A wide range of policies and configurations may be implemented, such as setting a maximum energy use for an entity for a period of time, setting a maximum energy cost for an entity for a period of time, setting a maximum carbon production for an entity for a period of time, setting a maximum pollution emission for an entity for a period of time, setting carbon offset requirements, setting renewable energy credit requirements, setting energy mix requirements (e.g., requiring a minimum percentage of renewable energy), setting minimum rates of return based on energy and other marginal costs for production entities, setting minimum storage baselines for energy storage entities (e.g., to provide a safety margin for disaster recovery), and many others.

[0104] The set of energy provisioning and governance solutions 156 may include a set of energy governance smart contract solutions 538 that enable users of the platform 102 to design, generate, configure, and / or deploy smart contracts that automatically provide a degree of governance of energy transactions, such that the smart contracts take operational, market, or other contextual inputs (e.g., energy usage information collected by edge devices about operational assets) as inputs and automatically configure a set of contracts that adhere to a set of policies regarding the purchase, sale, reservation, sharing, or other transactions of energy, energy-related credits, etc. For example, a smart contract may automatically aggregate carbon offset credits needed to balance detected carbon generation across a set of machines used in an enterprise's operations.

[0105] The set of energy supply and governance solutions 156 may include a set of automated energy financial control solutions 540 that enable users and / or intelligent agents of the platform 102 to design, generate, configure, or deploy policies related to controlling financial factors associated with energy generation, storage, delivery, and / or utilization. For example, a user may set a policy requiring a minimum marginal profit for a machine to continue operating, and the policy may be presented to a machine operator, manager, etc. As another example, a policy may be incorporated into a machine's control system that takes a set of inputs required to determine marginal profit (e.g., costs of inputs and other non-energy resources used in production, cost of energy, predicted energy required to produce output, and market price of output) and automatically determines whether and at what level to continue production in order to maintain marginal profit. Such policies may incorporate further inputs, such as those related to predicted market or customer behavior, such as based on the elasticity of demand for the associated output. Figure 6: Details of the Intelligence Realization System

[0106] Referring to FIG. 6 , further details are provided regarding an embodiment of the intelligence enabling system 112, including an intelligent data layer 130, a distributed ledger and smart contract system 132, an adaptive energy digital twin system 134, and an energy simulation system 136.

[0107] The intelligent data layer 130 can optionally autonomously, with user supervision, or semi-supervision, undertake any of a wide range of data processing capabilities noted throughout this disclosure and the documents incorporated herein by reference, including extraction, transformation, loading, normalization, cleansing, compression, routing selection, protocol selection, storage self-organization, filtering, transmission timing, encoding, decoding, and many others. The intelligent data layer 130 may include an energy generation data layer 602 (e.g., generating, automatically configuring, and routing streams or batches of data related to energy generation by a collection of entities, such as a company's operating assets), an energy storage data layer 604 (e.g., generating, automatically configuring, and routing streams or batches of data related to energy storage by a collection of entities, such as a company's operating assets or a collection of customers' assets), an energy delivery data layer 608 (e.g., generating, automatically configuring, and routing streams or batches of data related to energy delivery by a collection of entities, such as delivery via power lines, delivery via pipelines, delivery via portable energy storage, etc.), and an energy consumption data layer 610 (e.g., generating, automatically configuring, and routing streams or batches of data related to energy consumption by a collection of entities, such as a company's operating assets, a collection of customers, a collection of vehicles, etc.).

[0108] The distributed ledger and smart contract system 132 can provide a set of underlying functions to enable energy-related transactions, such as purchases, sales, leases, and futures contracts for energy generation, storage, delivery, or consumption, as well as related types of transactions, such as renewable energy credits, carbon reduction credits, pollution reduction credits, asset leasing, shared economy transactions for asset use, shared consumption contracts, bulk purchases, mobile resource provisioning, and many other transactions. This can include an energy trading blockchain 612 or a set of distributed ledgers for recording energy transactions, including generation, storage, delivery, and consumption transactions. A set of energy trading smart contracts 614 can operate on blockchain events and other input data to enable, configure, and execute the aforementioned types of transactions and other transactions. In embodiments, a set of energy trading intelligent agents 618 can be configured to design, generate, and deploy smart contracts 614; optimize trading parameters; automatically discover counterparties, arbitrage opportunities, etc.; recommend and / or automatically initiate steps toward offering or executing contracts; resolve contracts upon completion based on blockchain data; and perform many other functions.

[0109] The adaptive energy digital twin system 134 may include digital twins of energy-related entities, such as an enterprise's operational assets that generate, store, deliver, or consume energy, and may include an energy generation digital twin 622 (e.g., displaying content from an event log or stream or batch of data related to energy generation by a set of entities, such as an enterprise's operational assets), an energy storage digital twin 624 (e.g., displaying energy storage status information, usage patterns, etc., for a set of entities, such as an enterprise's operational assets or customer assets), an energy delivery digital twin 628 (e.g., displaying status data, events, workflow, etc., related to energy delivery by a set of entities, such as delivery by power lines, delivery by pipelines, delivery by portable energy storage, etc.), and an energy consumption digital twin 630 (e.g., displaying data related to energy consumption by a set of entities, such as an enterprise's operational assets, a set of customers, a set of vehicles, etc.). The adaptive energy digital twin system 134 may include various types of digital twins described throughout this disclosure and / or documents incorporated by reference herein, such as those fed by data streams from edge and IoT devices, those that adapt based on user role or context, those that adapt based on market context, those that adapt based on operational context, and many others.

[0110] The collection of energy simulation systems 136 may include a wide range of systems for simulating energy-related operations based on historical patterns, current states (including contextual information, operational information, market information, and other information), and predicted / projected states of entities involved in energy generation, storage, delivery, and / or consumption. This may include, among others, energy generation simulation 632, energy storage simulation 634, energy delivery simulation 638, and energy consumption simulation 640. Simulation systems 136 may employ a wide range of simulation capabilities, such as 3D visualization simulation of physical behavior, presentation of simulation output in a digital twin, generation of simulated financial results for a set of different operational scenarios, generation of simulated operational results, and many others. Simulations may be based on a set of models, such as models of the energy generation, storage, delivery, and / or consumption operations of a machine or system, or a fleet of machines or systems (potentially aggregated based on underlying models and / or based on projections from a subset of models into a larger set). The model can be iteratively improved by feedback of results from operations and / or by feedback comparing model-based predictions with actual results and / or predictions by other models or human experts. Simulations can be performed using stochastic methods, random walk or random forest algorithms, predicting trends from historical data about current conditions, etc. Simulations can be based on behavioral models, such as models of corporate or individual behavior based on various factors, such as past behavior, economic factors (e.g., elasticity of demand or supply to price changes), energy usage models, etc. Simulations may use predictions from artificial intelligence, including artificial intelligence trained by machine learning (including deep learning, supervised learning, semi-supervised learning, etc.).Simulations may be configured for presentation in augmented reality, virtual reality, and / or mixed reality interfaces and systems (collectively referred to as "XR") to allow users to interact with aspects of the simulation to train in controlling machines, setting policies, managing a factory or other entity containing multiple machines, handling a fleet of machines or factories, etc. As an example, a factory simulation may simulate the energy consumption of all machines in the factory and present other data (such as operating data, input costs, production costs, computational costs, market price data, and other content within the simulation). In the simulation, a user may configure the factory, such as by setting the output level of each machine, and the simulation may simulate the profitability of the factory under various simulated market conditions. Thus, a user may be trained to configure the factory under a variety of different market conditions. Figure 7: Details of an AI-based energy orchestration, optimization, and automation system

[0111] Referring to FIG. 7 , more details are provided regarding a set of AI-based energy orchestration, optimization, and automation systems 114, each of which may use various other capabilities, services, functions, modules, components, or other elements of the platform 102 to orchestrate energy-related entities, workflows, etc. on behalf of an enterprise or other user. Orchestration may use, for example, robotic process automation to facilitate the automatic orchestration of energy-related entities and resources based on training datasets and / or human oversight based on historical human interaction data. As another example, orchestration may include the design, configuration, and deployment of a set of intelligent agents that can automatically orchestrate a set of energy-related workflows based on operational, market, contextual, and other inputs. Orchestration may include the design, configuration, and deployment of autonomous control systems, such as systems that control energy-related activities based on operational data collected by or from onboard sensors, edge devices, IoT devices, etc. Orchestration may include optimization, such as simulation-based multivariate decision optimization, optimization based on real-time inputs, etc. Orchestration may involve the use of artificial intelligence for pattern recognition, prediction, and forecasting, such as based on historical data sets or current conditions.

[0112] The collection of AI-based energy orchestration, optimization, and automation systems 114 may include, among others, a collection of energy generation orchestration systems 138, a collection of energy consumption orchestration systems 140, a collection of energy storage orchestration systems 142, a collection of energy market orchestration systems 146, and a collection of energy delivery orchestration systems 147.

[0113] The set of energy generation orchestration systems 138 may include, among other things, a set of generation timing orchestration systems 702 and a set of location orchestration systems 704. The set of timing orchestration systems 702 may orchestrate the timing of energy generation to ensure that the timing of generation meets mission-critical or operational needs, complies with policies and plans, is optimized to improve financial or operational metrics, and / or (in the case of energy generated for sale) is timed based on fluctuations in energy market prices. The orchestration of generation timing can be based on models, simulations, or machine learning on historical datasets. The orchestration of generation timing can be based on current conditions (operating, market, etc.).

[0114] The set of power generation location orchestration systems 704 can orchestrate the location of power generation assets, including mobile or portable power generation assets such as portable generators, solar systems, wind systems, and modular nuclear systems, as well as the selection of locations for large-scale fixed infrastructure power generation assets such as power plants, generators, and turbines, for example, to ensure that for any operating location, available generating capacity (baseline capacity and peak capacity) meets mission-critical or operational needs, complies with policies and plans, is optimized to improve financial or operational metrics, and / or (in the case of energy generated for sale) is appropriately located based on regional fluctuations in energy market prices. The orchestration of power generation locations can be based on models, simulations, or machine learning of historical datasets. The orchestration of power generation locations can be based on current conditions (operating, market, etc.).

[0115] The set of energy consumption orchestration systems 140 may include, among other things, a set of consumption timing optimization systems 718 and a set of operational prioritization systems 720. The set of consumption timing optimization systems 718 shift consumption of non-critical activities to lower cost energy resources (e.g., by shifting to off-peak hours to obtain lower electricity prices for grid energy consumption (e.g., renewable energy systems instead of the grid)), shift consumption to more profitable activities (e.g., shifting consumption to machines with higher marginal profit per time period based on current market and operating conditions (as detected by a combination of edge and IoT devices and market data sources)), etc.

[0116] A set of operational prioritization systems 720 may enable a user, intelligent agent, etc. to set operational priorities by rules or policies, by setting target metrics (e.g., efficiency, marginal profit production, etc.), by declaring mission-critical operations (e.g., safety, disaster recovery, emergency systems, etc.), by declaring priority among a set of operational assets or activities. In an embodiment, energy consumption orchestration may take input from operational prioritization to provide a set of recommendations or control instructions for optimizing energy consumption by a machine, component, set of machines, factory, or fleet of assets.

[0117] The set of energy storage orchestration systems 142 may include a set of storage location orchestration systems 708 and a set of safety margin orchestration systems 710. The set of storage location orchestration systems 708 coordinates the locations of storage assets, including mobile or portable power generation assets such as portable batteries, fuel cells, and nuclear storage systems, as well as large-scale arrays of batteries, fuel storage systems, thermal energy storage systems (e.g., using molten salt), and the like, to ensure that, for any operational location, available storage capacity meets mission-critical or operational needs, complies with policies and plans, is optimized to improve financial or operational metrics, and / or (in the case of energy stored and offered for sale) is well-positioned based on regional fluctuations in energy market prices. Storage location orchestration can be based on models, simulations, or machine learning on historical datasets, such as behavioral models that indicate usage patterns by individuals or businesses. Storage location orchestration can be based on current conditions (operational, market, etc.), as well as many other factors. For example, storage capacity can be brought to locations where grid capacity is offline or abnormally constrained (e.g., for disaster recovery).

[0118] The safety margin orchestration system 710 set may be used to orchestrate storage capacity to maintain a safety margin, such as a minimum amount of stored energy to power critical systems (e.g., life support systems, perimeter security systems, etc.) or high-priority systems (e.g., high-margin manufacturing) for a defined period in case of loss of critical energy capacity (e.g., due to a grid outage or brownout) or a lack of renewable energy production (e.g., insufficient wind, hydro, or solar power due to weather conditions, drought, etc.). The minimum amount may be set by rules or policies, or may be adaptively learned by an intelligent agent, etc., based on a training dataset of results and / or based on historical, current, and predicted conditions (e.g., climate and weather forecasts). The safety margin orchestration system 710, in embodiments, may take input from the energy supply and governance solution 156.

[0119] The set of energy market orchestration systems 146 may include a set of transaction aggregation systems 722 and futures market optimization systems 724 .

[0120] The suite of transaction aggregation system 722 system can automatically orchestrate a suite of energy-related transactions, such as purchases, sales, orders, futures contracts, hedging contracts, limit orders, stop-loss orders, etc., for energy generation, storage, delivery, or consumption, renewable energy credits, carbon reduction credits, pollution reduction credits, etc., to take advantage of volume discounts, lock in current or day-ahead prices when advantageous, enable fractional ownership by a set of owners, operators, or consumers of blocks of energy generation, storage, or delivery capacity, etc. For example, a company may aggregate energy purchases across a collection of assets in different jurisdictions. An intelligent agent aggregates the collection of energy purchases in futures markets across jurisdictions and represents the aggregated purchases in a centralized location, such as the company's operating digital twin.

[0121] The futures market optimization system 724 can automatically assemble a collection of futures market contracts, such as energy, renewable energy credits, carbon offset or reduction credits, and pollution reduction credits, based on predictions of an individual's or business's future energy needs. Predictions can be based on historical usage patterns, current operating conditions, current market conditions, anticipated operational needs, and the like. Predictions may be generated by an intelligent agent using predictive models and / or based on machine learning of outcomes, human output, human-labeled data, and the like. Predictions may be generated by deep learning, supervised learning, semi-supervised learning, and the like. Based on the predictions, the intelligent agent can design, configure, and execute a series of futures market transactions across various jurisdictions to meet the anticipated timing, location, and type of needs.

[0122] The set of energy delivery orchestration systems 147 may include a set of delivery routing orchestration systems 712 and a set of energy delivery type orchestration systems 714.

[0123] The set of energy delivery routing orchestration systems 712 can orchestrate the routing of energy deliveries using the various components, modules, facilities, services, functions, and other elements of platform 102, such as the location, timing, and type of need, available power generation and storage capacity at the location where energy is needed, available energy sources for routing (e.g., liquid fuels, portable energy generation systems, portable energy storage systems, etc.), available routes (e.g., main pipelines, pipeline branches, power lines, wireless power transmission systems, transportation infrastructure (roads, railroads, waterways, etc.)), market factors (energy prices, commodity prices, profit margins for production activities, timing of events requiring energy, etc.), environmental factors (weather, etc.), operational priorities, etc. A set of artificial intelligence systems trained in the various ways disclosed herein can be trained to recommend or configure routes based on the foregoing inputs and sets of training data such as human routing activities, route optimization models, iterations between multiple simulated scenarios, or any combination of the foregoing, etc. For example, a set of control instructions may direct valves and other elements of an energy pipeline to supply large amounts of fluid-based energy to one location, while directing mobile or portable resources to another location where energy availability would otherwise be reduced, based on pipeline routing instructions.

[0124] The set of energy delivery type orchestration systems 714 uses the various components, modules, facilities, services, capabilities, and other elements of platform 102 to determine the location, timing, and type of needs, available power generation and storage capacity where energy is needed, available energy sources for routing (e.g., liquid fuels, portable energy generation systems, portable energy storage systems, etc.), available routes (main pipelines, pipeline branches, power transmission lines, wireless power transmission systems, transportation infrastructure (roads, railroads, waterways, etc.), etc.), market factors (energy prices, commodity prices, profit margins for production activities, timing of events requiring energy, etc.), environmental factors (weather, etc.), operational priorities, etc. A set of artificial intelligence systems trained in various ways disclosed herein can be trained to recommend or configure energy type combinations based on the foregoing inputs and a set of training data such as human type selection activities, delivery type optimization models, iterations between multiple simulated scenarios, or any combination of the foregoing. For example, a set of recommendations or control instructions may select a set of portable modular energy resources that fit the needs (e.g., specifying renewable sources if there is high storage capacity to meet operational needs so that inexpensive, intermittent sources are preferred), while the instructions may select more expensive natural gas energy if storage capacity is limited or nonexistent and usage is continuous (such as for a 24 / 7 data center operated remotely from the energy grid).

[0125] Many other examples of AI-based energy orchestration, optimization, and automation 114 are provided throughout this disclosure. Figure 8: Configurable Data Intelligence Modules and Services Details

[0126] 8 , the set of configurable data and intelligence modules and services 118 may include, among many other things, a set of energy tradable enablement systems 144, a set of stakeholder energy digital twins 148, and a set of data integration microservices 150. These data and intelligence modules may include various components, modules, services, subsystems, and other elements necessary to configure data streams or batches, configure intelligence to provide specific types of outputs, etc., to enable other elements of the platform 102 and / or various stakeholder solutions.

[0127] The set of energy trade enablement systems 144 may include, among other things, a set of counterparty and arbitrage discovery systems 802, a set of automated trade setup systems 804, and a set of energy investment and sale recommendation systems 808. The set of counterparty and arbitrage discovery systems 802 may be configured to operate on various data sources related to operational energy needs, contextual factors, and the energy markets, renewable energy credits, carbon offsets, pollution abatement credits, or other energy-related market offers by the set of counterparties to determine a recommendation or selection of a set of counterparties and offers. Intelligent agents in the counterparty and arbitrage discovery system 802 may initiate trades with the set of counterparties based on the recommendation or selection. Factors may include cost, counterparty credibility, size of the counterparty's offer, timing, location of energy needs, and many others.

[0128] A suite of automated trade setting systems 804 can recommend or automatically set terms for trades, either automatically or with human supervision, based on contextual factors (e.g., weather), past, current, or forecasted / predicted market data (e.g., related to energy prices, production costs, storage costs, etc.), timing and location of operational needs, and other factors. The automation may be through artificial intelligence, such as trained on human configuration interactions, trained by deep learning on outcomes, or trained by iterative improvement through a series of trials and adjustments (e.g., of neural network inputs and / or weights).

[0129] A set of energy investment and split recommendation systems 808 can automatically, or with human supervision, recommend or automatically configure the terms of investments or split transactions based on contextual factors (e.g., weather), historical, current, or forecast / predictive market data (e.g., related to energy prices, production costs, storage costs, etc.), the timing and location of operational needs, and other factors. The automation may be artificial intelligence, such as trained on human configuration interactions, trained by deep learning on outcomes, or trained by iterative improvement through a series of trials and adjustments (e.g., of neural network inputs and / or weights). For example, a set of energy investment and divestiture recommendation systems 808 may output a recommendation to invest in additional modular portable power generation units to support the location of planned energy exploration activities, or to divest a relatively inefficient plant where energy costs are predicted to produce a negative marginal return.

[0130] The set of stakeholder energy digital twins 148 may include, among many others, a set of financial energy digital twins 810, a set of operational energy digital twins 812, and a set of managerial energy digital twins 814. The set of financial energy digital twins 810 may represent a set of entities such as an enterprise's operating assets along with energy-related financial data such as, for example, the cost of energy used or projected to be used by a fleet of machines, components, plants, or assets, the price at which energy can be sold, the cost or price of renewable energy credits available through the use of renewable energy generating capacity, the cost or price of carbon offsets required to offset current and projected future operations, the cost of pollution abatement offsets or credits, etc. The financial energy digital twin 810 may integrate with other financial reporting systems and interfaces, such as an enterprise resource planning suite, a financial accounting suite, a tax system, etc.

[0131] A set of operational energy digital twins 812 can represent operational entities involved in generating, storing, delivering, or consuming energy, along with associated specification data, historical, current, or expected / projected operating conditions or parameters, and other information, allowing operators to view, for example, components, machines, systems, plants, and various combinations and sets thereof at an individual or aggregate level. The operational energy digital twin 812 can display energy and energy-related data relevant to operations, such as power generation, storage, delivery, consumption data, carbon production, pollution emissions, and waste heat production. A set of intelligent agents can provide alerts to the digital twin. The digital twin can automatically adapt by highlighting important changes, critical operations, maintenance, or replacement needs, etc. The operational energy digital twin 812 can obtain data from onboard sensors, IoT devices, and edge devices located at or near relevant operations to provide real-time, up-to-date data.

[0132] A set of executive energy digital twins 814 can display entities involved in generating, storing, delivering, or consuming energy, along with associated specification data, past, current, or expected / projected operating conditions or parameters, and other information, allowing executives to view key energy-driven performance metrics at an individual or aggregate level for components, machines, systems, plants, and various combinations and sets thereof. The executive energy digital twin 814 can display energy and energy-related data relevant to executive decision-making, such as generation, storage, delivery, and consumption data, carbon production, pollution emissions, waste heat production, etc., as well as financial performance data, competitive market data, etc. A set of intelligent agents can provide alerts configured to executive roles in the digital twin (e.g., financial data to the CFO, risk management data to the Chief Legal Officer, and aggregate performance data to the CEO or Chief Strategy Officer). The executive energy digital twin 814 can automatically adapt by highlighting important changes, critical operations, strategic opportunities, etc. The Executive Energy Digital Twin 814 captures data from on-board sensors, IoT devices and edge devices to provide real-time, up-to-date data.

[0133] The set of data integration microservices 150 may include a set of energy market data services 818, a set of operational data services 820, and a set of other contextual data services 822, among many others.

[0134] The energy market data services suite 818 may include market prices for a company's goods and services, feeds of past, current and / or future market energy prices in the company's operating jurisdictions (optionally weighted or ordered based on relative energy use across jurisdictions), a set of user or company preferences (e.g., displaying transactions related to a company's operational requirements or energy capacity), feeds of past, current, or future renewable energy credit prices, feeds of past, current, or future carbon offset prices, feeds of past, current, or future pollution reduction credit prices, and others.

[0135] A set of operational data services 820 may provide a composed, filtered, and / or otherwise processed feed of operational data, such as historical, current, and predicted / projected conditions and events of an enterprise's operational assets, collected by sensors, IoT devices, and / or edge devices and predicted or inferred based on the operation of a set of models, analytical systems, and / or artificial intelligence systems, such as intelligent predictive agents.

[0136] A set of other contextual data services 822 may provide feeds of a wide range of configured, filtered, or otherwise processed contextual data, such as weather data, user behavior data, population location data, demographic data, psychographic data, and many others.

[0137] Various types of configurable data integration microservices can provide a variety of configured outputs, such as batches, files, database reports, event logs, data streams, etc. Streams and feeds may be automatically generated and pushed to other systems, and services may be queried and / or pulled from sources (e.g., distributed databases, data lakes, etc.) or by application programming interfaces. Neural Network Example

[0138] Such neural networks have various nodes or neurons that can perform various functions on inputs, such as inputs received from sensors or other data sources, including other nodes. Functions include weights, features, feature vectors, etc. Neurons include perceptrons, neurons that mimic biological functions (such as human touch, vision, taste, hearing, and smell), etc. Continuous neurons, such as those with sigmoid activation, are used in the context of various forms of neural nets, such as when backpropagation is involved.

[0139] In many embodiments, an expert system or neural network may be trained, such as by a human operator or supervisor, or based on a data set, a model, or the like. Training may include presenting the neural network with one or more training data sets representing values, such as sensor data, event data, parameter data, and other types of data (including many types described throughout this disclosure), as well as one or more indicators of results, such as process results, computation results, event results, or activity results. Training may also include optimization training, such as training a neural network to optimize one or more systems based on one or more optimization approaches, such as a Bayesian approach, a parametric Bayesian classifier approach, a k-nearest neighbor classifier approach, an iterative approach, an interpolation approach, a Pareto optimization approach, or an algorithmic approach. Feedback may be provided in a variation-and-selection process, such as a genetic algorithm that evolves one or more solutions based on feedback through a series of rounds.

[0140] In embodiments, multiple neural networks may be deployed to a cloud platform that receives data streams and other inputs collected in one or more energy edge environments (e.g., by mobile data collectors) and transmits them to the cloud platform over one or more networks, including using network coding to provide efficient transmission. The cloud platform may optionally use massively parallel computing power to employ multiple different neural networks of various types (including modular, adaptive, hybrid, etc.) to perform prediction, classification, control functions, and provide other outputs, as described in connection with the expert systems disclosed throughout this disclosure. Different neural networks may be structured to compete with each other (optionally using evolutionary algorithms, genetic algorithms, etc.) so that an appropriate type of neural network, with an appropriate set of inputs, weights, node types, capabilities, etc., is selected by the expert system, etc., for the particular task involved in a given context, workflow, environmental process, system, etc.

[0141] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use feedforward neural networks. A feedforward neural network moves information unidirectionally from a data input, such as a data source related to at least one resource or parameter associated with a transaction environment, such as any of the data sources mentioned throughout this disclosure, through a series of neurons or nodes to an output. Data moves from the input node to the output node, optionally passing through one or more hidden nodes without loops. In embodiments, the feedforward neural network may be constructed with various types of units, such as binary McCulloch-Pitts neurons, the simplest of which are perceptrons.

[0142] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use capsule neural networks for prediction, classification, or control functions, etc., related to trading environments, such as in connection with one or more of the machines and automation systems described throughout this disclosure.

[0143] In embodiments, methods and systems described herein involving expert systems or self-organizing capabilities may use radial basis function (RBF) neural networks, which may be preferable in some situations involving interpolation in multidimensional space (e.g., where interpolation is useful in optimizing multidimensional functions, such as optimizing data marketplaces as described herein, optimizing the efficiency or output of power generation systems, factory systems, etc.). In embodiments, each neuron in an RBF neural network stores an example from a training set as a "prototype." The linearity involved in the functioning of this neural network provides RBFs with the advantage that they generally do not suffer from problems with local minima or maxima.

[0144] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities can use radial basis function (RBF) neural networks, such as those employing a distance criterion to the center (e.g., a Gaussian function). Radial basis functions may be applied as an alternative to hidden layers, such as sigmoid hidden layer transfer, in multilayer perceptrons. RBF networks may have two layers, with inputs mapped to each RBF in the hidden layer. In embodiments, the output layer is composed of a linear combination of hidden layer values, representing, for example, the average predicted output. The output layer values ​​can provide outputs that are the same as or similar to the outputs of regression models in statistics. In classification problems, the output layer is a sigmoid function of a linear combination of hidden layer values, representing posterior probabilities. In both cases, performance is often improved by shrinkage techniques such as ridge regression in classical statistics. In a Bayesian framework, this corresponds to a prior belief in small parameter values ​​(and therefore a smooth output function). RBF networks can avoid local minima because the only parameters adjusted in the learning process are the linear mapping from the hidden layer to the output layer. Linearity ensures that the error surface is quadratic and therefore has a single minimum. In regression problems, this can be found with a single matrix operation. In classification problems, the fixed nonlinearity introduced by a sigmoid output function can be handled using an iteratively reweighted least-squares function, etc.

[0145] RBF networks can use kernel methods such as support vector machines (SVMs) or Gaussian processes (where the RBF is the kernel function). Nonlinear kernel functions can be used to project the input data into a space where the learning problem can be solved using a linear model.

[0146] In embodiments, an RBF neural network may include an input layer, a hidden layer, and a summation layer. In the input layer, one neuron appears for each predictor variable. For categorical variables, Nl neurons are used, where N is the number of categories. In embodiments, the input neurons may standardize the range of values ​​by subtracting the median and dividing by the interquartile range. The input neurons send their values ​​to each neuron in the hidden layer. The hidden layer may use a variable number of neurons (determined by a learning process). Each neuron consists of a radial basis function centered at a point with the same dimensions as the number of predictor variables. The spread (e.g., radius) of the RBF function may be different in each dimension. The center and spread may be determined by training. When presented with a vector of input values ​​from the input layer, the hidden neuron calculates the Euclidean distance of the test case from the neuron's center point and then applies the RBF kernel function to this distance. The resulting value is passed to the summation layer. In the summation layer, the value coming out of a neuron in the hidden layer is multiplied by the weight associated with that neuron and then summed with the weighted values ​​of other neurons. This sum is the output. In classification problems, one output is generated for each target category (separate sets of weights and summation units). The output value for a category is the probability that the evaluated case belongs to that category. Training an RBF can determine various parameters, such as the number of neurons in the hidden layer, the coordinates of the centers of each hidden layer function, the spread of each function in each dimension, and the weights applied to the output when passing it to the summation layer. Training can be used by clustering algorithms (e.g., k-means clustering), evolutionary approaches, etc.

[0147] In an embodiment, a recurrent neural network may have time-varying real-valued (zero or greater than one) activations (outputs). Each connection has a modifiable real-valued weight. Some of the nodes are called labeled nodes, output nodes, and hidden nodes. In supervised learning in a discrete-time setting, a training sequence of real-valued input vectors results in a sequence of activations of the input nodes, one input vector at a time. At each time step, each non-input unit can compute its current activation as a nonlinear function of the weighted sum of the activations of all units to which it receives connections. The system can explicitly activate some output units at a given time step (independent of the input signal).

[0148] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use a self-organizing neural network, such as a Kohonen self-organizing neural network, for visualization of views of data, such as a lower-dimensional view of high-dimensional data. The self-organizing neural network can apply competitive learning to a set of input data, such as from one or more sensors or other data inputs from a trading environment, including any machine or component associated with the trading environment. In embodiments, the self-organizing neural network can be used to identify structure in data, such as unlabeled data, such as data sensed from various data sources around the trading environment or sensors within the trading environment, where the source of the data is unknown (e.g., an event may originate from any of a variety of unknown sources). The self-organizing neural network can organize structures or patterns in the data so that they can be recognized, analyzed, and labeled, such as identifying market behavior structures as corresponding to other events or signals.

[0149] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use recurrent neural networks, which may allow for bidirectional flow of data, such that connected units (e.g., neurons or nodes) form a directed cycle. Such networks may be used to model or represent dynamic temporal behavior, such as that involved in dynamic systems, such as the wide variety of automated systems, machines, and devices described throughout this disclosure, including automated agents that interact with markets for purposes of data collection, testing spot market transactions, execution trading, etc., where dynamic system behavior includes complex interactions that a user may wish to understand, predict, control, and / or optimize. For example, recurrent neural networks may be used to predict market conditions, such as those involving dynamic processes or actions, such as changes in the state of resources traded in or enabling the trading environment. In embodiments, recurrent neural networks may use internal memory to process a series of inputs, such as from other nodes and / or from sensors and / or other data inputs from or related to the trading environment, of the various types described herein. In embodiments, recurrent neural networks may also be used for pattern recognition, such as to recognize machines, components, agents, or other items based on behavioral signatures, profiles, sets of feature vectors (such as in an audio file or image), etc. In a non-limiting example, a recurrent neural network may recognize shifts in a market or machine's operating mode by learning to classify the shifts from a training dataset composed of data streams from one or more data sources of sensors applied to or relating to one or more resources.

[0150] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use modular neural networks, which may consist of a series of independent neural networks (such as those of the various types described herein) coordinated by a mediator. Each of the independent neural networks in a modular neural network operates on separate inputs and can accomplish subtasks that make up the task the modular network as a whole is intended to perform. For example, a modular neural network may consist of a recurrent neural network for pattern recognition, such as recognizing the type of machine or system sensed by one or more sensors provided as input channels to the modular network, and a RBF neural network for optimizing the operation of the machine or system once understood. The mediator can accept the inputs of the individual neural networks, process them, and create outputs for the modular neural networks, such as appropriate control parameters, state predictions, etc.

[0151] Combinations between pairs, triplets, or larger combinations of the various neural network types described herein are encompassed by this disclosure. This may include combinations in which an expert system uses one neural network to recognize patterns (e.g., patterns indicative of a problem or fault condition) and a different neural network to self-organize activities or workflows based on the recognized patterns (e.g., provide outputs that govern autonomous control of the system in response to the recognized condition or pattern). This may also include combinations in which an expert system uses one neural network to classify items (e.g., identify machines, components, or operating modes) and another neural network to predict item states (e.g., fault states, operational states, prognostic states, maintenance states, etc.). Modular neural networks also include situations in which an expert system uses one neural network to determine a state or context (such as a machine state, a process, a workflow, a market, a storage system, a network, a data collector, etc.) and uses a different neural network to self-organize a process that includes the state or context (e.g., a data storage process, a network coding process, a network selection process, a data marketplace process, a power generation process, a manufacturing process, a refining process, a drilling process, a borehole process, or any other process described herein).

[0152] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may employ physical neural networks in which one or more hardware elements are used to perform or simulate neural operations. In embodiments, one or more hardware neurons may be configured to stream voltage values, current values, etc., representing sensor data to calculate information from analog sensor inputs representing energy consumption, energy production, etc., by one or more machines that provide or consume energy for one or more transactions. One or more hardware nodes may be configured to stream output data resulting from the activity of the neural net. The hardware nodes may be comprised of one or more chips, microprocessors, integrated circuits, programmable logic controllers, application-specific integrated circuits, field programmable gate arrays, etc., and may be provided to optimize an energy-producing or energy-consuming machine or to optimize another parameter of a portion of any type of neural net described herein. Hardware nodes can include hardware for accelerating computations (such as specialized processors for performing basic or more advanced calculations on input data to provide output, specialized processors for filtering or compressing data, specialized processors for decompressing data, specialized processors for compressing specific files or data types (e.g., for processing image data, video streams, acoustic signals, thermal images, heat maps, etc.)), etc. Physical neural nets can be embodied within the data collectors, including those that can be reconfigured by switching or routing inputs in various configurations to provide different neural net configurations within the data collector for processing different types of inputs (the switching and configurations are optionally under the control of an expert system that can include software-based neural nets located on the data collector or remotely).A physical, or at least partially physical, neural net can include physical hardware nodes located within a storage system in a transactional environment, such as a machine, data storage system, distributed ledger, mobile device, server, or cloud resource, for storing data, or for accelerating input / output functions to one or more storage elements that feed or retrieve data from the neural net. A physical, or at least partially physical, neural net can include physical hardware nodes located within a network, such as for accelerating input / output functions to one or more network nodes within the net, for transmitting data within, to, or from an energy edge environment, such as for accelerating relay functions. In physical neural network embodiments, electrically tunable resistive materials can be used to emulate the function of neural synapses. In embodiments, physical hardware emulates neurons, and software emulates the neural network between neurons. In embodiments, neural networks complement traditional algorithmic computers. Neural networks are versatile and can be trained to perform appropriate functions without requiring instructions, such as classification, optimization, pattern recognition, control, selection, and evolution.

[0153] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use multi-layer feedforward neural networks for complex pattern classification of one or more items, phenomena, modes, states, etc. In embodiments, the multi-layer feedforward neural networks are trained by optimization techniques, such as genetic algorithms, to explore a large, complex space of options to find an optimal or near-optimal global solution. For example, multi-layer feedforward neural networks may be trained using one or more genetic algorithms to classify complex phenomena, such as recognizing complex machine operating modes, including modes involving complex interactions between machines (including interference effects, resonance effects, etc.), modes involving nonlinear phenomena, and modes involving catastrophic failures where multiple failures occur simultaneously, making root cause analysis difficult. In embodiments, multi-layer feedforward neural networks may be used to classify results from market monitoring, such as monitoring systems, such as automated agents operating within markets, and monitoring of market-enabling resources, such as computing, networking, energy, data storage, energy storage, and other resources.

[0154] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use feedforward, backpropagation multilayer perceptron (MLP) neural networks to process one or more remote sensing applications, such as to take inputs from sensors distributed across various trading environments. In embodiments, MLP neural networks may be used for classification of trading and resource environments, such as lending markets, spot markets, forward markets, energy markets, renewable energy credit (REC) markets, networking markets, advertising markets, spectrum markets, ticketing markets, rewards markets, computational markets, and others mentioned throughout this disclosure, as well as physical resources and the environments that produce them, such as energy resources (including renewable energy environments, mining environments, exploration environments, drilling environments, etc.), including classification of geological structures (including subsurface and surface features), classification of materials (including fluids, minerals, metals, etc.), and other problems. This may include fuzzy classification.

[0155] In embodiments, expert systems or methods and systems described herein with self-organizing capabilities may use structure-adaptive neural networks, where the structure of the neural network is adapted, such as based on rules, sensed conditions, contextual parameters, etc. For example, if the neural network fails to converge to a solution, such as classifying an item or arriving at a prediction, when acting on an input set after some training, the neural network is changed from a feedforward neural network to a recurrent neural network, such as by switching the data paths between some subset of nodes from unidirectional to bidirectional data paths. The adaptation of the structure may occur under the control of the expert system, such as triggering adaptation upon the occurrence of a trigger, rule, or event (such as recognizing the occurrence of a threshold (such as not converging to a solution within a predetermined time) or recognizing a phenomenon requiring a different or additional structure (such as recognizing that the system is changing dynamically or nonlinearly)). In one non-limiting example, when the expert system receives an indication that a continuously variable transmission is used to drive a generator, turbine, etc. in the system under analysis, the expert system may switch from a simple neural network structure, such as a feedforward neural network, to a more complex neural network structure, such as a recurrent neural network, a convolutional neural network, etc.

[0156] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use autoencoders, autoassociators, or diavolo neural networks, which may be similar to multilayer perceptron (MLP) neural networks, such as those with an input layer, an output layer, and one or more hidden layers connecting them. However, the output layer of an autoencoder may have the same number of units as the input layer, where the goal of an MLP neural network is to reconstruct its own input (rather than simply generating a target value). Thus, autoencoders operate as unsupervised learning models. Autoencoders may be used for unsupervised learning of efficient encoding, such as for dimensionality reduction or for learning generative models of data. In embodiments, autoencoding neural networks may be used to self-learn efficient network coding for the transmission of analog sensor data from machines over one or more networks or for the transmission of digital data from one or more data sources. In embodiments, autoencoding neural networks may be used to self-learn efficient storage approaches for the storage of streams of data.

[0157] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use a probabilistic neural network (PNN), which may, in embodiments, comprise a multi-layer (e.g., four-layer) feed-forward neural network, where the layers may include an input layer, a hidden layer, a pattern / sum layer, and an output layer. In one embodiment of a PNN algorithm, the parent probability distribution function (PDF) of each class may be approximated, such as by a Parzen window and / or a nonparametric function. The PDF of each class may then be used to estimate the class probability of a new input, and Bayes' rule may be employed, such as assigning it to the class with the highest posterior probability. A PNN may embody a Bayesian network and employ statistical algorithms or analytical techniques, such as the Kernel-Fisher discriminant analysis technique. A PNN may be used for classification and pattern recognition in any of the broad embodiments disclosed herein. In one non-limiting example, a probabilistic neural network may be used to predict engine fault conditions based on a collection of data inputs from engine sensors and instruments.

[0158] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use a time-delay neural network (TDNN), which may form a feed-forward architecture for sequential data that recognizes features independent of sequence position. In embodiments, delays are added to one or more inputs or between one or more nodes to account for time shifts in the data, and multiple data points (from different points in time) are analyzed together. The time-delay neural network may form part of a larger pattern recognition system, such as using a perceptron network. In embodiments, the TDNN may be trained using supervised learning, where connection weights are trained under backpropagation or feedback. In embodiments, the TDNN may be used to process sensor data from different streams, such as a stream of speed data, a stream of acceleration data, a stream of temperature data, and a stream of pressure data, and a time delay is used to align the data streams in time, such as to help understand patterns involving understanding the various streams (e.g., changing price patterns in spot or forward markets).

[0159] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use convolutional neural networks (sometimes referred to as CNNs, ConvNets, shift-invariant neural networks, or space-invariant neural networks), where units are connected in a pattern similar to the visual cortex of the human brain. Neurons respond to stimuli in limited regions of space called receptive fields. The receptive fields may overlap, collectively covering the entire (e.g., visual) cortex. Node responses can be mathematically calculated, such as by convolutional operations like multilayer perceptrons, using minimal preprocessing. Convolutional neural networks can be used for recognition in images and video streams, such as to recognize types of machinery in large environments using camera systems deployed on mobile data collection devices such as drones and mobile robots. In embodiments, convolutional neural networks can be used to provide recommendations based on data inputs, including sensor inputs and other contextual information, such as recommending routes for mobile data collectors. In embodiments, convolutional neural networks can be used to process inputs, such as natural language processing of instructions provided by one or more parties involved in a workflow within an environment. In embodiments, convolutional neural networks can be deployed using a large number of neurons (e.g., 100,000, 500,000, or more), multiple layers (e.g., 4, 5, 6, or more), and a large number of parameters (e.g., millions). A convolutional neural network can use one or more convolutional nets.

[0160] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use regulatory feedback networks, such as to recognize emergent phenomena (such as new types of behavior not previously understood in a trading environment).

[0161] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities can use self-organizing maps (SOMs) with unsupervised learning. A set of neurons learns to map points in an input space to coordinates in an output space. The input space can have different dimensions and topology than the output space, and the SOM can preserve these while mapping phenomena to groups.

[0162] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use learning vector quantization neural nets (LVQs). Prototype representatives of classes may be parameterized in a distance-based classification scheme along with an appropriate distance measure.

[0163] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use echo state networks (ESNs), which may be composed of recurrent neural networks with sparsely connected random hidden layers. The weights of the output neurons may be modified (e.g., the weights may be trained based on feedback). In embodiments, ESNs may be used to process time series patterns, such as recognizing patterns of market-related events, such as, by way of example, patterns of price changes in response to stimuli.

[0164] In embodiments, methods and systems described herein with expert system or self-organizing capabilities can use bidirectional recurrent neural networks (BRNNs) that use a finite sequence of values ​​(e.g., voltage values ​​from a sensor) to predict or label each element of the sequence based on both the element's past and future context. This can be done by adding the outputs of two RNNs, one processing the sequence from left to right and the other from right to left. The combined output is a prediction of a target signal, such as provided by a teacher or supervisor. Bidirectional RNNs can be combined with long short-term memory RNNs.

[0165] In embodiments, methods and systems described herein with expert systems or self-organizing capabilities may use hierarchical RNNs that connect elements in various ways to decompose hierarchical operations into useful subprograms, etc. In embodiments, a hierarchical RNN may be used to manage one or more hierarchical templates for data collection in a transactional environment.

[0166] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use probabilistic neural networks, which may introduce random variation into the network, which may be considered a form of statistical sampling, such as Monte Carlo sampling.

[0167] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities can use genetically scaled recurrent neural networks. In such embodiments, an RNN (often an LSTM) is used, where a sequence is decomposed into several scales, and every scale signals the dominant length between two consecutive points. The first-order scale consists of a regular RNN, and the second-order scale consists of all points separated by two indices. An Nth-order RNN connects the first and last nodes. The outputs from all the various scales can be treated as a committee of members, and the associated scores can be used genetically for the next iteration.

[0168] In embodiments, methods and systems described herein with expert systems or self-organizing capabilities can use a Committee of Machines (CoM), which consists of a collection of different neural networks that "vote" together for a given example. Because neural networks can suffer from local minima, using randomly different initial weights often produces different results, even when starting with the same architecture and training. A CoM tends to stabilize results.

[0169] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use associative neural networks (ASNNs), such as those involving a committee of machines extension that combines multiple feedforward neural networks and k-nearest neighbor techniques. Correlations between ensemble responses can be used as a measure of distance between analyzed cases for kNNs, thereby correcting for bias in the neural network ensemble. Associative neural networks can have a memory consistent with the training set. As new data becomes available, the network instantly improves its predictive ability, providing data approximation (self-learning) without retraining. Another important feature of ASNNs is the potential for interpreting neural network results by analyzing correlations between data cases in the model space.

[0170] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use instantly trained neural networks (ITNNs), where hidden and output layer weights are mapped directly from training vector data.

[0171] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities can use spiking neural networks that can explicitly consider the timing of inputs. The network's inputs and outputs can be represented as a series of spikes (e.g., delta functions or more complex shapes). SNNs can process time-domain information (e.g., signals that change over time, such as signals related to the dynamic behavior of a market or trading environment). They are often implemented as recurrent networks.

[0172] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use dynamic neural networks that address nonlinear multivariate behavior and include learning of time-dependent behavior such as transients and lag effects. Transients may include the behavior of shifting market variables such as price, available quantity, available counterparties, etc.

[0173] In embodiments, cascade correlation can be used as an architecture and supervised learning algorithm to complement weight tuning in networks with fixed topology. Cascade correlation can start with a minimal network and automatically learn to add new hidden units one by one, creating a multi-layered structure. When a new hidden unit is added to the network, its input weights are frozen. This unit becomes a permanent feature detector within the network and can generate outputs or create other, more complex feature detectors. Cascade correlation architectures can learn quickly, determine their own size and topology, and retain their structure across different training sets, without requiring backpropagation.

[0174] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities can use neuro-fuzzy networks, such as fuzzy inference systems within the core of an artificial neural network. Depending on the type, several layers can simulate the processes involved in fuzzy inference, such as fuzzification, inference, aggregation, and defuzzification. Embedding a fuzzy system within the general structure of a neural network has the advantage of being able to use available learning methods to find the parameters of the fuzzy system.

[0175] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use configuration pattern generating networks (CPPNs), such as variants of associative neural networks (ANNs) that differ in their sets of activation functions and how they are applied. Typical ANNs use sigmoid functions (and

[0176] A CPPN can contain both types of functions (sometimes Gaussian functions). Furthermore, a CPPN can be applied to the entire space of possible inputs, allowing it to represent a complete image. Because a CPPN is a composition of functions, it can encode an image with virtually infinite resolution, sampling it at the optimal resolution for a particular display.

[0177] This type of network allows new patterns to be added without retraining. In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use a one-shot associative memory network, such as by creating a specific memory structure, which assigns each new pattern to an orthogonal plane using a contiguously connected hierarchical array.

[0178] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may employ hierarchical temporal memory (HTM) neural networks that incorporate structural and algorithmic characteristics of the neocortex. The HTM may employ biomimetic models based on memory-prediction theory. The HTM may be used to discover and infer high-level causes of observed input patterns or sequences.

[0179] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities can use a holographic associative memory (HAM) neural network, which constitutes an analog, correlation-based, associative, stimulus-response system. Information can be mapped to complex phase orientations. This memory is useful for associative memory tasks, generalization, and pattern recognition with shifting attention.

[0180] In embodiments, various embodiments, including network coding, may be used to code transmitted data between network nodes in a neural net, such as when the nodes are located in one or more data collectors or machines in a transactional environment.

[0181] 9-37, embodiments of the present disclosure, including expert systems, self-organization, machine learning, artificial intelligence, and the like, can benefit from the use of neural networks, such as neural networks trained for pattern recognition, for classification of one or more parameters, characteristics, or phenomena, for supporting autonomous control, and for other purposes. Throughout this disclosure, references to neural networks include Dual Process Artificial Neural Networks (DPANNs), feedforward neural networks, radial basis function neural networks, self-organizing neural networks (e.g., Kohonen self-organizing neural networks), recurrent neural networks, modular neural networks, artificial neural networks, physical neural networks, multilayer neural networks, convolutional neural networks, hybrids of neural networks and other expert systems (e.g., hybrid fuzzy logic-neural network systems), autoencoder neural networks, probabilistic neural networks, time delay neural networks, convolutional neural networks, regulatory feedback neural networks, radial basis function neural networks, recurrent neural networks, Hopfield neural networks, Boltzmann machine neural networks, self-organizing map (SOM) neural networks, learning vector quantization (LVQ) neural networks, fully recurrent neural networks, simple recurrent neural networks,Echo state neural network, long short-term memory neural network, bidirectional neural network, hierarchical neural network, probabilistic neural network, genetically scaled RNN neural network, machine committee neural network, associative neural network, physical neural network, instantaneously trained neural network, spiking neural network, neocognitron neural network, dynamic neural network, cascade neural network, neuro-fuzzy neural network, configurational pattern generating neural network, memory neural network, hierarchical temporal memory neural network, deep feedforward neural network, gated recurrent unit (GCU) neural network, autoencoder neural network, variational autoencoder neural network, denoising autoencoder neural network, sparse autoencoder neural network neural networks, Markov chain neural networks, restricted Boltzmann machine neural networks, deep belief neural networks, deep convolutional neural networks, deconvolutional neural networks, deep convolutional inverse graphics neural networks, generative adversarial neural networks, liquid state machine neural networks, extreme learning machine neural networks, echo state neural networks, deep residual neural networks, support vector machine neural networks, neural Turing machine neural networks, and / or holographic associative memory neural networks, or hybrids or combinations thereof, or in combination with other expert systems such as rule-based systems, model-based systems (including those based on physical models, statistical models, flow-based models, biological models, biomimetic models, etc.).

[0182] In an embodiment, the platform 102 includes a dual-process artificial neural network (DPANN) system. The DPANN system includes an artificial neural network (ANN) having behaviors and behavioral processes (e.g., decision-making) that are the product of a training system and a retraining system. The training system is configured to perform automated execution of trained ANN operations. The retraining system performs effortful, analytical, and deliberate retraining of the ANN based on one or more relevant aspects of the ANN, such as memory, one or more input datasets (including temporal information about elements in such datasets), one or more goals or objectives (including those that may change dynamically, periodically, and / or based on contextual changes, such as those related to the context of use of the ANN), and / or other factors. When memory-based retraining is included, the memory can include original / past training data and refined training data. The DPANN system includes a dual-process learning function (DPLF) 902 configured to manage and execute an ongoing data retention process. The DPLF 902 (including memory management processes, if applicable) facilitates retraining and refinement of the ANN's operation. DPLF902 provides a framework for an ANN to create outputs, such as predictions, classifications, recommendations, conclusions, and / or other outputs, based on historical inputs, new inputs, and new outputs, including outputs configured for a particular use case and determined by parameters of the usage context (which may include performance parameters such as latency parameters, accuracy parameters, consistency parameters, bandwidth utilization parameters, processing power utilization parameters, prioritization parameters, energy utilization parameters, etc.).

[0183] In embodiments, the DPANN system stores training data, thereby enabling constant relearning based on the results of the ANN's decisions, predictions, and / or other operations, and analyzing the training data at the ANN's output. Management of entities stored in memory enables the construction and execution of new models, such as those that may be processed, executed, or otherwise performed by or under the control of the training system. The DPANN system uses instances of memory to validate actions (e.g., in a manner similar to that of biological neural networks, including retrospective or self-reflective thinking about whether actions performed under given circumstances were optimal) and to train the ANN (including intentionally training the ANN with an appropriate set of memories (i.e., those that will produce favorable results given the ANN's performance requirements)).

[0184] In an embodiment, FIG. 9 illustrates an exemplary process for DPLF 902. DPLF 902 may be or include the ongoing process of maintaining one or more training datasets and / or memories stored in memory over time. DPLF 902 thereby enables an ANN to apply existing neural functions and utilize sets of past events (including those intentionally altered and / or curated for a specific purpose) to frame its understanding and behavior within current, recent, and / or new scenarios, including in simulations, in the training process, and in fully operational deployments of the ANN. DPLF 902 may provide an ANN with a framework for analyzing, evaluating, and / or managing data, such as data related to the past, present, and future. In this manner, DPLF 902 plays a key role in the training and retraining of ANNs via training and retraining systems.

[0185] In an embodiment, the DPLF 902 is configured to perform dual process operations to manage existing training processes and to manage and / or execute new training processes, i.e., retraining processes. In an embodiment, each instance of the ANN is configured to be trained via a training system and retrained via a retraining system. The ANN encodes, stores, and retrieves training and / or retraining data sets both during training by the training system and during retraining by the retraining system. A DPANN system can recognize whether a dataset (the term dataset in this context optionally includes various subsets, supersets, combinations, permutations, elements, metadata, augmentations, etc., of the base dataset used for training or retraining), memory activity, processing operations, and / or output have characteristics natively advantageous to a training system versus a retraining system based on their respective inputs, processing (e.g., based on its structure, type, model, operation, execution environment, resource utilization, etc.), and / or results (including type of result, performance requirements (including contextual or dynamic requirements)), for example. A DPANN system can resolve poor performance of a training system on a classification task by finding that the ANN was inappropriately trained (e.g., due to the type of dataset, the nature of the input model and / or feedback, the amount of training data, the quality of tagging or labeling, the quality of supervision, etc.), or the ANN's processing operations are unsuitable (e.g., if it is prone to known vulnerabilities due to, for example, the type of neural network used, the type of model used, etc.) by engaging a retraining system to retrain the model and teach it to learn to solve a new classification problem (e.g., by providing many labeled instances of correctly classified items).By periodically or continuously evaluating the performance of the ANN, the DPANN system can determine that the training system is ready to replace the retraining system (or, more favorably, if both are involved) when ANN performance is very stable (e.g., when multiple retrainings by the retraining system result in only slight improvements to the ANN). Over a longer period of time, various performance cycles may emerge, with the DPANN retraining system being used, as needed, to retrain the ANN and / or to augment the ANN by providing a second source of output (which may be fused or combined with the output of the ANN to provide a single result, with various weightings between them, or provided in parallel, allowing for comparison, selection, averaging, or other contextual or situational adaptation of their respective outputs).

[0186] In an embodiment, the ANN is configured to learn new functions in conjunction with data collection through a dual-process training of the ANN via a training system and a retraining system. The DPANN system performs initial training of the ANN through an analysis of the ANN via the training system, such that the ANN acquires new internal functions (or internal functions are subtracted or modified, for example, if existing functions do not contribute to favorable results). After initial training, the DPANN system retrains the ANN via the retraining system. To perform retraining, the retraining system evaluates the ANN's memory and past processing and constructs a target DPLF902 process for retraining. The DPLF902 process may be specific to the identified scenario. ANN processes can run in parallel with the DPLF902 process. For example, after initial training via the training system, the ANN can function to operate a specific make and model of autonomous vehicle. The DPANN system can retrain the ANN's functions via the retraining system to enable the ANN to operate different makes and models (e.g., with different cameras, accelerometers, other sensors, different physical characteristics, different performance requirements, etc.) or different types of vehicles, such as bicycles or spacecraft.

[0187] In embodiments, as the quality of the ANN's output and / or operation improves, and as long as the ANN's performance requirements and usage remain fairly stable, performing the dual-process training process can become less demanding. As such, the DPANN system may require fewer ANN neurons to perform its operations and / or processing, require less intensive performance monitoring (e.g., longer intervals between performance checks), and / or determine that retraining is no longer necessary (at least for a period of time, such as until a long-term maintenance period and / or a significant change in usage context occurs). As the ANN continues to refine existing features and / or add new features via the dual-process learning process, the ANN may simultaneously perform other, sometimes more "intellectually demanding" (e.g., re-learning intensive) tasks. For example, the ANN may utilize the knowledge of the features or processes being learned via the dual-process learning process to solve unrelated complex problems or make concurrent re-learning decisions. Retraining can include supervision, such as an agent (e.g., a human supervisor or intelligent agent) instructing the ANN on the retraining objectives (e.g., "master this new function") and providing a set of training tasks and feedback functions (e.g., supervisory ratings) for the retraining. In embodiments, the ANN can be used to organize, seed, supervise, train, and retrain other dual-process trained ANNs.

[0188] In embodiments, one or more operations and operational processes (e.g., decision-making) of the ANN may be the product of training and re-learning processes, respectively, facilitated by a training system and a re-learning system. The training system may be configured to perform automatic training of the ANN, such as by various data sources or by continually adding additional instances of training data collected from various data sources. The re-training system may be configured to perform effortful, analytical, and deliberate re-training of the ANN based on memory (e.g., saved or refined training data) and / or optionally based on inference or other factors. For example, in a deployment management context, the training system may be associated with a standard response by the ANN, while the re-training system may perform DPLF902 re-training and / or network adaptation of the ANN. In some cases, re-training an ANN beyond a factory or “out-of-the-box” training level may involve more than re-training by the re-learning system. The success of tuning the ANN through one or more network adaptations may depend on the operation of one or more network adjustments of the training system.

[0189] In embodiments, the training system can facilitate rapid operation and training of the ANN by applying the ANN's existing neural functions based on training the ANN with previous datasets. Typical operational activities of an ANN that may heavily rely on a training system may include, but are not limited to, one or more of the methods, processes, workflows, systems, etc. described throughout this disclosure and the documents incorporated herein: defined functions within a network (discovering available networks and connections, establishing connections within a network, provisioning network bandwidth between devices and systems, routing data within a network, steering traffic to available network paths, load balancing among network resources, and many others); recognition and classification (images, text, symbols, objects, video content, music and other audio content, voice content, and many others); spoken language; prediction of states and events (predicting machine and system failure modes, predicting events in a workflow, predicting behavior in shopping and other activities, and many others); control (controlling autonomous or semi-autonomous systems, automated agents (e.g., automated call center operations, chatbots, and so on)); and / or optimization and recommendation (products, content, decision-making, and many others). ANNs may also be suitable for training datasets for scenarios requiring only output. Standard operational activities may not require active analysis of what is required of the ANN beyond operating with well-defined data inputs to compute well-defined outputs for well-defined use cases. The operation of a training and / or retraining system can be based on one or more historical data training data sets, and can use parameters from the historical data training data sets to compute results based on new input values, and can run with little or no modification to the ANN or its input types.In an embodiment, an instance of the training system can be trained to classify whether the ANN will perform well in a given situation, such as by recognizing whether the image or sound to be classified by the ANN is of a type that has historically been classified with high accuracy (e.g., above a threshold).

[0190] In embodiments, network adaptation of the ANN by one or both of the training and retraining systems can include multiple defined network functions, knowledge, and intuition-like behaviors of the ANN when new input values ​​are received. In such embodiments, the retraining system can apply new input values ​​to the DPLF902 system to adjust the functional response of the ANN, thereby retraining the ANN. The DPANN system can determine that retraining of the ANN with network adjustments is necessary, for example, but not by way of limitation, when the functional neural network is assigned activities and assignments that require the ANN to provide solutions to novel problems, engage in network adaptation or other higher-order cognitive activities, apply concepts outside the domain for which the DPANN was originally designed, support a different context for deployment (e.g., when use cases, performance requirements, available resources, or other factors change), etc. The ANN can be trained to recognize where a retraining system is needed by training the ANN to recognize degradation in training system performance, high variability of input datasets relative to historical datasets used to train the training system, new functionality or performance requirements, dynamic changes in use cases or contexts, or other factors. The ANN can apply inference to evaluate performance and provide feedback to the retraining system. ANNs may be trained and / or retrained to perform intuitive functions, optionally through a combination or recombination process (e.g., inputs (e.g., data sources), processes / functions (e.g., ANNs are tested in relation to each other (either in simulation or production deployment) in a series of rounds or evolutionary steps, etc., to promote advantageous variants until a preferred ANN or preferred set of ANNs is identified for a given scenario, use case, or set of requirements)).This may include generating a set of input "ideas" (e.g., combinations of different conclusions regarding causality in a diagnostic process) for retraining the system and subsequent training, and / or processing by an explicit reasoning process, such as a Bayesian reasoning process, a sophistic or conditional reasoning process, a deductive reasoning process, an inductive reasoning process, or others (including combinations of the above), as described in this disclosure or in documents incorporated herein by reference.

[0191] In an embodiment, the DPLF 902 may perform an encoding process to process the dataset into a stored form for future use, such as retraining the ANN by a retraining system. The encoding process allows the dataset to be captured, understood, and modified by the DPLF 902 to better support storage to and use from memory. The DPLF 902 may apply current functional knowledge and / or reasoning to integrate new input values. The memory may include a short-term memory (STM) 906, a long-term memory (LTM) 912, or a combination thereof. The dataset may be stored in one or both of the STM 906 and the LTM 912. The STM 906 may be implemented by applying specialized operations within the ANN (such as a gated or ungated recurrent neural network or a long-term short-term neural network). The LTM 912 may be implemented by storing scenarios, associated data, and / or raw data that can be applied to discover new scenarios. The encoding process may involve, for example, visual encoding data (e.g., processed through a convolutional neural network), acoustic sensor encoding data (e.g., what something sounds like), speech encoding data (e.g., processed through a deep neural network (DNN), optionally including for phoneme recognition), word semantic encoding data (e.g., using a hidden Markov model (HMM) to determine semantic meaning), and / or motion and / or tactile encoding data (such as operations on vibration / accelerometer sensor data, touch sensor data, position or geolocation data, etc.). Datasets can enter the DPLF902 system through one of these modes, but the format in which the dataset is stored may differ from the dataset's original format and may be passed through a neural processing engine to be compressed and / or encoded into a context-relevant format. For example, an unsupervised instance of an ANN may be used to train historical data into a compressed format.

[0192] In an embodiment, the encoded data set is maintained within the DPLF 902 system. The encoded data set is first stored in the short-term DPLF 902, i.e., STM 906. For example, sensor data sets may be primarily stored in the STM 906 and maintained there on a regular basis. Data sets stored in the STM 906 are active and function as a kind of immediate response to new input values. The DPANN system may remove data sets from the STM 906 in response to changes in the data stream, for example, due to the STM 906 running out of capacity as new data is imported, processed, and / or stored. For example, the short-term DPLF 902 may only last between 15 and 30 seconds. The STM 906 may only store a small amount of data, typically embedded within the ANN.

[0193] In embodiments, the DPANN system may measure attention based on training system utilization, overall DPANN system utilization, and / or the like, such as by consuming various indicators of attention to the ANN and / or utilization of output from the ANN and transmitting such indicators in response to the ANN (analogous to a "moment of awareness" in the brain, where attention passes over something and the cognitive system says "aha!"). In embodiments, attention may be measured by the sheer amount of activity of one or both systems on a data stream. In embodiments, a system using output from an ANN may explicitly indicate attention, such as by an operator instructing the ANN to pay attention to specific activity (e.g., to respond to a diagnosed problem, among many other possibilities). The DPANN system may manage data inputs to facilitate attention measurement, such as by prompting and / or calculating greater attention for data with high inherent variability from past patterns (e.g., rate of change, deviation from norms, etc.), data showing high variability in past performance (e.g., data with characteristics similar to datasets involved in situations in which the ANN performed poorly in training), etc.

[0194] In an embodiment, the DPANN system may retain encoded datasets within the DPLF902 system according to and / or as part of one or more storage processes. The DPLF902 system may store encoded datasets in the LTM912 as needed after the encoded datasets have been stored in the STM906 and are determined to be no longer necessary and / or of low priority for the ANN's current operation, training process, retraining process, etc. The LTM912 may be implemented by storing scenarios, and the DPANN system may apply associated and / or unprocessed data to discover new scenarios. For example, data from certain processed data streams, such as semantically encoded datasets, may be stored primarily in the LTM912. The LTM912 may also store image (and sensor) datasets in encoded form, among many other examples.

[0195] In embodiments, LTM 912 may have a relatively high storage capacity, and datasets stored within LTM 912 may, in some scenarios, be stored virtually indefinitely. The DPANN system may be configured to remove datasets from LTM 912, e.g., by passing LTM 912 data through a series of memory structures with progressively longer retrieval periods or progressively higher threshold requirements for triggering utilization (similar to how a biological brain “thinks very hard” to find precedents for addressing difficult problems), thereby increasing the salience of more recent or more frequently used memories while preserving the ability to retrieve (at greater time / effort) older memories when the situation warrants more comprehensive memory utilization. In this manner, the DPANN system may arrange datasets stored in LTM 912 chronologically by storing older memories (as measured by time of origin and / or recency of utilization) in a separate and / or slower system, by penalizing older memories by imposing artificial delays on their retrieval, and / or by imposing threshold requirements before utilization (such as an indicator of a higher demand for improved results). Additionally or alternatively, LTM 912 can cluster according to other classification protocols, such as by topic. For example, all memories that are temporally proximate to a regularly recognized person may be clustered to be searched together, and / or all memories that were related to a scenario may be clustered to be searched together.

[0196] In embodiments, the DPANN system may modularize and link LTM912 datasets in catalogs, hierarchies, clusters, knowledge graphs (directed / undirected or with conditional logic), etc., to facilitate retrieval of related memories. For example, all memory modules with instances containing people, topics, items, processes, n-tuple concatenations of such (e.g., all memory modules containing a selected pair of entities), etc. The DPANN system may select subgraphs of the knowledge graph for DPLF902 to implement in one or more domain-specific and / or task-specific applications, such as training a model to predict the behavior of a robot or human agent using memories related to a specific robot or human agent and / or a collection of similar robots or human agents. The DPLF902 system may cache frequently used modules for different speeds and / or usage probabilities. High-value modules (e.g., those with high-quality results, performance characteristics, etc.) can be used for other functions, such as selecting / training the STM906 retention / forget process.

[0197] In embodiments, the DPANN system can modularize and link LTM datasets, such as in the various ways described above, to facilitate retrieval of relevant memories. For example, memory modules with instances containing n-tuples of people, topics, items, processes, and the like (e.g., all memory modules containing a selected pair of entities), or all memories related to a scenario, may be linked and retrieved. The DPANN system may select a subset of scenarios (e.g., a subgraph of a knowledge graph) for the DPLF 902 for domain-specific and / or task-specific use, such as training a model to predict the behavior of a robot or human agent using memories associated with a specific set of robots or human agents. Frequently used modules or scenarios can be cached for different speed / usage probability or other performance characteristics. High-value modules or scenarios (those that produce high-quality results) can be used for other functions, such as selecting / training a keep / forget process for the STM 906.

[0198] In embodiments, a DPANN system can perform LTM planning, such as finding a procedural course of action for a declaratively described system to reach a goal while optimizing an overall performance metric. A DPANN system can perform LTM planning, for example, when the problem can be described in a declarative manner, when the DPANN system has domain knowledge that should not be ignored, when the problem is structured in a way that makes the problem difficult for pure learning techniques, and / or when the ANN needs to be trained and / or retrained to be able to explain the specific course of action taken by the DPANN system. In embodiments, a DPANN system can be applied to plan recognition problems, i.e., the inverse problem of a planning problem: instead of a goal state, a set of possible goals is given, and the objective in plan recognition is to find which goals are achieved and how.

[0199] In embodiments, the DPANN system can facilitate long-term (LTM) scenario planning by users to develop long-term plans. For example, LTM scenario planning for risk management use cases can focus on identifying extreme or unusual, yet possible, risks and opportunities not typically considered in daily operations, such as those that fall outside of a bell curve or normal distribution but occur more frequently than expected in "long-tail" or "fat-tail" situations. LTM scenario planning can involve analyzing the relationships between forces (e.g., social, technological, economic, environmental, and / or political trends) to explain the current situation and / or providing scenarios of potential future states.

[0200] In embodiments, the DPANN system can facilitate LTM scenario planning to forecast and predict possible alternative futures along with the ability to respond to predicted states. LTM planning may be derived from expert domain knowledge or predicted from current scenarios. This is because many scenarios (e.g., those involving the outcome of combinatorial processes that result in new entities or behaviors) have never occurred and therefore cannot be predicted by probabilistic means that rely entirely on past distributions. The DPANN system may prepare applications to the LTM 912 to generate many different scenarios and explore various possible futures for the DPLM, both expected and unexpected. This may be facilitated or augmented by genetic programming and inference techniques, particularly as described above.

[0201] In embodiments, the DPANN system can implement LTM scenario planning to facilitate the transformation of risk management into a plan-aware problem and apply DPLF902 to generate potential solutions. LTM scenario induction addresses several challenges inherent in predictive planning. LTM scenario induction is applicable, for example, when the models used for forecasting have inconsistent, missing, or unreliable observations; when it is possible to generate many future plans rather than just one; and / or when LTM domain knowledge can be captured and encoded to improve forecasts (e.g., when domain experts tend to outperform available computational models). LTM scenarios can focus on applying LTM scenario planning to risk management. LTM scenario planning can provide situational awareness of relevant risk drivers by detecting emerging storylines. Furthermore, LTM scenario planning can generate future scenarios that enable DPLMs and operators to reason about and plan for future contingencies and opportunities.

[0202] In embodiments, the DPANN system may be configured to execute a search process via the DPLF 902 to access the ANN's stored datasets. The search process can determine how well the ANN performs on tasks designed to test recall. For example, an ANN may be trained to perform a controlled vehicle parking maneuver in which an autonomous vehicle returns to a designated spot or exit by correlating previous visits via a search of data stored in the LTM 912. The datasets stored in the STM 906 and the LTM may be searched by different processes. The datasets stored in the STM 906 may be searched in response to specific inputs and / or by the order in which the datasets were stored (e.g., a list of consecutive numbers). The datasets stored in the LTM 912 may be searched through association and / or matching of events with historical activity, for example, through complex association and indexing of large datasets.

[0203] In embodiments, the DPANN system can implement scenario monitoring as at least part of the search process. Scenarios can provide context for context-based decision-making processes. In embodiments, scenarios may include explicit reasoning (such as causal inference, Bayesian inference, sophistry inference, conditional logic, etc., or a combination thereof), and their output declares what LTM-stored data to search (e.g., a timeline of events to be evaluated, and potentially other timelines containing events that follow similar causal patterns). For example, diagnosing a machine or workflow failure can search not only historical sensor data but also LTM data regarding various failure modes of that type of machine or workflow (and / or similar processes, including diagnosing problem states or conditions, recognizing events or behaviors, failure modes (e.g., financial failures, contract breaches, etc.), or many others).

[0204] In embodiments, FIGS. 10-37 depict exemplary neural nets, with FIG. 10 depicting a legend indicating the various components of the neural nets depicted throughout FIGS. 10-37. FIG. 10 depicts the various neural net components depicted in cells with assigned functions and requirements. In embodiments, various neural net examples may include (from top to bottom in the example of FIG. 10) backfeed data / sensor input cells, data / sensor input cells, noise input cells, and hidden cells. Neural net components may also include stochastic hidden cells, spike hidden cells, output cells, coincidence input / output cells, recurrent cells, memory cells, difference memory cells, kernels, and convolution or pooling cells.

[0205] In an embodiment, Figure 11 depicts an exemplary perceptron neural network that may be connected to, integrated with, or interfaced with platform 102. The platform may also be associated with additional neural net systems, such as a feedforward neural network (Figure 12), a radial basic neural network (Figure 13), a deep feedforward neural network (Figure 14), a recurrent neural network (Figure 15), a long / short-term neural network (Figure 16), and a gated recurrent neural network (Figure 17). The platform may also be associated with additional neural net systems, such as an autoencoder neural network (Figure 18), a variational neural network (Figure 19), a denoising neural network (Figure 20), a sparse neural network (Figure 21), a Markov chain neural network (Figure 22), and a Hopfield network neural network (Figure 23). The platform may further be associated with additional neural net systems, such as a Boltzmann machine neural network (Figure 24), a restricted BM neural network (Figure 25), a deep belief neural network (Figure 26), a deep convolutional neural network (Figure 27), a deconvolutional neural network (Figure 28), and a deep convolutional inverse graphics neural network (Figure 29). The platform may also be associated with further neural net systems, such as a generative adversarial neural network (Figure 30), a liquid state machine neural network (Figure 31), an extreme learning machine neural network (Figure 32), an echo state neural network (Figure 31), a deep convolutional inverse graphics neural network (Figure 32), an echo state neural network (Figure 33), a deep residual neural network (Figure 34), a Kohonen neural network (Figure 35), a support vector machine neural network (Figure 36), and a neural Turing machine neural network (Figure 37).

[0206] Such neural networks have various nodes or neurons that can perform various functions on inputs, such as inputs received from sensors or other data sources, including other nodes. Functions include weights, features, feature vectors, etc. Neurons include perceptrons, neurons that mimic biological functions (such as human touch, vision, taste, hearing, and smell), etc. Continuous neurons, such as those with sigmoid activation, are used in the context of various forms of neural nets, such as when backpropagation is involved.

[0207] In many embodiments, an expert system or neural network may be trained, such as by a human operator or supervisor, or based on a data set, a model, or the like. Training may include presenting the neural network with one or more training data sets representing values, such as sensor data, event data, parameter data, and other types of data (including many types described throughout this disclosure), as well as one or more indicators of results, such as process results, computation results, event results, or activity results. Training may also include optimization training, such as training a neural network to optimize one or more systems based on one or more optimization approaches, such as a Bayesian approach, a parametric Bayesian classifier approach, a k-nearest neighbor classifier approach, an iterative approach, an interpolation approach, a Pareto optimization approach, or an algorithmic approach. Feedback may be provided in a variation-and-selection process, such as a genetic algorithm that evolves one or more solutions based on feedback through a series of rounds.

[0208] In embodiments, multiple neural networks may be deployed to a cloud platform that receives data streams and other inputs collected in one or more energy edge environments (e.g., by mobile data collectors) and transmits them to the cloud platform over one or more networks, including using network coding to provide efficient transmission. The cloud platform may optionally use massively parallel computing power to employ multiple different neural networks of various types (including modular, adaptive, hybrid, etc.) to perform prediction, classification, control functions, and provide other outputs, as described in connection with the expert systems disclosed throughout this disclosure. Different neural networks may be structured to compete with each other (optionally using evolutionary algorithms, genetic algorithms, etc.) so that an appropriate type of neural network, with an appropriate set of inputs, weights, node types, capabilities, etc., is selected by the expert system, etc., for the particular task involved in a given context, workflow, environmental process, system, etc.

[0209] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use feedforward neural networks. A feedforward neural network moves information unidirectionally from a data input, such as a data source related to at least one resource or parameter associated with a transaction environment, such as any of the data sources mentioned throughout this disclosure, through a series of neurons or nodes to an output. Data moves from the input node to the output node, optionally passing through one or more hidden nodes without loops. In embodiments, the feedforward neural network may be constructed with various types of units, such as binary McCulloch-Pitts neurons, the simplest of which are perceptrons.

[0210] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use capsule neural networks for prediction, classification, or control functions, etc., related to trading environments, such as in connection with one or more of the machines and automation systems described throughout this disclosure.

[0211] In embodiments, methods and systems described herein involving expert systems or self-organizing capabilities may use radial basis function (RBF) neural networks, which may be preferable in some situations involving interpolation in multidimensional space (e.g., where interpolation is useful in optimizing multidimensional functions, such as optimizing data marketplaces as described herein, optimizing the efficiency or output of power generation systems, factory systems, etc.). In embodiments, each neuron in an RBF neural network stores an example from a training set as a "prototype." The linearity involved in the functioning of this neural network provides RBFs with the advantage that they generally do not suffer from problems with local minima or maxima.

[0212] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities can use radial basis function (RBF) neural networks, such as those employing a distance criterion to the center (e.g., a Gaussian function). Radial basis functions may be applied as an alternative to hidden layers, such as sigmoid hidden layer transfer, in multilayer perceptrons. RBF networks may have two layers, with inputs mapped to each RBF in the hidden layer. In embodiments, the output layer is composed of a linear combination of hidden layer values, representing, for example, the average predicted output. The output layer values ​​can provide outputs that are the same as or similar to the outputs of regression models in statistics. In classification problems, the output layer is a sigmoid function of a linear combination of hidden layer values, representing posterior probabilities. In both cases, performance is often improved by shrinkage techniques such as ridge regression in classical statistics. In a Bayesian framework, this corresponds to a prior belief in small parameter values ​​(and therefore a smooth output function). RBF networks can avoid local minima because the only parameters adjusted in the learning process are the linear mapping from the hidden layer to the output layer. Linearity ensures that the error surface is quadratic and therefore has a single minimum. In regression problems, this may be found with a single matrix operation. In classification problems, the fixed nonlinearity introduced by a sigmoid output function may be handled using an iteratively reweighted least-squares function, for example. RBF networks may use kernel methods such as support vector machines (SVMs) or Gaussian processes (where the RBF is the kernel function). A nonlinear kernel function may be used to project the input data into a space where the learning problem can be solved using a linear model.

[0213] In embodiments, an RBF neural network may include an input layer, a hidden layer, and a summation layer. In the input layer, one neuron appears for each predictor variable. For categorical variables, Nl neurons are used, where N is the number of categories. In embodiments, the input neurons may standardize the range of values ​​by subtracting the median and dividing by the interquartile range. The input neurons send their values ​​to each neuron in the hidden layer. The hidden layer may use a variable number of neurons (determined by a learning process). Each neuron consists of a radial basis function centered on a point with the same dimensions as the number of predictor variables. The spread (e.g., radius) of the RBF function may be different for each dimension. The center and spread are determined by training. When presented with a vector of input values ​​from the input layer, the hidden neuron calculates the Euclidean distance of the test case from the neuron's center point and applies the RBF kernel function to this distance, such as using a spread value. The resulting value is passed to the summation layer. In the summation layer, the value coming out of a neuron in the hidden layer is multiplied by the weight associated with that neuron and then summed with the weighted values ​​of other neurons. This sum is the output. In classification problems, one output is generated for each target category (separate sets of weights and summation units). The output value for a category is the probability that the evaluated case belongs to that category. Training an RBF can determine various parameters, such as the number of neurons in the hidden layer, the coordinates of the centers of each hidden layer function, the spread of each function in each dimension, and the weights applied to the output when passing it to the summation layer. Training can be used by clustering algorithms (e.g., k-means clustering), evolutionary approaches, etc.

[0214] In an embodiment, a recurrent neural network may have time-varying real-valued (zero or greater than one) activations (outputs). Each connection has a modifiable real-valued weight. Some of the nodes are called labeled nodes, output nodes, and hidden nodes. In supervised learning in a discrete-time setting, a training sequence of real-valued input vectors results in a sequence of activations of the input nodes, one input vector at a time. At each time step, each non-input unit can compute its current activation as a nonlinear function of the weighted sum of the activations of all units to which it receives connections. The system can explicitly activate some output units at a given time step (independent of the input signal).

[0215] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use self-organizing neural networks, such as the Kohonen self-organizing neural network, for visualization of views of data, such as lower-dimensional views of high-dimensional data. The self-organizing neural network can apply competitive learning to a set of input data, such as from one or more sensors or other data inputs from or associated with the trading environment, including any machines or components associated with the trading environment. In embodiments, the self-organizing neural network can be used to identify structures in data, such as unlabeled data, such as data sensed from various data sources around the transaction environment or sensors within the transaction environment, where the source of the data is unknown (e.g., an event may originate from any of a variety of unknown sources). The self-organizing neural network can organize structures or patterns in the data so that they can be recognized, analyzed, and labeled.

[0216] In embodiments, methods and systems described herein with expert systems or self-organizing capabilities may use recurrent neural networks, which may allow for bidirectional flow of data, such that connected units (e.g., neurons or nodes) form a directed cycle. Such networks may be used to model or represent dynamic temporal behavior, such as that involved in dynamic systems, such as the wide variety of automated systems, machines, and devices described throughout this disclosure, including automated agents that interact with markets for purposes of data collection, testing spot market transactions, execution trading, etc., where dynamic system behavior includes complex interactions that a user may wish to understand, predict, control, and / or optimize. For example, recurrent neural networks may be used to predict market conditions, such as those involving dynamic processes or actions, such as changes in the state of resources traded in or enabling the trading environment. In embodiments, recurrent neural networks may use internal memory to process a series of inputs, such as from other nodes and / or from sensors and / or other data inputs from or related to the trading environment, of the various types described herein. In embodiments, recurrent neural networks may also be used for pattern recognition, such as to recognize machines, components, agents, or other items based on behavioral signatures, profiles, sets of feature vectors (such as in an audio file or image), etc. In a non-limiting example, a recurrent neural network may recognize shifts in a market or machine's operating mode by learning to classify the shifts from a training dataset composed of data streams from one or more data sources of sensors applied to or relating to one or more resources.

[0217] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use modular neural networks, which may consist of a series of independent neural networks (such as those of the various types described herein) coordinated by a mediator. Each of the independent neural networks in a modular neural network operates on separate inputs and can accomplish subtasks that make up the task the modular network as a whole is intended to perform. For example, a modular neural network may consist of a recurrent neural network for pattern recognition, such as recognizing the type of machine or system sensed by one or more sensors provided as input channels to the modular network, and a RBF neural network for optimizing the operation of the machine or system once understood. The mediator can accept the inputs of the individual neural networks, process them, and create outputs for the modular neural networks, such as appropriate control parameters, state predictions, etc.

[0218] Combinations between pairs, triplets, or larger combinations of the various neural network types described herein are encompassed by this disclosure. This may include combinations in which an expert system uses one neural network to recognize patterns (e.g., patterns indicative of a problem or fault condition) and a different neural network to self-organize activities or workflows based on the recognized patterns (e.g., provide outputs that govern autonomous control of the system in response to the recognized condition or pattern). This may also include combinations in which an expert system uses one neural network to classify items (e.g., identify machines, components, or operating modes) and another neural network to predict item states (e.g., fault states, operational states, prognostic states, maintenance states, etc.). Modular neural networks also include situations in which an expert system uses one neural network to determine a state or context (such as a machine state, a process, a workflow, a market, a storage system, a network, a data collector, etc.) and uses a different neural network to self-organize a process that includes the state or context (e.g., a data storage process, a network coding process, a network selection process, a data marketplace process, a power generation process, a manufacturing process, a refining process, a drilling process, a borehole process, or any other process described herein).

[0219] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may employ physical neural networks in which one or more hardware elements are used to perform or simulate neural operations. In embodiments, one or more hardware neurons may be configured to stream voltage values, current values, etc., representing sensor data to calculate information from analog sensor inputs representing energy consumption, energy production, etc., by one or more machines that provide or consume energy for one or more transactions. One or more hardware nodes may be configured to stream output data resulting from the activity of the neural net. The hardware nodes may be comprised of one or more chips, microprocessors, integrated circuits, programmable logic controllers, application-specific integrated circuits, field programmable gate arrays, etc., and may be provided to optimize an energy-producing or energy-consuming machine or to optimize another parameter of a portion of any type of neural net described herein. Hardware nodes can include hardware for accelerating computations (such as specialized processors for performing basic or more advanced calculations on input data to provide output, specialized processors for filtering or compressing data, specialized processors for decompressing data, specialized processors for compressing specific files or data types (e.g., for processing image data, video streams, acoustic signals, thermal images, heat maps, etc.)), etc. Physical neural nets can be embodied within the data collectors, including those that can be reconfigured by switching or routing inputs in various configurations to provide different neural net configurations within the data collector for processing different types of inputs (the switching and configurations are optionally under the control of an expert system that can include software-based neural nets located on the data collector or remotely).A physical, or at least partially physical, neural net can include physical hardware nodes located within a storage system in a transactional environment, such as a machine, data storage system, distributed ledger, mobile device, server, or cloud resource, for storing data, or for accelerating input / output functions to one or more storage elements that feed or retrieve data from the neural net. A physical, or at least partially physical, neural net can include physical hardware nodes located within a network, such as for accelerating input / output functions to one or more network nodes within the net, for transmitting data within, to, or from an energy edge environment, such as for accelerating relay functions. In physical neural network embodiments, electrically tunable resistive materials can be used to emulate the function of neural synapses. In embodiments, physical hardware emulates neurons, and software emulates the neural network between neurons. In embodiments, neural networks complement traditional algorithmic computers. Neural networks are versatile and can be trained to perform appropriate functions without requiring instructions, such as classification, optimization, pattern recognition, control, selection, and evolution.

[0220] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use multi-layer feedforward neural networks for complex pattern classification of one or more items, phenomena, modes, states, etc. In embodiments, the multi-layer feedforward neural networks may be trained using optimization techniques, such as genetic algorithms, to explore a large, complex space of options to find an optimal or near-optimal global solution. For example, multi-layer feedforward neural networks may be trained using one or more genetic algorithms to classify complex phenomena, such as recognizing complex machine operating modes, including modes with complex interactions between machines (including interference effects, resonance effects, etc.), modes with nonlinear phenomena, and modes with catastrophic failures where multiple failures occur simultaneously, making root cause analysis difficult. In embodiments, multi-layer feedforward neural networks may be used to classify results from market monitoring, such as monitoring systems, such as automated agents operating within markets, and monitoring of market-enabling resources, such as computing, networking, energy, data storage, energy storage, and other resources.

[0221] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use feedforward, backpropagation multilayer perceptron (MLP) neural networks to process one or more remote sensing applications, such as to take inputs from sensors distributed across various trading environments. In embodiments, MLP neural networks may be used for classification of energy edge and resource environments, such as spot markets, forward markets, energy markets, renewable energy credit (REC) markets, network markets, advertising markets, spectrum markets, ticket markets, reward markets, computational markets, and others mentioned throughout this disclosure, as well as physical resources and the environments that produce them, such as energy resources (including renewable energy environments, mining environments, exploration environments, drilling environments, etc.), including classification of geological structures (including subsurface and surface features), classification of materials (including fluids, minerals, metals, etc.), and other problems. This may include fuzzy classification. In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use structure-adaptive neural networks, where the structure of the neural network is adapted, such as based on rules, sensed conditions, contextual parameters, etc. For example, if the neural network fails to converge to a solution, such as classifying an item or arriving at a prediction, when acting on an input set after some training, the neural network is changed from a feedforward neural network to a recurrent neural network, such as by switching the data paths between some subset of nodes from unidirectional to bidirectional data paths. The adaptation of the structure may occur under the control of the expert system, such as triggering adaptation upon the occurrence of a trigger, rule, or event (such as recognizing the occurrence of a threshold (such as not converging to a solution within a predetermined time) or recognizing a phenomenon requiring a different or additional structure (such as recognizing that the system is changing dynamically or nonlinearly)).In one non-limiting example, when the expert system receives an indication that a continuously variable transmission is used to drive a generator, turbine, etc. in the system under analysis, the expert system may switch from a simple neural network structure, such as a feedforward neural network, to a more complex neural network structure, such as a recurrent neural network, a convolutional neural network, etc.

[0222] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use autoencoders, autoassociators, or diavolo neural networks, which may be similar to multilayer perceptron (MLP) neural networks, such as those with an input layer, an output layer, and one or more hidden layers connecting them. However, the output layer of an autoencoder may have the same number of units as the input layer, where the goal of an MLP neural network is to reconstruct its own input (rather than simply generating a target value). Thus, autoencoders operate as unsupervised learning models. Autoencoders may be used for unsupervised learning of efficient encoding, such as for dimensionality reduction or for learning generative models of data. In embodiments, autoencoding neural networks may be used to self-learn efficient network coding for the transmission of analog sensor data from machines over one or more networks or for the transmission of digital data from one or more data sources. In embodiments, autoencoding neural networks may be used to self-learn efficient storage approaches for the storage of streams of data.

[0223] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use a probabilistic neural network (PNN), which may, in embodiments, comprise a multi-layer (e.g., four-layer) feed-forward neural network, where the layers may include an input layer, a hidden layer, a pattern / sum layer, and an output layer. In one embodiment of a PNN algorithm, the parent probability distribution function (PDF) of each class may be approximated, such as by a Parzen window and / or a nonparametric function. The PDF of each class may then be used to estimate the class probability of a new input, and Bayes' rule may be employed, such as assigning it to the class with the highest posterior probability. A PNN may embody a Bayesian network and may use statistical algorithms and analytical techniques, such as the Kernel-Fisher discriminant analysis technique. A PNN may be used for classification and pattern recognition in any of the broad embodiments disclosed herein. As a non-limiting example, a probabilistic neural network may be used to predict engine fault conditions based on a collection of data inputs from engine sensors and instruments.

[0224] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use a time-delay neural network (TDNN), which may constitute a feed-forward architecture for sequential data that recognizes features independent of sequence position. In embodiments, delays are added to one or more inputs or between one or more nodes to account for time shifts in the data, and multiple data points (from different points in time) are analyzed together. The time-delay neural network may also constitute part of a larger pattern recognition system, such as one using a perceptron network. In embodiments, the TDNN may be trained using supervised learning, where connection weights are trained under backpropagation or feedback. In embodiments, the TDNN may be used to process sensor data from different streams, such as a stream of velocity data, a stream of acceleration data, a stream of temperature data, and a stream of pressure data, and a time delay is used to align the data streams in time, such as to help understand patterns involving understanding the various streams (e.g., changing price patterns in spot or forward markets).

[0225] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use convolutional neural networks (sometimes referred to as CNNs, ConvNets, shift-invariant neural networks, or space-invariant neural networks), where units are connected in a pattern similar to the visual cortex of the human brain. Neurons respond to stimuli in limited regions of space called receptive fields. Receptive fields may overlap or collectively cover the entire (e.g., visual) cortex. Node responses may be mathematically calculated, such as by convolutional operations using multilayer perceptrons with minimal preprocessing. Convolutional neural networks may be used for recognition in images and video streams, such as to recognize types of machinery in large environments using camera systems deployed on mobile data collection devices such as drones and mobile robots. In embodiments, convolutional neural networks may be used to provide recommendations based on data inputs, including sensor inputs and other contextual information, such as recommending routes for mobile data collectors. In embodiments, convolutional neural networks may be used to process inputs, such as natural language processing of instructions provided by one or more parties involved in a workflow within an environment. In embodiments, convolutional neural networks can be deployed using a large number of neurons (e.g., 100,000, 500,000, or more), multiple layers (e.g., 4, 5, 6, or more), and a large number of parameters (e.g., millions). A convolutional neural network can use one or more convolutional nets.

[0226] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use regulatory feedback networks, such as to recognize emergent phenomena (such as new types of behavior not previously understood in a trading environment).

[0227] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use self-organizing maps (SOMs) with unsupervised learning. A set of neurons learns to map points in an input space to coordinates in an output space. The input space may have different dimensions and topology than the output space, and the SOM may preserve these while mapping phenomena to groups.

[0228] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use learning vector quantization neural nets (LVQs). Prototype representatives of classes may be parameterized in a distance-based classification scheme along with an appropriate distance measure.

[0229] In embodiments, methods and systems described herein with expert system or self-organizing capabilities may use echo state networks (ESNs), which may be composed of recurrent neural networks with sparsely connected random hidden layers. The weights of the output neurons may be modified (e.g., the weights may be trained based on feedback). In embodiments, ESNs may be used to process time series patterns, such as recognizing patterns of market-related events, such as, by way of example, patterns of price changes in response to stimuli.

[0230] In embodiments, methods and systems described herein with expert system or self-organizing capabilities can use bidirectional recurrent neural networks (BRNNs) that use a finite sequence of values ​​(e.g., voltage values ​​from a sensor) to predict or label each element of the sequence based on both the element's past and future context. This can be done by adding the outputs of two RNNs, one processing the sequence from left to right and the other from right to left. The combined output is a prediction of a target signal, such as provided by a teacher or supervisor. Bidirectional RNNs can be combined with long short-term memory RNNs.

[0231] In embodiments, methods and systems described herein with expert systems or self-organizing capabilities may use hierarchical RNNs that connect elements in various ways to decompose hierarchical operations into useful subprograms, etc. In embodiments, a hierarchical RNN may be used to manage one or more hierarchical templates for data collection in a transactional environment.

[0232] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use probabilistic neural networks, which may introduce random variation into the network, which may be considered a form of statistical sampling, such as Monte Carlo sampling.

[0233] In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities can use genetically scaled recurrent neural networks. In such embodiments, an RNN (often an LSTM) is used, where a sequence is decomposed into several scales, and every scale signals the dominant length between two consecutive points. The first-order scale consists of a regular RNN, and the second-order scale consists of all points separated by two indices. An Nth-order RNN connects the first and last nodes. The outputs from all the various scales can be treated as a committee of members, and the associated scores can be used genetically for the next iteration.

[0234] In embodiments, methods and systems described herein with expert systems or self-organizing capabilities can use a Committee of Machines (CoM), which consists of a collection of different neural networks that "vote" together for a given example. Because neural networks can suffer from local minima, using randomly different initial weights often produces different results, even when starting with the same architecture and training. A CoM tends to stabilize results.

[0235] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use associative neural networks (ASNNs), such as an extension of machine learning that combines multiple feedforward neural networks and k-nearest neighbor techniques. Correlations between ensemble responses can also be used as a measure of distance between analysis cases in kNNs, thereby correcting for bias in the neural network ensemble. Associative neural networks may have a memory consistent with the training set. As new data becomes available, the network instantly improves its predictive capabilities, providing data approximation (self-learning) without retraining. Another important feature of ASNNs is the possibility to interpret neural network results by analyzing correlations between data cases in the model space.

[0236] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use instantly trained neural networks (ITNNs), where hidden and output layer weights are mapped directly from training vector data.

[0237] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use spiking neural networks that can explicitly consider the timing of inputs. The network's inputs and outputs may be represented as a series of spikes (e.g., delta functions or more complex shapes). SNNs may process time-domain information (e.g., signals that change over time, such as signals related to the dynamic behavior of a market or trading environment). They are often implemented as recurrent networks.

[0238] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may use dynamic neural networks that address nonlinear multivariate behavior and include learning of time-dependent behavior such as transients and lag effects. Transients may include the behavior of shifting market variables such as price, available quantity, available counterparties, etc.

[0239] In embodiments, cascade correlation can be used as an architecture and supervised learning algorithm to complement weight tuning in networks with fixed topology. Cascade correlation can start with a minimal network and automatically learn to add new hidden units one by one, creating a multi-layered structure. When a new hidden unit is added to the network, its input weights are frozen. This unit becomes a permanent feature detector within the network and can generate outputs or create other, more complex feature detectors. Cascade correlation architectures can learn quickly, determine their own size and topology, and retain their structure across different training sets, without requiring backpropagation.

[0240] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities can use neuro-fuzzy networks, such as fuzzy inference systems within the core of an artificial neural network. Depending on the type, several layers can simulate the processes involved in fuzzy inference, such as fuzzification, inference, aggregation, and defuzzification. Embedding a fuzzy system within the general structure of a neural network has the advantage of being able to use available learning methods to find the parameters of the fuzzy system.

[0241] In embodiments, methods and systems described herein with expert system or self-organizing capabilities can use configuration pattern generating networks (CPPNs), such as variants of associative neural networks (ANNs), that differ in their sets of activation functions and how they are applied. While typical ANNs often contain only sigmoidal functions (and sometimes Gaussian functions), CPPNs can contain both types of functions, as well as many others. Furthermore, CPPNs can be applied to the entire space of possible inputs, allowing them to represent complete images. Because CPPNs are compositions of functions, they can encode images with virtually infinite resolution and sample at the optimal resolution for a particular display.

[0242] This type of network allows new patterns to be added without retraining. In embodiments, the methods and systems described herein with expert systems or self-organizing capabilities may use one-shot associative memory networks, such as by creating a specific memory structure, which uses adjacently connected hierarchical arrays to assign each new pattern to an orthogonal plane.

[0243] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities may employ hierarchical temporal memory (HTM) neural networks that incorporate structural and algorithmic characteristics of the neocortex. The HTM may employ biomimetic models based on memory-prediction theory. The HTM may be used to discover and infer high-level causes of observed input patterns or sequences.

[0244] In embodiments, the methods and systems described herein with expert system or self-organizing capabilities can use a holographic associative memory (HAM) neural network, which constitutes an analog, correlation-based, associative, stimulus-response system. Information can be mapped to complex phase orientations. This memory is useful for associative memory tasks, generalization, and pattern recognition with shifting attention. Quantum Computing Services

[0245] FIG. 38 illustrates an exemplary quantum computing system 3800 according to some embodiments of the present disclosure. In embodiments, the quantum computing system 3800 provides a framework for providing a set of quantum computing services to one or more quantum computing clients. In some embodiments, the quantum computing system 3800 framework may be at least partially replicated in each quantum computing client. In these embodiments, individual clients may include some or all of the functionality of the quantum computing system 3800, thereby adapting the quantum computing system 3800 to the particular functions performed by the quantum computing client's subsystems. Additionally or alternatively, in some embodiments, the quantum computing system 3800 may be implemented as a set of microservices such that different quantum computing clients can utilize the quantum computing system 3800 through one or more APIs exposed to the quantum computing clients. In these embodiments, the quantum computing system 3800 may be configured to perform various types of quantum computing services that may be adapted to different quantum computing clients. In any of these configurations, a quantum computing client may provide a request to the quantum computing system 3800 requesting performance of a particular task (e.g., optimization). In response, the quantum computing system 3800 performs the requested task and returns a response to the quantum computing client.

[0246] Referring to FIG. 38 , in some embodiments, quantum computing system 3800 may include a quantum adaptation service library 3802, a quantum general service library 3804, a quantum data service library 3806, a quantum computing engine library 3808, a quantum computing configuration service 3810, a quantum computing execution system 3812, and a quantum computing API interface 3814.

[0247] In an embodiment, quantum computing engine library 3808 includes quantum computing engine configuration 3816 and quantum computing process module 3818 based on various supported quantum models. In an embodiment, quantum computing system 3800 can support many different quantum models, including, but not limited to, the quantum circuit model, quantum Turing machines, adiabatic quantum computers, spintronic computing systems (e.g., using spin-orbit coupling to generate spin-polarized electronic states in non-magnetic solids, such as those using diamond materials), one-way quantum computers, quantum annealing, and various quantum cellular automata. In the quantum circuit model, quantum circuits are based on quantum bits (qubits). A qubit can be in a quantum state of 1 or 0, or in a quantum state in superposition of 1 and 0. However, when a qubit measures the result of a measurement, the qubit is always in a quantum state of 1 or 0. The probabilities associated with these two outcomes depend on which quantum state the qubit was in immediately before the measurement. Computations are performed by manipulating the qubits using quantum logic gates, which are somewhat similar to classical logic gates.

[0248] In embodiments, the quantum computing system 3800 can be physically implemented using analog or digital approaches. Analog approaches include, but are not limited to, quantum simulation, quantum annealing, and adiabatic quantum computing. In embodiments, a digital quantum computer uses quantum logic gates for computation. Both analog and digital approaches can use quantum bits (qubits).

[0249] In an embodiment, the quantum computing system 3800 includes a quantum annealing module 3820, which may be configured to use quantum fluctuations to find a global minimum or maximum of a given objective function over a set of given candidate solutions (e.g., candidate states). As used herein, quantum annealing may refer to a meta-procedure for finding an absolute minimum or maximum, such as size, length, cost, time, distance, or other measure, among a potentially very large but finite set of solutions, using quantum fluctuation computation instead of classical computation. The quantum annealing module 3820 can be used for problems with a discrete search space and many local minima (e.g., combinatorial optimization problems), such as spin glass ground states or the traveling salesman problem.

[0250] In an embodiment, the quantum annealing module 3820 starts with a quantum mechanical superposition of all possible states (candidate states) with equal weights. The quantum annealing module 3820 may then evolve to follow the time-dependent Schrödinger equation, which is the natural quantum mechanical evolution of a system (e.g., a physical system, a logical system, etc.). In an embodiment, the amplitudes of all candidate states change, quantum parallelism is achieved according to the time-dependent transverse field strength, and quantum tunneling between states occurs. If the rate of change of the transverse field is sufficiently slow, the quantum annealing module 3820 may remain near the ground state of the instantaneous Hamiltonian. If the rate of change of the transverse field is accelerated, the quantum annealing module 3820 may temporarily leave the ground state but is more likely to end up at the final problem energy state or ground state of the Hamiltonian.

[0251] In an embodiment, the quantum computing system 3800 can include any number of qubits, transport ions to spatially distinct locations within an array of ion traps, and build large-scale entangled states via a photonically connected network of remotely entangled ion chains.

[0252] In some implementations, the quantum computing system 3800 includes a trapped ion computer module 3822, which may be a quantum computer that applies trapped ions to solve complex problems. The trapped ion computer module 3822 has low quantum decoherency and can construct a large number of solution states. Ions (charged atomic particles) can be trapped and suspended in free space using electromagnetic fields. Quantum bits (qubits) are stored in each ion's stable electronic state, and quantum information can be transferred through the collective quantized motion (Coulomb interaction) of the ions within the shared trap. Lasers can be applied to induce coupling between qubit states (for single-qubit manipulation) or between internal qubit states and external motional states (for entanglement between qubits).

[0253] In some embodiments of the present invention, a conventional computer including a processor, memory, and a graphical user interface (GUI) may be used to provide output from the design, compilation, and execution, and a quantum computing system 3800 may be used to execute the machine instructions. In some embodiments of the present invention, the quantum computing system 3800 may be simulated by a computer program executed by a conventional computer. In such embodiments, a superposition of states of the quantum computing system 3800 may be prepared based on input from initial conditions. Because the initialization operations available on quantum computers can only initialize qubits to either the |0| or |1| state, initialization to a superposition of states is physically impractical. However, for simulation purposes, it may be useful to bypass the initialization process and initialize the quantum computing system 3800 directly.

[0254] In some embodiments, the quantum computing system 3800 provides various quantum data services, including quantum input filtering, quantum output filtering, quantum application filtering, and a quantum database engine.

[0255] In embodiments, quantum computing system 3800 may include quantum input filtering service 3824. In embodiments, quantum input filtering service 3824 may be configured to select whether to run a model on quantum computing system 3800 or on a classical computing system. In some embodiments, quantum input filtering service 3824 may filter data for later modeling on a classical computer. In embodiments, quantum computing system 3800 may provide input to a classical computation platform while filtering unwanted information flowing into the distributed system. In some embodiments, platform 3800 may be trusted through specified experience filtered for intelligent agents.

[0256] In embodiments, a system in a system of systems can include a model or system for automatically determining, based on a set of inputs, whether to deploy quantum computing resources or quantum algorithmic resources to an activity, whether to deploy conventional computing resources and algorithms, or whether to apply a hybrid or combination thereof. In embodiments, inputs to the model or automated system can include demand information, supply information, financial data, energy cost information, capital costs of computing resources, development costs (such as algorithms), energy costs, operational costs (including labor and other costs), performance information of available resources (quantum and conventional), such as using any of the wide variety of simulation techniques described herein and / or in documents incorporated by reference, and / or any of many other datasets that can be used to predict the difference in results between quantum and non-quantum optimization results. Machine-learned models (including DPANN systems) can be trained by deep learning on results or by datasets from human expert judgment to determine what set of resources to deploy given input data for a given request. The model itself may be deployed on quantum computing resources and / or may use quantum algorithms such as quantum annealing to determine whether, where, and when to use quantum systems, classical systems, and / or hybrids or combinations.

[0257] In some embodiments of the invention, quantum computing system 3800 may include quantum output filtering service 3826. In embodiments, quantum output filtering service 3826 may be configured to select a solution from multiple neural network solutions. For example, multiple neural networks may be configured to generate solutions to a particular problem, and quantum output filtering service 3826 may select the best solution from the set of solutions.

[0258] In some embodiments, the quantum computing system 3800 interfaces with and directs the neural network development or selection process. In this embodiment, the quantum computing system 3800 can directly program the neural network weights so that the neural network gives a desired output. This quantum-programmed neural network can operate without the supervision of the quantum computing system 3800, but operates within the expected parameters of the desired computational engine.

[0259] In embodiments, the quantum computing system 3800 includes a quantum database engine 3828. In embodiments, the quantum database engine 3828 is configured with in-database quantum algorithm execution. In embodiments, a quantum query language may be employed to query the quantum database engine 3828. In some embodiments, the quantum database engine may have an embedded policy engine 3830 for prioritizing and / or allocating quantum workflows, including prioritizing query workloads as well as overall priority and comparative advantage of using quantum computing resources. In embodiments, the quantum database engine 3828 may assist with entity recognition by establishing a single identity valid across interactions and touchpoints. The quantum database engine 3828 is configured to perform data matching optimization and intelligent classical computation optimization to match individual data elements. The quantum computing system 3800 may include a quantum data obfuscation system for obfuscating data.

[0260] Quantum computer systems 3800 include, but are not limited to, analog quantum computers, digital computers, and error-correcting quantum computers. Analog quantum computers can directly manipulate interactions between qubits without decomposing these operations into primitive gate operations. In embodiments, quantum computers that can implement analog machines include, but are not limited to, quantum annealers, adiabatic quantum computers, and direct quantum simulators. Digital computers can operate by executing desired algorithms using primitive gate operations on physical qubits. Error-correcting quantum computers are a more robust version of gate-based quantum computers by incorporating quantum error correction (QEC), which emulates noisy physical qubits into stable logical qubits to achieve reliable operation for any computation. Furthermore, quantum information products include, but are not limited to, computational power, quantum predictions, and quantum inventions.

[0261] In some embodiments, the quantum computing system 3800 is configured as an engine that can be used to optimize classical computers, integrate data from multiple sources into decision-making processes, etc. The data integration process can include real-time capture and management of interaction data with extensive tracking capabilities directly and indirectly related to value chain network activity. In embodiments, the quantum computing system 3800 can be configured to accept cookies, email addresses and other contact data, social media feeds, news feeds, event and transaction log data (including transaction events, network events, computational events, and many others), event streams, web crawling results, distributed ledger information (including blockchain updates and state information), results of distributed or federated queries of data sources, streams of data from chat rooms and discussion forums, and many others.

[0262] In an embodiment, quantum computing system 3800 includes a quantum register having a plurality of qubits. Additionally, quantum computing system 3800 may include a quantum control system for performing elementary operations on each qubit in the quantum register, and a control processor for coordinating the required operations.

[0263] In embodiments, the quantum computing system 3800 is configured to optimize pricing for a set of goods or services. In embodiments, the quantum computing system 3800 may utilize quantum annealing to provide optimized pricing. In embodiments, the quantum computing system 3800 may use qubit-based computational methods to optimize pricing.

[0264] In embodiments, the quantum computing system 3800 is configured to automatically discover smart contract configuration opportunities, which may be based on published APIs to the marketplace and machine learning (e.g., via robotic process automation (RPA)) of stakeholders, assets, and transaction types.

[0265] In embodiments, quantum certs or other blockchain-enabled smart contracts enable frequent transactions to occur between a network of parties, eliminating manual or duplicative work performed by counterparties for each transaction. The quantum certs or other blockchains act as a shared database providing a secure, single source of truth, and smart contracts automate approvals, calculations, and other transaction activities that are prone to lag and errors. Smart contracts can use software code to automate tasks, and in some embodiments, this software code can include quantum code, enabling highly optimized results.

[0266] In embodiments, quantum computing system 3800 or other systems within the system may include a quantum-enabled or other risk identification module configured to perform risk identification and / or mitigation. Steps that may be performed by the risk identification module may include, but are not limited to, risk identification, impact assessment, etc. In some embodiments, the risk identification module determines a risk type from a set of risk types. In embodiments, the risk includes, but is not limited to, preventable risks, strategic risks, and external risks. Preventable risks may refer to risks that arise internally and can typically be managed at a rules-based level, such as by monitoring operating procedures and employing employee and management guidance and direction. Strategic risks refer to risks that are voluntarily assumed in order to obtain greater rewards. External risks refer to risks that arise externally and are beyond the company's control (e.g., natural disasters). External risks are neither preventable nor desirable. In embodiments, the risk identification module may determine projected costs for many categories of risk. The risk identification module may perform calculations of current and potential impacts on the overall risk profile. In embodiments, the risk identification module may determine the probability and importance of specific events. Additionally or alternatively, the risk identification module may be configured to predict events.

[0267] In an embodiment, quantum computing system 3800 or other systems of platform 3800 are configured for graph clustering analysis for anomaly and fraud detection.

[0268] In some embodiments, the quantum computing system 3800 includes a quantum prediction module configured to generate predictions. Additionally, the quantum prediction module can build a classical prediction engine to generate further predictions, which can reduce the need for ongoing quantum computing costs compared to classical computers.

[0269] In an embodiment, the quantum computing system 3800 may include a quantum principal component analysis (QPCA) algorithm that can process input vector data if the covariance matrix of the data can be efficiently obtained as a density matrix under certain assumptions about the vectors given in quantum mechanical form. It may be assumed that users have quantum access to training vector data in quantum memory. Furthermore, it may be assumed that each training vector is stored in quantum memory in terms of its difference from the class mean. These QPCA algorithms can be applied to perform dimensionality reduction using the computational advantages of quantum methods.

[0270] In an embodiment, the quantum computing system 3800 is configured for graph clustering analysis to authenticate the randomness of a proof-of-stake blockchain. Quantum cryptography schemes can utilize quantum mechanics in their design, allowing them to rely on potentially unbreakable laws of physics for their security. Quantum cryptography schemes are information-theoretically secure, and their security is not based on any non-fundamental assumptions. Blockchain system designs do not prove information-theoretic security. Rather, classical blockchain technology typically relies on security arguments that make assumptions about the resource limitations of attackers.

[0271] In embodiments, quantum computing system 3800 is configured to detect adversarial systems, such as adversarial neural networks, including adversarial convolutional neural networks. For example, quantum computing system 3800 or other systems of platform 3800 may be configured to detect false trading patterns.

[0272] In an embodiment, the quantum computing system 3800 includes a quantum continuous learning (QCL) system 3832, which enables the autonomous, incremental development of complex skills and knowledge by continuously and adaptively learning about the external world and updating quantum models to account for different tasks and data distributions. The QCL system 3832 operates on realistic timescales where data and / or tasks are only available during operation. Superposition of previous quantum states onto the quantum engine can provide the power of the QCL. Because the QCL system 3832 is not constrained to a finite number of variables that can be processed deterministically, it can continuously adapt to future states, creating a dynamic, continuous learning capability. The QCL system 3832 can have applications where the data distribution remains relatively static, but data is continuously received. For example, the QCL system 3832 can be used in quantum recommendation applications or quantum anomaly detection systems, where data is continuously received and quantum models are continuously refined to provide different outcomes, predictions, etc. QCL enables asynchronous interleaved learning of tasks, updating quantum models only based on real-time data available from one or more streaming sources at a particular moment in time.

[0273] In embodiments, the QCL system 3832 operates in complex environments where target data continues to change based on uncontrolled hidden variables. In embodiments, the QCL system 3832 can scale in intelligence while processing increasing amounts of data and maintaining a realistic number of quantum states. The QCL system 3832 applies quantum techniques to significantly reduce historical data storage requirements while enabling continuous computations to deliver detail-driven optimal results. In embodiments, the QCL system 3832 is configured for unsupervised streaming perceptual data, as it continuously updates its quantum model with new available data.

[0274] In embodiments, the QCL system 3832 enables multimodal and multitask quantum learning. The QCL system 3832 is not constrained to a single stream of sensory data, but allows for many streams of sensory data from different sensors and input modalities. In embodiments, the QCL system 3832 can solve multiple tasks by replicating quantum states and performing computations on the replicated quantum environment. A key advantage of QCLs is that the superposition state retains information relevant to all past inputs, eliminating the need to retrain quantum models with past data. Multimodal and multitask quantum learning facilitates quantum optimization by empowering quantum machines with inference capabilities through the application of vast state information.

[0275] In an embodiment, quantum computing system 3800 supports quantum superposition, i.e., the ability to superimpose a set of states onto a single quantum environment.

[0276] In embodiments, quantum computing system 3800 supports quantum teleportation, e.g., passing information between photons on a chipset without the photons being physically linked.

[0277] In embodiments, the quantum computing system 3800 may include a quantum transfer pricing system. Quantum transfer pricing allows for establishing prices for goods and / or services exchanged between subsidiaries, affiliates, or commonly controlled companies that are part of a larger enterprise and may be used to provide tax savings to the enterprise. In embodiments, resolving transfer pricing problems involves testing the resiliency of each system within the system in a series of tests. In these embodiments, tests may be performed periodically in batches and then repeated. As described herein, transfer prices may refer to the prices that one division within a company charges another division within the company for goods or services.

[0278] In embodiments, the quantum transfer pricing system continuously consolidates all transfer pricing relevant financial data for all entities of an organization throughout the year, where the consolidation includes applying quantum entanglement to superimpose the data into a single quantum state. In embodiments, the financial data includes profit data, loss data, data from intercompany invoices (which may include quantities and prices), etc.

[0279] In embodiments, the quantum transfer pricing system may interface with a reporting system that reports segmented profits and losses, transaction matrices, tax optimization results, etc. based on the overlay data. In embodiments, the quantum transfer pricing system automatically generates forecast calculations to assess expected local profits for any set of quantum states.

[0280] In an embodiment, the quantum transfer pricing system can be integrated with a simulation system for running simulations. Through the integrated quantum workflow and quantum teleportation communication state, optimal new product prices can be proposed and discussed across borders.

[0281] In an embodiment, quantum transfer pricing is used to actively manage profit allocation within a multinational enterprise (MNE).

[0282] In embodiments, the QCL system 3832 can use a number of methods to calculate the quantum transfer price, including the quantum comparative uncontrolled price (QCUP) method, the quantum cost plus percentage method (QCPM), the quantum resale price method (QRPM), the quantum transaction net margin method (QTNM), and the quantum profit split method.

[0283] The QCUP method applies quantum computing to find comparable transactions between related and unrelated organizations by sharing quantum superposition data. Benchmark prices can be determined by comparing the prices of goods and / or services in business-to-business transactions with prices used by independent parties using a quantum comparison engine.

[0284] The QCPM method measures cost-plus markup (the actual profit derived from a product) by comparing gross margin to cost of goods sold. Once this markup is determined, it should equal the profit a third party would make from an equivalent transaction under similar external market conditions. In embodiments, the Quantum Engine can simulate external market conditions.

[0285] The QRPM method focuses on groups of transactions rather than individual transactions and is based on the gross margin or difference between the purchase price of a product and the sale price to a third party. In embodiments, a quantum engine can be applied to calculate the price difference and record the transaction in the superimposed system.

[0286] The QTNM method is based on the net profit of a managed transaction, rather than comparable external market prices. The calculation of net profit is accomplished by a quantum engine that can optimally solve for product prices, taking into account various factors. That net profit may then be compared to the net profits of independent companies using quantum teleportation.

[0287] The quantum profit split method is used when two associated companies separately conduct the same business. In such cases, quantum transfer pricing is calculated based on profits. The quantum profit split method applies quantum calculations to determine how the profits associated with a particular transaction are divided between independent parties.

[0288] In embodiments, the quantum computing system 3800 may utilize one or more artificial networks to fulfill the requests of quantum computing clients. For example, the quantum computing system 3800 may utilize a set of artificial neural networks to identify patterns in images (e.g., using image data from a liquid lens system), perform binary matrix factorization, perform topical content targeting, perform similarity-based clustering, perform collaborative filtering, perform opportunity mining, or the like.

[0289] In embodiments, the system of systems may include a hybrid computing allocation system for prioritizing and allocating quantum computing resources and conventional computing resources. In embodiments, the prioritization and allocation of quantum computing resources and conventional computing resources may be measurement-based (e.g., measuring the degree of dominance of quantum resources over other available resources), cost-based, optimality-based, speed-based, impact-based, etc. In some embodiments, the hybrid computing allocation system is configured to perform time-division multiplexing between the quantum computing system 3800 and conventional computing systems. In embodiments, the hybrid computing allocation system may automatically track and report computational resource allocation, computational resource availability, computational resource costs, etc.

[0290] In an embodiment, the quantum computing system 3800 can be utilized for queue optimization for quantum computing resource utilization, including context-based queue optimization.

[0291] In an embodiment, the quantum computing system 3800 may support quantum-aware location-based data caching.

[0292] In embodiments, quantum computing system 3800 can be utilized for optimization of various system resources within a system, including optimization of quantum computing resources, conventional computing resources, energy resources, human resources, robotic fleet resources, smart container fleet resources, I / O bandwidth, storage resources, network bandwidth, attention resources, etc.

[0293] The quantum computing system 3800 has all its functionality available and available as part of a configured quantum computing service, which may consist of a subset of these functionality and perform certain predefined functions, generate newly defined functions, or perform various combinations of both.

[0294] 39 illustrates quantum computing service request processing according to some embodiments of the present disclosure. A directed quantum computing request 3902 may come from one or more quantum-aware devices or stacks of devices, the request is for a known application configured with a particular quantum instance(s), quantum computing engine(s), or other quantum computing resource, and data associated with the request may be pre-processed or otherwise optimized for use in quantum computing.

[0295] A general quantum computing request 3904 may come from any system within the system or configured services where the requester determines that quantum computing resources may provide added value or other improved results. The quantum computing service may also perform some form of monitoring or analysis to suggest improved results. In a general quantum computing request 3904, the input data may not be structured or formatted as required for quantum computing.

[0296] In embodiments, external data requests 3906 may include any available data that may be needed to train a new quantum instance. The sources of such requests may be public data, sensors, ERP systems, and many others.

[0297] The received operation request and associated data may be analyzed using a standardized approach to identify one or more possible sets of known quantum instances, quantum computing engines, or other quantum computing resources that may be applied to perform the requested operation. Possible existing sets may be identified in a quantum set library 3908.

[0298] In an embodiment, quantum computing system 3800 includes a quantum computing configuration service 3810. The quantum computing configuration service, alone or in cooperation with intelligence service 3834, can select the best available configuration using resource and priority analysis, including the requester's priorities. The quantum computing configuration service either provides a solution (YES) or determines that a new configuration is needed (NO).

[0299] In some examples, the requested set of quantum computing services may not exist in quantum set library 3908. In this example, one or more new quantum instances must be developed (trained) in conjunction with intelligence services 3834 using the available data. In embodiments, alternative configurations may be developed with the assistance of intelligence services 3834 to identify alternative ways of providing all or part of the requested quantum computing services until appropriate resources are available. For example, a hybrid quantum / classical model is possible, where the requested services may be provided at a slower rate.

[0300] In embodiments, information services 3834 may assist in developing alternative configurations and identifying alternative, possibly temporary, ways of providing all or part of the requested quantum computing service. For example, a hybrid quantum-classical model may be envisioned that provides the requested service at a slower rate. It may also include a feedback learning loop to adjust the service in real time and improve the stored library elements.

[0301] Once a quantum computing configuration is identified and available, it is allocated and programmed to execute and deliver one or more quantum states (solutions). Biologically Based Systems, Methods, Kits, and Devices

[0302] 40 and 41 show a thalamic service 4000 together with a set of input sensors streaming data from various sources throughout a system 4002 with a centralized data source 4004. The thalamic service 4000 filters the data to the control system 4002 so that the control system is not overwhelmed with the amount of information. In an embodiment, the thalamic service 4000 provides an information throttling mechanism for the information flow within the system. This mechanism monitors all data streams and removes irrelevant data streams by ensuring that the maximum data flow from all input sensors is constrained at all times.

[0303] The thalamus service 4000 may be the gateway for all communications responsive to the prioritization of the control system 4002. The control system 4002 may decide to change the prioritization of data streaming from the thalamus service 4000, for example, during a known fire in an isolated area, this event may instruct the thalamus service 4000 to continue providing flame sensor information even though the majority of this data is not anomalous. The thalamus service 4000 may be an integral part of the overall system communication framework.

[0304] In an embodiment, the thalamus services 4000 include an ingestion management system 4006. The ingestion management system 4006 may be configured to receive and process multiple large data sets by converting them into sized and organized data streams for subsequent use by one or more central control systems 4002 operating within the system. For example, a robot may include vision and sensing systems that its central control system 4002 uses to identify and navigate its environment in real time. The ingestion management system 4006 may facilitate the robot's decision-making by analyzing, filtering, classifying, or otherwise reducing the size and increasing the usefulness of multiple large data sets that would otherwise overwhelm the central control system 4002. In an embodiment, the ingestion management system may include an ingestion controller 4008 that, in cooperation with the intelligence service 4010, evaluates the incoming data and takes action-based evaluation results. The evaluation and action may include a specific set of instructions received by the thalamus services 4000, such as using a specific set of compression and prioritization tools defined in the "Networking" library module. In another example, the thalamic service inputs can direct the use of specific filtering and suppression techniques. In a third example, the thalamic service inputs can specify data filtering related to areas of interest, such as specific types of financial transactions. The ingestion management system can also be configured to recognize and manage datasets that are in vectorized formats, such as PCMP, and either passed directly to a central control or disassembled and processed separately. The ingestion management system 4006 can include a learning module that receives data from external sources, enabling the improvement and creation of application and data management library modules. In some cases, the ingestion management system can request external data to augment existing datasets.

[0305] In an embodiment, the control system 4002 can instruct the thalamic service 4000 to change its filtering to provide more input from a particular set of sources, causing the thalamic service 4000 to process more input by throttling other information flows, thereby constraining the total data flow within the capacity of the central control system.

[0306] The thalamus service 4000 can operate by suppressing data based on several different factors, and in an embodiment, the default factor may be data anomaly, which involves constantly monitoring all input sensors to determine if the data is anomaly.

[0307] In some embodiments, the thalamic service 4000 can suppress data based on geospatial factors. The thalamic service 4000 can be aware of the geospatial location of all sensors and can look for anomalous patterns in the data based on the geospatial context and suppress the data accordingly.

[0308] In some embodiments, the thalamic service 4000 can throttle data based on temporal factors. Data can be throttled temporally, for example, when the cadence of the data can be reduced so that the overall data stream is filtered to a level that can be processed by a central processing unit.

[0309] In some embodiments, the thalamic service 4000 can suppress data based on contextual factors. In embodiments, context-based filtering is filtering events where the thalamic service 4000 recognizes some context-based event. In this context, filtering is performed to suppress information flow that is not related to data from the event.

[0310] In an embodiment, the control system 4002 may decide to override the thalamic filtering and focus on an entirely different region for a particular reason.

[0311] In embodiments, the system may include a vector module. In embodiments, the vector module may be used to convert data into a vectorized format. In many instances, converting long sequences of similar numbers into vectors that may contain short-term future predictions reduces the size of communications and is inherently forward-looking. In embodiments, forecasting methods may include: moving averages; weighted moving averages; Kalman filters; exponential smoothing; autoregressive moving averages (ARMA) (where predictions depend on past values ​​of the variable being predicted and past prediction errors); autoregressive integrated moving averages (ARIMA) (where ARMA relates to period-to-period changes in the variable being predicted); extrapolation; linear forecasting; trend estimation (predicting a variable as a linear or polynomial function of time); growth curves (e.g., statistics), and recurrent neural networks.

[0312] In an embodiment, the system can include a Predictive Model Communication Protocol (PMCP) system that supports vector-based predictive models. Under the PMCP protocol, instead of traditional streams of individual data items, vectors are communicated, representing how the data is changing or what the predicted trends of the data are. The PMCP system can transmit actual model parameters and receiving units so that edge devices can apply the vector-based predictive model to determine future states. For example, each automated device in the network can train a regression model or neural network to constantly adapt the data stream to current input data. All automated equipment utilizing the PMCP system can react before an event actually occurs, rather than waiting for a certain product to run out of stock, for example. Continuing with this example, stateless automated equipment can react to the predicted future state and make any necessary adjustments, such as ordering more of that item.

[0313] In embodiments, the PMCP system allows for the communication of vectorized information with algorithms that process the vectorized information to refine known information about a state based on a set of probabilities. For example, the PMCP system supports the communication of vectorized information collected at each point of sensor reading, but can also add algorithms that process the information. When applied to environments with numerous sensors of varying accuracy and reliability, the PMCP system's probabilistic vector-based mechanism allows for the combination of numerous, if not all, data streams to generate a refined model that represents the current state, past state, and likely future states of the item. Approximation methods may include importance sampling, and the resulting algorithms are known as particle filters, condensation algorithms, or Monte Carlo localization.

[0314] In embodiments, the vector-based communication of the PMCP system allows future security events to be predicted, for example, by a simple edge node device operating semi-autonomously. The edge device can be responsible for building a set of predictive models that represent trends in the data. The parameters of this set of predictive models can be transmitted using the PMCP system.

[0315] Security systems are constantly looking for vectors that indicate a change in state, as anomalous events tend to trigger multiple vectors, indicating an abnormal pattern. In a security environment, seeing multiple anomalous vectors simultaneously can trigger escalation and response by control systems, for example. Furthermore, one of the primary concerns in communications security is the protection of stored data.

[0316] In embodiments, PMCP data can be stored directly in a queryable database where the actual data is dynamically reconstructed in response to a query. In some embodiments, the PMCP data stream can be used to recreate granular data to be part of an ETL (Extract Transform and Load) process.

[0317] In embodiments where edge devices with very limited capacity are present, an edge communications device can be added to convert data into PMCP format. For example, to protect distributed medical devices from hacking, many manufacturers choose not to connect the devices to any kind of network. To overcome this limitation, medical devices can be monitored using sensors such as cameras, sound monitors, voltage detectors for power usage, and chemical sniffers. Functional unit learning and other data techniques may be used to determine the actual usage of medical devices that are disconnected from network functional units.

[0318] Future state is kept in perspective through communication using vectorized data, which allows future state to be communicated and allows various entities to proactively address future state requirements without needing access to fine-grained data.

[0319] In embodiments, the PMCP protocol can be used to communicate relevant information about production levels and future production trends. This PMCP data feed, with built-in data obfuscation, allows for real-world contextual information about production levels to be shared with consumers, regulators, and other parties without sharing sensitive data. For example, if a consumer is purchasing a new car and red paint is in short supply, they can be advised to choose a different color to maintain the desired delivery date. PMCP and vector data enable the creation of interactive systems based on simple data information without building complex big data engines. For example, consider an upstream manufacturer with the highly complex task of coordinating many downstream consumer locations. Using PMCP, manufacturers can provide realistic information to consumers without storing detailed data or building complex models.

[0320] In embodiments, edge device units may communicate via the PMCP system to indicate direction of movement and likely future locations, for example, a moving robot may communicate its likely future trajectory.

[0321] In embodiments, the PMCP system enables a visual representation of vector-based data (e.g., via a user interface) to highlight areas of concern without the need to process vast amounts of data. This representation allows for many monitored vector inputs to be displayed. The user interface can then display information related to key items of interest, particularly vectors that indicate anomalous or problematic movement. This mechanism allows advanced models built at edge nodes on edge devices to be reflected in end-user communications in a visually informative manner.

[0322] Functional units constantly produce "boring" data. Shifting from generating data to monitoring issues highlights problems in logistics modules without the need to scrutinize detailed data. In embodiments, the vectorization process can constantly manage predictive models that indicate future conditions. In a maintenance context, these changes to the predictive model's parameters are themselves predictors of changes in operational parameters, potentially indicating the need for maintenance. In embodiments, functional areas are not always designed to be connected, but by allowing external devices to virtually monitor devices, functional areas that do not allow connectivity can become part of the information flow of goods. This concept extends to effectively monitoring functional areas with limited connectivity by decorating the data stream with vectorized monitoring information. Placing automated devices near functional units with limited or no connectivity can capture information from the devices without requiring connectivity. There is also the possibility of adding training data capture functional units for such unconnected or limited-connectivity functional areas. These training data capture functional units are typically very expensive and can provide high-quality monitoring data, which is used as input to nearby edge device monitoring devices to provide data for supervised learning algorithms.

[0323] Locations often have a lot of electrical interference, creating fundamental communication challenges. Traditional approaches that stream all the granular data rely on the integrity of the data stream. For example, if an edge device goes offline for 10 minutes, the streaming data and its information is lost. With vectored communication, the offline unit can continue to refine its predictive model until it reconnects, and then transmit the updated model via the PMCP system.

[0324] In embodiments, systems and devices can be based on the PMCP protocol. For example, cameras and vision systems (e.g., liquid lens systems), user devices, sensors, robots, smart containers, etc., may use PMCP and / or vector-based communication. Using vector-based cameras, for example, only information about the movement of an item is transmitted. This reduces the amount of data and, by its nature, filters out information about static items, displaying only changes in the image and focusing data communication on the elements of change. This shift in overall communication to communication of changes is similar to how the human visual process works, where stationary items are not even communicated to higher levels of the brain.

[0325] Radio frequency identification allows for real-time tracking of large numbers of moving tags. In embodiments, tag movements may be communicated as vector information via the PMCP protocol, as this form of communication is naturally suited to passing information about the location of tags within merchandise. By adding the ability to indicate the future state of a location using a predictive model that can use previous movement paths, the item can change its basic communication mechanism to one in which units consuming the data stream consume information about the likely future state of the item. In embodiments, each tagged item can be represented as a probability-based location matrix indicating the likelihood that the tagged item is at a certain location in space. The propagation of movement indicates the transformation of the location probability matrix into a new set of probabilities. This probabilistic location view provides for constant modeling of areas likely to be intersected by the moving unit, allowing for the refinement of the probabilistic view of the item's location. By moving to a vector-based probability matrix, the unit can constantly handle the inherent uncertainty in measuring the status of various items, entities, etc. In embodiments, status includes, but is not limited to, location, temperature, movement, and power consumption.

[0326] In embodiments, continuous connection is not required for continuous monitoring of sensor inputs in a PMCP-based communication system. For example, a mobile robotic device with multiple sensors can continue to build models and predictions of a data stream while disconnected from the network, and upon reconnection, updated models are communicated. Furthermore, other systems or devices that use inputs from the monitored system or device can apply the best-known, typically last-communicated, vector prediction to continue to maintain a probabilistic understanding of the state of the item. Additional Exemplary Embodiments AI-based Energy Edge Platform

[0327] In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the system having a system for communicating data across a set of nodes in a network, each node adapted to operate on an energy dataset of energy generation, storage, or consumption data, the set of nodes configured with at least one of an algorithm or rule set for filtering, compressing, or routing the energy dataset based on at least one of network conditions, network error correction requirements, data size, data granularity, or data content. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the system having a system for communicating data across a set of nodes in a network, at least a subset of the nodes configured with at least one rule or algorithm adapted to set at least one parameter of the data communication based on a set of indicators of current network conditions to optimize energy used in the data communication. In an embodiment, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, comprising a system for communicating data across a set of nodes in a network, at least a subset of the nodes configured with at least one rule or algorithm adapted to set routing instructions for data communications based on a set of indicators of current network conditions to optimize energy used in the data communications.In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform comprising a system for communicating data across a set of nodes in a network, wherein at least a subset of the nodes are configured with at least one rule or algorithm adapted to set path parameters for the data communication based on a set of indicators of current network conditions to optimize energy used in the data communication. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform comprising a system for communicating data across a set of nodes in a network, wherein at least a subset of the nodes are configured with at least one rule or algorithm adapted to set error correction parameters for the data communication based on a set of indicators of current network conditions to optimize energy used in the data communication. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform comprising a system for communicating data across a set of nodes in a network, wherein at least a subset of the nodes are configured with at least one rule or algorithm adapted to set compression parameters for the data communication based on a set of indicators of current network conditions to optimize energy used in the data communication.In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform comprising a system for communicating data across a set of nodes in a network, at least a subset of the nodes configured with at least one rule or algorithm adapted to set storage parameters for data communication based on a set of indicators of current network conditions to optimize energy used in the data communication. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform comprising a system for communicating data across a set of nodes in a network, at least a subset of the nodes configured with at least one rule or algorithm adapted to set timing parameters for data communication based on a set of indicators of current network conditions to optimize energy used in the data communication. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform comprising an edge device artificial intelligence system for operating on data communicated via the edge device to collectively optimize energy used by the edge device and by a set of systems controlled by the edge device. In embodiments, provided herein is an AI-based platform for enabling intelligent power and energy orchestration and management, having a system for automated and coordinated governance or provisioning of a set of grid energy equipment and a set of distributed edge energy resources that are electrically independent from the grid.In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, comprising a system for automatic discovery of energy generation or storage resources that are electrically independent from the electric grid, in data communication with a set of grid energy equipment and a system for coordinated governance or provisioning of a set of distributed edge energy resources that are electrically independent from the grid. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, comprising a system for automatic discovery of energy generation or storage resources that are electrically independent from the electric grid, in data communication with a set of grid energy equipment and a system for coordinated governance or provisioning of a set of distributed edge energy resources that are electrically independent from the grid, wherein the automatic discovery of the grid-independent resources is through artificial intelligence processing of a dataset. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, comprising a system for automated discovery of energy generation or storage resources that are electrically independent from the electric grid in data communication with a set of grid energy equipment and a system for coordinated governance or provisioning of a set of distributed edge energy resources that are electrically independent from the grid, wherein the automated discovery of the grid independent resources is through natural language processing of social data content.In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, comprising a system for automatic discovery of energy generation or storage resources that are electrically independent of the electric grid in data communication with a set of grid energy equipment and a system for coordinated governance or provisioning of a set of distributed edge energy resources that are electrically independent of the grid, wherein the automatic discovery of the grid-independent resources is through computer vision processing of satellite imagery content or web imagery content. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, comprising a system for automatic discovery of energy generation or storage resources that are electrically independent of the electric grid in data communication with a set of grid energy equipment and a system for coordinated governance or provisioning of a set of distributed edge energy resources that are electrically independent of the grid, wherein the automatic discovery of the grid-independent resources is through automatic processing of a set of energy transaction logs. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the AI-based platform having a system for automated discovery of energy generation or storage resources that are electrically independent from the electric grid, in data communication with a set of grid energy equipment and a system for coordinated governance or provisioning of a set of distributed edge energy resources that are electrically independent from the grid, wherein the discovery of the grid-independent resources is done by applying an artificial intelligence system trained on historical training datasets of grid and off-grid energy patterns to recognize the existence of off-grid energy resources.In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, having an artificial intelligence system operating on a dataset of energy generation, storage, or consumption data for a set of infrastructure assets, generated at least in part by a set of sensors included in or governed by the set of edge devices, to generate output operating parameters for energy generation by at least one of the legacy infrastructure assets. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, having an artificial intelligence system operating on a dataset of energy generation, storage, or consumption data for a set of infrastructure assets, generated at least in part by a set of sensors included in or governed by the set of edge devices, to generate output operating parameters for energy storage for at least one of the legacy infrastructure assets. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, having an artificial intelligence system operating on a dataset of energy generation, storage, or consumption data for a set of infrastructure assets, the dataset being generated at least in part by a set of sensors included in or managed by the set of edge devices, to generate output operating parameters for energy consumption by at least one of the legacy infrastructure assets. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform having a set of edge devices for collecting energy generation data for the set of infrastructure assets based on a set of sensors included in or governed by the edge devices. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform having a set of edge devices for collecting energy storage data for the set of infrastructure assets based on a set of sensors included in or governed by the edge devices. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform having a set of edge devices for collecting energy consumption data for the set of infrastructure assets based on a set of sensors included in or governed by the edge devices. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform having a set of adaptive and autonomous data processing systems for collecting and transmitting energy edge data. In an embodiment, provided herein is an AI-based platform for enabling intelligent power and energy orchestration and management, having a set of adaptive and autonomous data processing systems for energy edge data collection and transmission, wherein the data processing systems are trained with a training set of data to recognize a set of events or signals indicative of energy consumption by a set of edge devices or a set of systems connected to or controlled by the edge devices.In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform having a set of adaptive and autonomous data processing systems for collecting and transmitting energy edge data, the data processing systems being trained based on a set of training data to recognize a set of events or signals indicative of energy accumulation by a set of edge devices or a set of systems connected to or controlled by the edge devices. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform having a set of adaptive and autonomous data processing systems for collecting and transmitting energy edge data, the data processing systems being trained based on a set of training data to recognize a set of events or signals indicative of energy generation by a set of edge devices or a set of systems connected to or controlled by the edge devices. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, the platform having a digital twin of a mine with mine-level sensing of a set of parameters represented in the digital twin. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, having a digital twin of a drilling operation with drilling platform-level sensing of a set of parameters represented in the digital twin. In embodiments, provided herein is an AI-based platform for enabling intelligent orchestration and management of power and energy, having a digital twin of a hydropower system with hydropower system-level sensing of a set of paramete...

Claims

1. An AI-based platform that enables intelligent orchestration and management of power and energy, 1. An artificial intelligence system trained with a set of energy generation, energy storage, energy supply, and / or energy consumption outcomes, the artificial intelligence system comprising: analyzing a dataset of current energy generation, current energy storage, current energy supply, and / or current energy consumption information; and providing a recommendation comprising at least one operating parameter that satisfies both a mobile energy need or a fixed-location energy need in a defined area.

2. The AI-based platform of claim 1 , wherein the defined domain comprises a defined geographic location and a defined time period.

3. 10. The AI-based platform of claim 1, wherein the operational parameters indicate production instructions for the set of energy-generating resources.

4. 10. The AI-based platform of claim 1, wherein the operational parameters indicate storage instructions for a set of energy storage resources.

5. 10. The AI-based platform of claim 1, wherein the operational parameters indicate delivery instructions for a set of energy delivery resources.

6. The AI-based platform of claim 1 , wherein the operational parameters indicate consumption instructions for a set of energy-consuming entities.

7. 10. The AI-based platform of claim 1, wherein the artificial intelligence system is further configured to adapt transport of data over a network and / or a communication system, the adaptation comprising: Congestion, Delay and waiting conditions, Packet loss condition, Error rate conditions, transportation costs, Quality of Service (QoS) requirements, Terms of use, market factors conditions, or User-defined conditions, AI-based platforms based on one or more of the following:

8. energy stakeholder groups; Energy distribution resources, Stakeholder Information Technology; network infrastructure entities; Energy-dependent stakeholders' production facilities; Stakeholder transportation systems; market conditions; Energy use preferences, or 10. The AI-based platform of claim 1, further comprising an adaptive energy digital twin representing one or more of:

9. providing a visual and / or analytical indication of energy consumption by one or more energy consumers; filtering of energy data, Highlighting energy data, or Energy data adjustment, 10. The AI-based platform of claim 1, further comprising an adaptive energy digital twin configured to perform one or more of the following:

10. One or more machines, one or more factories, or one or more vehicles in a vehicle fleet; 10. The AI-based platform of claim 1, further comprising an adaptive energy digital twin configured to generate visual and / or analytical indicators of energy consumption by one or more of:

11. 10. The AI-based platform of claim 1, wherein the artificial intelligence system further comprises: Extracting energy-related data; detecting and / or correcting errors in energy-related data; Transforming, converting, normalizing, and / or cleansing energy-related data; Analysis of energy-related data, Detecting patterns, content, and objects in energy-related data; Compressing energy-related data, Streaming Energy Data filtering energy-related data; Loading and / or storing energy-related data; Routing and / or transmission of energy-related data; or Maintaining the security of energy-related data; 1. An AI-based platform configured to perform one or more of the following:

12. 2. The AI-based platform of claim 1, wherein the dataset is based on one or more public data resources, the public data resources comprising: Weather data resources, satellite data resources, census, demographic, demographic, and / or psychographic data resources; Market Data Resources, or e-commerce data resources, AI-based platforms, including one or more of the following:

13. 2. The AI-based platform of claim 1, wherein the dataset is based on one or more enterprise data resources, the enterprise data resources comprising: resource planning data, Sales and / or marketing data, Financial planning data, Demand planning data, supply chain data, Procurement data, Price data, Customer data, Product data, or Operational data, AI-based platforms, including one or more of the following:

14. The artificial intelligence system is trained based on a training data set, the training data set comprising: one or more human tags and / or labels; the interaction of a hardware and / or software system with one or more humans; one or more results, one or more AI-generated training data samples; The supervised learning training process, Semi-supervised learning training process, or Deep learning training process, The AI-based platform of claim 1, wherein the AI-based platform is based on one or more of:

15. 10. The AI-based platform of claim 1, wherein the artificial intelligence system is further configured to orchestrate delivery of energy to one or more points of consumption, the delivery of energy comprising: one or more fixed transmission lines; one or more instances of wireless energy transmission; one or more deliveries of fuel; or supplying one or more stored energy sources; AI-based platforms, including one or more of the following:

16. 10. The AI-based platform of claim 1, wherein the artificial intelligence system is further configured to record one or more energy-related events in a distributed ledger and / or blockchain, the one or more energy-related events comprising: Energy buying and selling events, service charges related to energy purchase and / or sales events; Energy consumption events, Power generation events, Energy distribution events, Energy storage events, carbon emission events, Carbon Emissions Reduction Events, Renewable Energy Credit Event, the occurrence of pollution, or pollution abatement events, AI-based platforms, including one or more of the following:

17. 2. The AI-based platform of claim 1, wherein the artificial intelligence system is deployed in an off-grid environment, the off-grid environment comprising: Off-grid energy generation systems, Off-grid energy storage systems, or Off-grid energy mobilization systems, AI-based platforms, including one or more of the following:

Images