Adaptive Power Grid Management System
The adaptive power grid management system uses machine learning to detect and respond to network disruptions, enhancing reliability and resilience by dynamically optimizing grid operations and forming self-adjusting cells.
Patent Information
- Application Number
- JP2024522485
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-04-06
- Filing Date
- 2022-10-17
- Publication Date
- 2024-11-13
AI Technical Summary
Complex, geographically distributed device networks such as power grids are susceptible to service interruptions due to various conditions like weather, sabotage, aging, or malfunctioning devices, making it difficult to identify and remediate these issues in a timely manner.
An adaptive power grid management system utilizing machine learning algorithms to train context models, detect alert conditions, and deploy scout applications on field agent devices to dynamically adjust and optimize grid operations in response to changing conditions, including physical stress, environmental hazards, and other disruptions.
Enhances system reliability, stability, and resilience by providing dynamic observability, faster response times, and optimized grid formation plans, reducing vulnerabilities through decentralized control and adaptive formation of operating cells.
Smart Images

Figure 2024541847000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates generally to managing a network of devices. [Background technology]
[0002] Complex, geographically distributed device networks, such as the electric power grid (GRID), are often particularly susceptible to a variety of conditions that can lead to service interruptions due to the sheer number of components in the network. External causes, such as weather or sabotage, and internal causes, such as aging or malfunctioning devices, can be difficult to identify, troubleshoot, and remediate in a timely manner. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] US Patent Application Publication No. 2018 / 0262525 [Brief description of the drawings]
[0004] Disclosed herein are embodiments of systems and methods for providing adaptive device network management. This specification includes the following drawings: [Figure 1] FIG. 1 is a system diagram of a power grid system according to some embodiments. [Diagram 2] FIG. 1 is a block diagram of an adaptive power grid management system according to some embodiments. [Diagram 3] FIG. 1 is a block diagram of an adaptive power grid management system and operating cells according to some embodiments. [Figure 4] FIG. 1 is a block diagram of a power grid system according to some embodiments. [Diagram 5]FIG. 1 is a flow diagram of a network management system according to some embodiments. [Figure 6] FIG. 2 is a flow diagram of an operating cell according to some embodiments. [Figure 7] FIG. 1 is a block diagram of a network management system according to some embodiments. [Figure 8] FIG. 1 is a block diagram of modules of an adaptive power grid management system and operating cells according to some embodiments. [Figure 9A] FIG. 2 is a flow diagram of modules of a network management system according to some embodiments. [Figure 9B] FIG. 2 is a flow diagram of modules of a network management system according to some embodiments. [Figure 10A-1] FIG. 2 is a flow diagram of modules of a network management system according to some embodiments. [Figure 10A-2] FIG. 2 is a flow diagram of modules of a network management system according to some embodiments. [Figure 10B] FIG. 2 is a flow diagram of modules of a network management system according to some embodiments. [Figure 11] FIG. 2 is a flow diagram of modules of a network management system according to some embodiments.
[0005] Elements in the figures are illustrated for simplicity and clarity and are not necessarily drawn to scale. For example, the dimensions and / or relative positions of some elements in the figures may be exaggerated relative to other elements to improve understanding of the various embodiments of the present invention. Also, common but well-understood elements that are useful or necessary in commercially feasible embodiments are often not depicted in order to less obscure the view of these various embodiments of the present invention. Although certain acts and / or steps may be described or depicted in a particular sequence of occurrence, those skilled in the art will understand that such specificity with respect to sequence is not actually required. Terms and expressions used in this specification have the ordinary technical meanings as given to such terms and expressions by those skilled in the art of the above-mentioned technical field, unless a different specific meaning is otherwise specified in this specification. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0006] Generally speaking, in accordance with various embodiments, systems, apparatus, and methods are provided herein for adaptively managing a network of devices.
[0007] In some embodiments, an adaptive power grid management system is provided. The system includes a network device database, a network adapter, and a processor. The processor is configured to train a context model with a machine learning algorithm using historical device signals and historical context information related to the network of devices, receive signals from the network of devices, determine context data related to a current state of the network of devices based on the context model, and determine a formation plan based on the context data, the formation plan including tasks to be performed by one or more devices in the network of devices, configure one or more scout applications based on the formation plan and device information stored in the network device database, and cause the one or more scout applications to be executed by the one or more devices in the device network via the network adapter.
[0008] In some embodiments, an operating cell device for power grid management is provided, the device comprising: a network adapter configured to communicate with a device network management system; and a processor coupled to the network adapter, the processor configured to receive, via the network adapter, a scout application having an associated role assignment from a server, authenticate the scout application, automatically execute the scout application upon authentication, and communicate, via the network adapter, with the device network management system or a plurality of devices in a network based on the role assignment of the scout application.
[0009] 1, a power grid with an adaptive grid management system is shown including a grid operation data center 110, an enterprise data center 120, cloud services 130, an adaptive grid management system 140, and a number of field agent devices 170, each configured to monitor and / or control control cells coupled to power grid components and / or subsystems.
[0010] The grid operations data center 110 may be comprised of a transmission and distribution management system, a distributed energy resource (DER) and renewable management system, a grid visibility and insight system, a historian system, and a grid devices and asset services system that communicate over an application integration platform. The components of the grid operations data center 110 may communicate with the enterprise data center 120 and the system operator 150 over a network to perform their respective functions. The enterprise data center 120 is generally comprised of enterprise applications for enterprise-level management. The system operator 150 may include external systems such as systems related to the power market, an electric power transmission system operator (TSO), an independent system operator (ISO), and a distributed system operator (DSO). The adaptive grid management system 140 can aggregate data from a system operator 150 , a grid operations data center 110 , and an enterprise data center 120 .In some embodiments, the adaptive power grid management system 140 can further aggregate data from cloud services 130 (including cloud applications and analytics and cloud service delivery) that aggregate power usage information, virtual power plants 180, and demand response / distributed energy response / electric vehicle (DR / DER / EV) aggregators 160.
[0011] The adaptive grid management system 140 is generally configured to aggregate data from devices and components of the network to detect alert conditions and determine a grid formation plan in response to the alert conditions. In some embodiments, the grid formation plan is determined based on artificial intelligence (AI) machine learning (ML) algorithms that compare contextual data associated with the alert conditions to historical contextual patterns to determine formations that will affect the formation of the grid. The adaptive grid management system 140 may be further configured to form and deploy a scouting application that runs on an operating cell, such as the field agent device 170, to execute the formation plan. In some embodiments, the adaptive grid management system 140 is further configured to receive additional data from the scouting application that runs on the field agent device 170 and generate a new formation plan based on updated information from the grid.
[0012] In some embodiments, the adaptive power grid management system 140 can automatically and contextually adapt the power grid to a new self-defined optimize operational formation in response to changing power system operating conditions. The operating conditions can include physical stress, environmental hazards, social unrest, sabotage, pandemics, and shelter-in-place operating conditions. In some embodiments, the adaptive power grid management system 140 includes a self-forming multi-dimensional, contextual, and cognitive state machine configured to drive the formation and operation of a self-acting grid. In some embodiments, the adaptive grid management system 140 includes a self-learning decision support system, called a grid-wide-mind module, that is configured to provide dynamic, condition-based context data and contextual constructs to assist the grid artificer module.In some embodiments, the adaptive grid management system 140 includes a system-wide federated alerts correlation processing engine with unique assigned ownership and colorization that utilizes grid-wide-mind distributed contextual sensing in support of grid-wide mindle. In some embodiments, the adaptive grid management system 140 includes a grid-wide federation command module that functions as a role-based command coordinator and executor, and operates as a grid-wide federation manager, grid foresight manager, and scouts command. The adaptive grid management system 140 according to some embodiments is described in further detail with reference to Figures 2-7 herein.
[0013] The field agent devices 170 include processor-based devices configured to communicate with the adaptive power grid management system 140 and other field agent devices 170 over a network, such as a protected private network, a virtual private network, a wireless network, a power lines communication (PLC) network, and / or the Internet. In some embodiments, the field agent devices 170 are geographically distributed and each is associated with a component or subsystem of the power grid, such as a windfarm, a digital substation, a solar farm, a power bank, an electric vehicle (EV) charging station, a 21339 controller, etc. In some embodiments, the field agent devices 170 may have a data connection with a control cell of an associated component of the network and may be configured to exchange data, send requests, and / or change the configuration of the network component via the control cell. In some embodiments, the field agent device 170 may be a user interface device (e.g., portable computer, cell phone) configured to provide instructions to a human operator via a user interface (UI) based on an executed scouting application. In some embodiments, the field agent device 170 is configured to execute one or more scouting applications received from the adaptive power grid management system 140 or from another field agent device to perform functions specified by the scouting application. In some embodiments, the field agent device 170 may be configured to operate for extended periods of time (e.g., hours, days) based on the instructions of the scouting application without communicating with the adaptive power grid management system.In some embodiments, a group of field agent devices 170 may form a federation based on roles assigned to the scout application executing on the devices. In some embodiments, a field agent device 170 executing a scout application that acts as a scout master for the federation may be configured to coordinate task execution and / or issue commands to other field agent devices 170 in the federation. In some embodiments, a field agent device 170 may be configured to install and run a copy of the executed scout application on another field agent device 170. In some embodiments, a field agent device 170 may be configured to automatically terminate the scout application after completion of one or more tasks or after a set period of time.
[0014] In some embodiments, the field agent devices 170 running one or more scouting applications may be referred to as operating cells of the adaptive power grid management system 140. In some embodiments, the adaptive power grid management system 140 is configured to provide adaptive formation of the power grid into geographically dispersed fit-for-purpose self-forming operating cells in response to dynamic power grid internal and external demands. In some embodiments, each operating cell or federation of cells is self-managed and can operate autonomously to achieve assigned objectives. In some embodiments, operating cells can further merge with other operating cells to create larger cells or join a federation to collectively achieve multiple assigned objectives. In some embodiments, an operating cell that is under stress (e.g., physical threat, cyber threat, insufficient processing power, memory, power supply, network connection stability, etc.) can devolve authority to other cells. In some embodiments, the adaptive grid management system 140 can be configured to provide dynamic formation of lateral access for adaptive distributed control to support shelter-in-place operating conditions and other emergency response operations. In some embodiments, the adaptive grid management system 140 can define federated transaction processing to enable transfer of authority and autonomous operation for multiple operating cells in response to commands or stress conditions.In some embodiments, the adaptive grid management system 140 provides operations and lifecycle management of scouting applications that support self-acting grid adaptive formation, grid formation, grid foresight, and optimization. In some embodiments, the system further includes a grid-wide-eye module to provide grid-wide observability and foresight with an extended line of sight and condition-based point of view to assist with grid formation.
[0015] In some embodiments, the adaptive grid management system 140 and the field agent devices 170 add one or more of the components shown in FIG. 1 to an existing power grid. Traditionally, components of the power grid operations data center 110 may communicate with the service and control cells of the power grid based on multiple messaging formats and / or protocols defined separately, such that components of the power grid operations data center 110 must each be configured to communicate with each type of control cell that the power grid operations center supports. In some embodiments, the adaptive grid management system 140 and the field agent devices 170 may act as an intermediary to unify communications between components of the power grid operations data center 110 and various controllers in the system. In some embodiments, each field agent may be configured to communicate with an associated controller, but communications between the field agent devices 170 and the adaptive grid management system 140 and / or the power grid operations data center 110 may be based on a unified messaging format and / or protocol.
[0016] 2, an adaptive power grid management system according to some embodiments is shown. In some embodiments, the adaptive power grid management system 200 may include one or more processor-based devices that execute computer-readable instructions stored in computer-readable memory storage devices. In some embodiments, the adaptive power grid management system may be implemented in a distributed and / or redundant server architecture on multiple networked processor-based devices. In some embodiments, the components of the illustrated adaptive power grid management system 200 may include software and / or hardware modules.
[0017] In some embodiments, the adaptive power grid management system 200 provides a self-behaving power grid that contextually adapts to new operating regimes in response to changing power system operating conditions. Operating conditions may include physical stress, environmental hazards, civil unrest, sabotage, pandemics, and shelter-in-place operating conditions. In some embodiments, the adaptive power grid management system 200 includes an alert correlation engine 201 based on uniquely assigned ownership and colorization utilizing distributed contextual sensing 272. In some embodiments, the adaptive power grid management system 200 includes a decision support module 202 that utilizes machine learning algorithms to provide dynamic, condition-based context data and contextual constructs to assist the power grid artificer module 210. In some embodiments, the adaptive power grid management system 200 includes a self-forming power grid artificer module 210 configured to drive self-acting power grid formation and operation to maintain physical and operational stability of the power grid in response to system context sensing operating conditions. In some embodiments, the adaptive power grid management system 200 includes a cognitive state machine ("CSM") operator 213 for operating a self-forming multi-dimensional, contextual, and cognitive state machine to drive the self-acting power grid formation and operation.In some embodiments, the adaptive grid management system 200 includes a grid federation command module 230 that functions as a role-based command coordinator and executor, operating as a grid-wide federation manager, grid foresight manager, and scouts command.
[0018] In some embodiments, the adaptive grid management system 200 is configured to provide adaptive formation of the grid into geographically dispersed fit-for-purpose self-forming operating cells 270 in response to dynamic grid internal and external imperatives. In some embodiments, each operating cell 270 is self-managed and can operate autonomously to achieve assigned objectives. Operating cells 270 can further merge with other operating cells to create larger cells or join federations to achieve multiple assigned objectives. In some embodiments, an operating cell 270 under stress can devolve authority to other cells. In some embodiments, the system is configured to provide dynamic formation of fit-for-purpose distributed control lateral access to support shelter-in-place and other emergency response activities. In some embodiments, the system can define federated transaction processing that allows for transfer of authority and autonomous operation for multiple operating cells in response to commands or stress conditions.
[0019] The alert correlation engine 201 is configured to acquire data from other data sources, such as grid operations data center 260 and contextual sensing 272, for alert condition detection. The alert correlation engine 201 may include an alert coordination manager module for coordinating geographically distributed alert acquisition processors. In some embodiments, the alert coordination manager module is configured to coordinate the processing and translation of alerts across related systems with unique system-wide assigned alert ownership. In some embodiments, the alert coordination manager further manages the ingestion and storage of related alerts to provide a dynamic, holistic view of the grid and its conditions. The stored information can be used for historical replay, pattern recognition, and situation coordination. In some embodiments, the stored information is used to train alert and / or context models using machine learning algorithms, such as deep neural network algorithms. The trained models may be configured by the alert correlation engine 201 to detect alerts related to the current condition of the grid and / or determine context information.
[0020] The alert correlation engine 201 receives signals from the alert coordination manager and / or the power grid operations data center 260, analyzes the signals, and detects abnormal conditions outside of an acceptable band. If an abnormal condition is detected, the alert correlation engine 201 can generate an action stress frame that includes a unique ID and signal attributes that triggers the alert correlation process. In some embodiments, the alert correlation engine 201 initiates a pattern recognition analysis based on the unique ID of the stress frame and generates a colored pattern plate based on the corresponding source, severity degree, and / or potential area of impact of the best match. In some embodiments, upon receiving the analysis result of the pattern plate, the alert correlation engine triggers an action stress frame constructor module to construct an action stress frame that includes a stress frame ID, pattern plates, and an associated time. In some embodiments, the alert correlation engine further creates a record in the learning engine 203 based on lessons learned from the analysis. The alert correlation engine 201 can then send the action stress frame to the grid-wide-mind module's decision support module 202 via a grid gatekeeper module. In some embodiments, the grid gatekeeper module is generally configured to provide authentication and security validation for communications between the grid-wide-mind module and the power grid artificer module.
[0021] The decision support module 202 is configured to validate the action stress frames via the grid-shaping analysis engine. In some embodiments, the grid-shaping analysis engine is configured to analyze the action stress frames through interaction with the learning engine 203 and send the analysis results and the context data to the context data builder. In some embodiments, the context data builder performs a correlation analysis on the results received from the grid-shaping analysis engine to determine associated vital attributes having value ranges. The context data builder further determines the context data ID and associated severity level (e.g., (for-id, (a1, sd1), (a2, sd2), . . . (an, sdn), ti). The decision support module 202 then sends "for-id, (attributes), t(x)" as context data to the grid artificer module 210 via the grid gatekeeper module.
[0022] The grid management database 205 stores temporal data, experience data, data ownership, alert ownership, grid profile, resource view data, application portfolio data, DNA mapping information, scouts portfolio data, and launch plans for use by the alert correlation engine 201, the decision support module 202, the learning engine 203, and / or the simulation engine 204.
[0023] The power grid artificer module 210 comprises a context configuration engine 211, a context abstraction plane database 212, a CSM operator 213, a CSM builder 214, and a CaCSM (Contextual and Cognitive State Machine) 215. In some embodiments, the power grid artificer module 210 can validate the context data identifier and request received from the decision assistance module 202. In some embodiments, the power grid artificer module 210 validates the "for-id, (attributes)" by extracting historical patterns of the "for-id, (attributes)", creates an activation key ("ga-authentication key"), and initiates power grid formation upon validation.
[0024] In some embodiments, when a misalignment is detected by the artificer module 210, the artificer module 210 may send the context data ID, the context data, and the time (tx1) to the decision assistance module 202 and request an accuracy analysis. In response, the decision assistance module 202 may check the difference in the system state between t(x) and t(x1) with the alert correlation engine 201 and perform an accuracy analysis. If the confidence degree from the accuracy analysis is within an acceptable range, the decision assistance module 202 sends the adjusted (for-id, (attributes), t(x), gwm-verification-key, ga-authentication-key) to the artificer module 210 via the gatekeeper module. If a change in system conditions causes the confidence to fall outside of the acceptable range, the decision support module 202 sends new context data to the grid artificer module 210, including (for-id, (attributes), t(x2), gwm-verification-key, ga-authentication-key-x). In some embodiments, if the grid artificer module 210 identifies an authentication mismatch, the verification step can be restarted. In some embodiments, the artificer module 210 may send the authentication key ("ga-authentication key") to the collaboration command module 230, the collaboration manager, the foresight manager, the scout command, and / or the auto operator. In some embodiments, the artificer module 210 then sends the context data (e.g., for-id, (a1, sd1), (a2, sd2), ... (an, sdn), t(y)) to the context construction engine 211.
[0025] The context construction engine 211 is configured to use the context data from the artificer module 210 to initiate context construction. The context construction engine 211 can send a request to the learning engine 203 to pinpoint the pattern and context associated with the received context data (e.g., for-id). In some embodiments, the context construction engine 211 can initiate context construction by mapping for each attribute (a1, n and sd1, n) the best matching context pattern with (for-id, (a1, sd1), (a2, sd2), ... , (an, sdn), t(y)). On the best matching context pattern, the context construction engine 211 constructs context abstraction panels (e.g., sub-context data), each panel representing a perspective and relative dimensions of a separate problem domain. In some embodiments, each panel is assigned a "cap-id" and may include panel positioning, operation classification, functional composition, and a severity index. The context construction engine 211 can then notify the artificer module 210 of the completion of construction and send the context abstraction plane data (cap-id, abstraction panels attribute, ga-authentication key) to the artificer module 210.
[0026] In some embodiments, in response to receiving a signal (cap-id, abstraction panels attribute, ga-authentication key) from the context construction engine 211, the power grid artificer module 210 validates the ga-authentication key and the corresponding "for-id, (attribute)" and activates the CSM entry creation. In some embodiments, the power grid artificer module 210 uses data from the decision support module 202 to map the abstraction panels and associated contexts and determine the level of context interdependence between the abstraction panels and associated indexes. In some embodiments, the power grid artificer module 210 then determines the most relevant abstraction panel context in relation to the "for-id, (attribute)" and marks the abstraction panel context based on the degree of relevance (e.g., high (most likely root cause), medium, low). Next, the power grid artificer module 210 creates a new composite ID (for-id, primary-context-id, attributes, ga-authentication key) and activates the cognitive state machine builder (CSM-builder) 214. In some embodiments, the CSM builder 214 extracts the context ID with the highest relevancy index (for-id, context-id, relevancy-index-high, attributes, ga-authentication key) from the signal received from the power grid artificer module 210 and sends the exact information to the learning engine 203.The learning engine 203 then extracts the context ID and the most closely matching pattern CSM (number of states, initial states, trigger, self-forming trigger, final state with corresponding performance index)-(for-id, primary-context-id, attributes, ga-authentication key) and sends the data to the CSM builder. In some embodiments, the CSM utilizes the template from the learning engine 203 to build a draft CSM entry. The creation of a new CSM draft entry by the CSM-Builder causes the power grid design module 210 to launch the simulation engine 204 for validation of the CSM. If the simulation engine 204 returns satisfactory results, the power grid artificer module 210 sends a signal with the certified simulation results to the CSM builder 214, allowing the CSM builder 214 to move the CSM entry into production. The CSM builder 214 then updates the draft CSM entry with the simulation results, creates a final blueprint, and sends the final blueprint with the associated for-ID to the learning engine 203. The CSM builder 214 further sends the final blueprint, the authenticated key, and a start command to the CSM operator 213. In some embodiments, the CSM operator 213 is configured to authenticate the signal from the CSM builder and, upon verification, verify that the blueprint is mapped to a CSM entry, activate self-forming capabilities in the associated cell or cells, and send a signal to the federation.
[0027] The grid-wide eye module 220 is configured to provide observability and predictability with an extended line of sight and condition-based point of view to aid in grid formation. In some embodiments, the grid-wide eye module 220 is configured to provide information to an alert engine including an alert correlation engine 201, an action stress frame constructor module, and a decision assistance module 202. In some embodiments, the adaptive grid management system 200 includes system-wide distributed context sensing 272 with embedded adaptive self-describing semantics and embedded methods for self-adjusting positioning. In some embodiments, the adaptive grid management system 200 includes an alert acquisition processor with built-in semantics that is deployed throughout the grid and configured for self-adjusting positioning.
[0028] The collaborative command module 230 is a role-based coordinator and functions as a grid formation manager, a grid foresight manager, and a scout coordinator. In some embodiments, the foresight manager creates an operational plan and optimizes a formation strategy based on signals from the artificer module 210. The foresight manager may further interact with the decision support module 202 to simulate the formation plan via the simulation engine 204 and derive operational capability attributes. In some embodiments, based on the simulation results, the foresight manager may optimize the formation plan, create a manifest template, define an operational cell DNA type, and record the manifest in a CSM cell with a corresponding DNA match. In some embodiments, if an operational cell does not exist, the collaborative command module 230 may request the CSM operator 213 to create a new operational cell. In some embodiments, once the CSM status cell formation is complete, the CSM operator 213 can set the cell as completion of incubation for the scout command module to initiate scout inception, training, and deployment preparation. Once incubation via the scout incubator 250 is complete, the CSM operator 213 updates the cell with a deployment ready flag and assigns a DNA.
[0029] In some embodiments, in response to a request from the grid foresight manager, the grid coordination command module 230 can initiate the definition and coordination of multiple distributed operating cells 270. In some embodiments, the coordination command module 230 may designate each cell as a virtual, physical, or hybrid base operation plan. In some embodiments, the cells are assigned an operational DNA that defines the cell's operational characteristics. In some embodiments, the cells may have knowledge of other cells, their locality, and DNA type. In some embodiments, the cells may operate autonomously for long periods of time. In some embodiments, the cells may be configured to decide to merge with other cells to create a larger cell with greater operational capabilities based on the detection of relevant trigger conditions. In some embodiments, a federation of multiple cells can be formed based on commands from a command hub, and the cells may be configured to create a new federation or join an existing federation of cells. In some embodiments, each federation of cells can include a coordination command hub that inherits relevant intelligence from the Grid Wide Mind module of the adaptive grid management system 200. In some embodiments, a federation may be managed by a Scout Master cell assigned by a command hub coordinator, and cells in the federation may be configured to receive commands managed by the Scout Master.
[0030] The Scout Command Module includes a Scout Incubator 250 and a Scout Launch Manager 251 that act on commands received from the Artificer Module 210. In some embodiments, the Scout application may be assigned the role of a Scout Messenger to provide optimized ingestion, transformation, and storage of relevant data on the move. In some embodiments, the Messenger Scout application may be automatically configured by the system to adapt to the region, people, assets, and applications. In some embodiments, the Scout application may be assigned the role of a Scout Coordinator that is automatically configured by the adaptive power grid management system 200 to provide coordination and management of the operating cells during assigned missions according to the mission plan. In some embodiments, the Scout application may be assigned the role of a Scout Inspector that is automatically configured by the system to observe the operation of the cells in relation to interactions, asset / device performance, people interactions, historical replay, pattern recognition, and machine learning, and provide the collected information to the decision support system. In some embodiments, the scout application may be assigned the role of a security guard that is automatically configured by the system to provide cell defense against physical threats, cyber threats, and functional inconsistencies. In some embodiments, the scout application running on an operating cell may be configured to clone itself to another operating cell in response to a trigger condition. In some embodiments, the scout application may be configured to change its configuration and / or tasks (i.e., self-form) after deployment in response to a trigger condition. In some embodiments, the scout application may be configured to self-terminate (i.e., demise) in response to a trigger condition or after a set period of time.
[0031] In some embodiments, the adaptive power grid management system 200 can provide operations and life cycle management of scout applications that aid in adaptive shaping of the power grid. In some embodiments, the scout command module can be configured to retrieve assignments and instructions from the CSM and map the received DNA description to a scout roster. In some embodiments, upon receiving instructions from the CSM, the scout command module can retrieve the associated assignment from the CSM operator 213. Based on the assigned DNA description, the scout command module can map the DNA to a scout in the scout roster, and if the matching scout is in an idle state, the scout command module can update the DNA, upload a mission plan, and assign a responsibility tag and alter ego level to the matched scout. If the matching scout is not in an idle state, the scout command module can initiate a cloning process to clone the scout, update the DNA, upload a mission plan, designate the responsibility tag, and designate an alter ego level for the cloned scout application. If no scout matches the received DNA, the system starts a new scout incubation and training process and notifies the CSM operator 213 to begin cell creation. The new scout application may be deployed once incubation and training is complete. In some embodiments, the scout command module may also manage assigned advance controllability and assign objective functions in relation to local grid morphing in the cell via the federated edge transaction manager 240.
[0032] The block diagram of Figure 2 is provided as an example configuration for an adaptive power grid management system. In some embodiments, one or more modules may be combined in function with another module or may be divided into multiple modules. In some embodiments, additional modules not shown in Figure 2 may be included to perform functions described herein.
[0033] In some embodiments, the systems and methods described herein are configured to increase system reliability, stability, and resilience to changing power system conditions, both when operating under normal and abnormal conditions. In some embodiments, the system provides dynamic power grid observability and situational awareness with an extended line of sight and state-based perspective, enabling faster response times to dynamic system conditions. In some embodiments, the system improves predictability of abnormal conditions in the power grid based on historical reconstruction, pattern recognition, and machine learning for operational optimization. In some embodiments, the system provides power grid assistance health management to monitor, analyze, and predict failures. In some embodiments, the system provides decentralized control applications and distributed computing to avoid vulnerabilities of centralized command and associated latencies. In some embodiments, the system enables a morphing grid with flexible integration and interoperability of diverse field devices, controllers, and communication paths at the edge. In some embodiments, the system further provides ecosystem value creation by enabling seamless interaction and collaboration between applications, assets, and people across different system domains. In some embodiments, the system may include E2E secure connectivity such that information flow between disparate entities across multiple network domains is protected. In some embodiments, the system provides full lifecycle device management requiring minimal manual intervention.In some embodiments, the system provides optimized capture, transformation, and storage of relevant data to meet power grid functions and future projected needs.
[0034] In some embodiments, the systems and methods described herein utilize a grid-wide artificial intelligence (AI) based decision support system to provide complex mission-critical machines that can transform into new formations in response to system stress, abnormal conditions, environmental hazards, pandemics, and shelter in place operating conditions with minimal manual intervention.
[0035] Referring now to Figure 3, an adaptive grid management system according to some embodiments is shown. In some embodiments, the adaptive grid management system 300 may include one or more processor-based devices that execute computer-readable instructions stored in memory storage devices. In some embodiments, the adaptive grid management system 300 may be implemented in a distributed and / or redundant server architecture on multiple networked processor-based devices. In some embodiments, the modules of the adaptive grid management system 300 may include software and / or hardware modules.
[0036] The adaptive grid management system 300 includes a grid-wide-mind module 310, a grid artificer module 315, a grid-wide command hub 340, and a federated grid data fabric 320. The grid-wide-mind module 310 includes a federated alert correlation engine and a decision support module. In some embodiments, the federated alert correlation engine is configured to detect alert conditions based on data from various sources including a grid operations data center 381, an enterprise data center, and cloud services 382. In some embodiments, the decision support module is configured to generate context data based on the alerts detected by the alert correlation engine. In some embodiments, the context data may be referred to as frame of reference ("for") data and may include alert type of severity and vital attributes. In some embodiments, the context data may be determined based on ML (machine learning) based pattern recognition and matching using context models trained based on historical data. In some embodiments, the GridWideMind module 310 may include the alert correlation engine 201, the decision support module 202, the learning engine 203, and the simulation engine 204 described with reference to FIG.
[0037] In some embodiments, the grid artificer module 315 may be configured to examine the context data from the grid-wide mind module 310 and determine a formation plan via a context construction engine and a context and cognitive state machine. In some embodiments, the context construction engine is configured to extract patterns and contexts related to the received context data and separate the context construction into abstraction panels that represent separate problems. The grid artificer module 315 then matches each abstraction plane with past components to determine a target grid formation and a formation plan to achieve the target formation through a cognitive state machine. In some embodiments, the grid artificer module 315 may include the grid artificer module 210, the context construction engine 211, the CSM operator 213, and the CSM builder 214 described with reference to FIG. 2.
[0038] Grid-wide command hub 340 includes a grid-wide-eye module 342 and a grid-wide federation command module 344. Grid-wide eye module 342 is configured to aggregate data from grid operations data center 381, enterprise data centers and cloud services 382, and operational cells and organize the data into views for various components of adaptive grid management system 300. In some embodiments, grid-wide eye module 342 may comprise grid-wide eye module 220 described with reference to FIG. 2.
[0039] The grid-wide federation command module 344 may include a federation manager module, a foresight manager module, and a scout command module. The foresight manager creates operational plans and optimizes formation strategies based on signals from the grid-wide mind module. The grid federation manager initiates the definition and coordination of multiple distributed operating cells. The scout command is configured to incubate and launch scout applications for execution in the operating cells 391, 392 and cells in the federation 390. The federated grid data fabric 320 may include a grid management system database that provides data for use by the grid-wide mind module 310 and the grid-wide federation command module 344. In some embodiments, the grid-wide federation command module 344 may be configured with the federation command module 230 described with reference to FIG. 2.
[0040] Each of the operational cells 391-395 includes a scout application that runs on a device in the power grid. In some embodiments, the devices may be physical field agent devices or virtual field agent devices that run as software components on other devices. In some embodiments, the operational cells may form a scout federation 390 that shares communications and commands. In some embodiments, one of the operational cells in the federation 390 may be designated as a scout master of the scout federation 390 and may be configured to coordinate task execution of the cells in the federation 390 to achieve a common goal. In some embodiments, each of the operational cells may be assigned a role (e.g., coordinator, messenger, inspector, guard) by the adaptive power grid management system 300 depending on the characteristics of the executed scout application. The operational cells may then communicate with the other operational cells and the adaptive power grid management system 300 based on the assigned role. In some embodiments, each operational cell 391 may be associated with a device, a sensor, or a person. In some embodiments, an operational cell may be configured to provide instructions, collect data, relay messages, affect configuration, and provide security protection according to assigned tasks in the scout application.
[0041] 4, an adaptive grid management system is shown according to some embodiments. In some embodiments, the adaptive grid management system 430 may include one or more processor-based devices that execute computer-readable instructions stored in memory storage devices. In some embodiments, the adaptive grid management system 430 may be implemented in a distributed and / or redundant server architecture on multiple networked processor-based devices. In some embodiments, the modules of the adaptive grid management system 430 may include software and / or hardware modules.
[0042] The adaptive grid management system 430 is coupled to a grid operations data center 410, cloud services 421, an enterprise data center 420, a grid system operator 425, and a market capabilities system 426. The grid operations data center 410 includes a transmission and distribution management system, a DER and renewable management system, a grid visibility and insight system, a historian system, a grid device and asset services system, and a security services system. The cloud services 421 includes cloud applications and service offerings. The enterprise data center includes enterprise applications and services, DR / DER / EV aggregators, virtual power plant systems, and other external data and alerts. The power grid system operators 425 consist of transmission system operators and distribution system operators. The market function systems 426 can include energy market systems and peer-to-peer electricity trading systems.In some embodiments, one or more components of grid operations datacenter 410, cloud services 421, enterprise datacenter 420, grid system operator 425, and market capability system 426 include a grid artificer virtual agent ("gAVA") that enables the components to exchange data and commands with adaptive grid management system 430. In some embodiments, gAVA can standardize communication messaging formats of protocols between components.
[0043] The adaptive grid management system 430 includes a grid-wide mind module 431, a grid artificer module 432, and a grid-wide command hub. The grid-wide mind module 431 includes a decision support engine, a coordinated alert engine, data storage and services. The grid-wide mind module 431 is configured to detect alert conditions in the grid, determine contextual data based on ML pattern recognition and matching, and simulate a formulation plan for execution based on current grid conditions. In some embodiments, the grid-wide mind module 431 includes the decision support module 202, the learning engine 203, the simulation engine 204, and the alert correlation engine 201 described with reference to FIG. 2. The grid artificer module 432 includes a contextual and cognitive state machine (CSM) for executing the formulation plan determined based on the contextual information received from the grid-wide mind module. In some embodiments, the grid artificer module includes the grid artificer module 210 described with reference to FIG. 2.
[0044] Grid-wide command hub 439 includes a grid-wide-eye module 435, a grid-wide federation command 436, a grid-wide formation manager 434, a grid-wide foresight manager 433, a scouts command 438, and a federated edge transaction manager 437. Grid-wide eye module 435 is configured to aggregate data from grid operations datacenter 410, enterprise datacenter 420, cloud services 421, grid system operator 425, and market capability system 426 and organize the data into views for use by various components of adaptive grid management system 430.
[0045] In some embodiments, the grid-wide coordination command 436 is configured to initiate the definition and coordination of multiple distributed operating cells. The grid-wide formation manager 434 is configured to define and coordinate the formation of the power grid through the operating cells. The foresight manager 433 is configured to create operational plans and optimize formation strategies based on signals from the grid-wide mind module 431. The scout command 438 is configured to incubate and launch scout applications for execution in the operating cells. The coordination edge transaction manager 437 is configured to assign advance controllability and objective functions in association with local grid morphing in the cells.
[0046] The operation cells of the adaptive power grid management system 430 may be configured with a distributed operation cell 440 that executes a scout application. The distributed operation cell 440 may include a processor-based device having control circuitry that executes computer-readable instructions stored in a memory storage device. In some embodiments, the distributed operation cell 440 is configured to communicate with one or more field agent devices based on the instructions of the scout application executing on the field agent device and a role assignment associated with the field agent device. The distributed operation cell 440 may execute software-defined control, contextual reasoning, virtual sensing, alert collection, and security services. In some embodiments, the security services may be configured to provide cybersecurity to the distributed operation cell 440 and verify and validate the scout application prior to execution. In some embodiments, the distributed operation cell further comprises a data fabric that stores data for decision making in the distributed operation cell 440. In some embodiments, the distributed operation cell 440 may be configured to coordinate the installation and execution of the scout application on multiple field agent devices. In some embodiments, the distributed operating cells 440 may be configured to perform their functions and make decisions based on scouting applications and stored data without communicating with the adaptive power grid management system 430 for extended periods of time (e.g., hours, days).
[0047] An operating cell of the adaptive power grid management system 430 may include an adaptive field agent device 450 that executes a scout application. In some embodiments, the field agent device 450 may include a physical and / or virtual device. The field agent device 450 may include a processor-based device having control circuitry that executes computer-readable instructions stored on a memory storage device. The field agent device 450 may be assigned a role based on the executed scout application and perform tasks based on the instructions of the scout application. The field agent device 450 may perform analytics, data services, semantic modeling, virtual sensing, and dynamic adaptation. In some embodiments, security services may be configured to verify and validate the scout application prior to execution. In some embodiments, the field agent device 450 may be associated with and / or communicate with components of the power grid, such as plants 461, controllers 462, sensors 463, devices 464, assets 465, and electric vehicles 466. In some embodiments, the field agent devices 450 may interface with power grid components to aggregate data, send commands, provide security services, or perform inspections. In general, the field agent devices running the scout application are configured to perform one or more tasks based on the directives of the scout application as part of a formation plan for the adaptive power grid management system 430. In general, the adaptive power grid management system 430 may have any number of distributed operation cells 440 and adaptive field agent devices 450 running the scout application. The distributed operation cells 440 may also be configured to command and coordinate any number of the adaptive field agent devices 450.
[0048] Although the descriptions of Figures 1-4 generally refer to power grid device networks, the teachings of the present disclosure are also applicable to other types of device networks, such as communications networks, transportation networks, Internet of Things (IoT) networks, enterprise networks, and the like.
[0049]
[0023] Referring now to Figure 5, a process for device network management is shown. In some embodiments, one or more steps of Figure 5 may be performed by a processor-based device or system, such as an adaptive power grid management system described herein.
[0050] Prior to step 501, the system receives device signals from a source other than the device running the scout application at step 512. In some embodiments, the signals received at step 512 may include device and system state information provided by another system monitoring the device or network of devices. In some embodiments, the data received at step 512 may consist of other contextual data such as weather information, customer feedback information, incident reports, news services, web content, etc. At step 501, the system aggregates signals from devices in the network. In some embodiments, the device signals include status and alert data from various components of the network including information such as connectivity, throughput, response time, etc. In some embodiments, the signals may be captured and processed through another system such as an operations center or enterprise data center prior to step 501. At step 502, an alert condition is detected. In some embodiments, the system may be configured to detect an alert condition based on detecting a signal outside of the expected band. In some embodiments, the alert condition may be detected based on an alert signal from one or more of the network devices or the operations center. In some embodiments, in the case of a power grid, the alert signal may correspond to an outage or loss of connection to a component, or a condition that is likely to lead to an outage or loss of connection. In step 503, the system performs an alert correlation analysis by comparing the detected device signal to historical device network data.In some embodiments, the system may first train a context model using past device signals and past context information associated with the network of devices. Current signals from the network of devices are then used as inputs to the context model to determine context data associated with the current state of the network of devices. In some embodiments, the alert condition may correspond to a current problem or a predicted future problem (e.g., outage, overload, component failure).
[0051] In step 504, context data is determined based on the alert correlation analysis. In some embodiments, the context data may include vital attributes and severities associated with each vital attribute. In some embodiments, the context data may further include a control area identifier, a timing identifier, and an alert condition identifier determined based on the signal from the device. For example, the context data may identify a geographic area associated with the device that generated the signal that led to the detection of the alarm condition. In step 505, the context data is separated into sub-context data, each associated with a distinct problem in the device network. In step 506, each sub-context data is compared to historical context data patterns from a historical context pattern database to determine a formation plan for affecting power grid formation in response to the alert in step 507. In some embodiments, each historical context pattern may identify a particular problem associated with the sub-context and a target grid formation or one or more tasks for affecting network formation in response to the problem. In some embodiments, the formation plan may include tasks to be performed by multiple operational cells in the device network. In some embodiments, the system may determine a state machine including a list of steps and trigger conditions for each step determined based on context information for incrementally reconfiguring the device network from a current state to a desired state, and the formation plan may be determined based on states defined in the state machine.In some embodiments, the tasks may include one or more of a coordination task, a messenger task, an inspection task, or a guard task. In step 508, the system simulates the formation plan based on the updated signals from the network devices to verify that the execution of the formation plan achieves the desired power grid formation. If the formation plan passes the simulation, the system configures the scout federation and scout applications for incubation and deployment based on the formation plan and / or state machine. In step 509, the scout applications assigned to each operating cell may be configured based on the tasks and roles of the operating cell assigned by the formation plan, as well as the characteristics of the operating cell, such as device type, device capability, etc. In some embodiments, if the directives of the formation plan match a previously configured scout application, the previously configured scout application may be searched for deployment. In some embodiments, if an existing scout application does not match the directives of the formation plan, the system may configure a new scout application based on the saved application components (e.g., DNA). The newly created scout application may be incubated and tested prior to deployment. In some embodiments, the system may further determine command and / or reconfiguration of existing operating cells already running the scout application based on the creation plan.
[0052] At step 510, the scout application is deployed for execution in operating cells, such as distributed control hubs and field agent devices. In some embodiments, the scout application may be deployed or triggered in stages based on the state of the state machine. At step 511, after the scout application is deployed, the system monitors data from the operating cells executing the scout application. In some embodiments, the system may further issue commands and instructions to the scout application as needed based on updated data from the network and / or state machine of the devices. After the scout application is deployed, the system continues to aggregate device signals to update the state machine and determine a new formation plan in response to remaining or new alert conditions. In some embodiments, the new formation plan may trigger an update or termination of the deployed scout application. In some embodiments, the scout application may be configured to cause the operating cell devices to perform tasks for extended periods of time in response to various trigger conditions without further commands or instructions from the network management system. Further details of the operating cell devices executing the scout application are described herein with reference to FIG. 6. In some embodiments, new formation plans may be directed to different sets of operating cells and device network components to address different alert conditions. In some embodiments, after successful execution of a formation plan that removes an alert condition, the detected device signals and formation plans may be used to train machine learning algorithm models for future context building and / or formation plan determination.
[0053]
[0023] Referring now to Figure 6, a process for device network management by an operational cell is shown. In some embodiments, one or more steps of Figure 6 may be performed by a processor-based device or system within a network of devices, such as the field agent devices and distributed control hubs described herein.
[0054] In step 601, a device in the power grid receives a scouting application from a network management system or another device in the network, such as a field coordinator device or a field agent device. In step 602, the device verifies the authenticity and timeliness of the received scouting application. In some embodiments, the authentication may be based on an authentication key. In step 603, the device executes the scouting application and begins executing instructions of a formation plan based on tasks defined in the scouting application. In some embodiments, the device can assume a role (e.g., coordinator, messenger, inspector, guard) based on the role assigned in the scouting application. In step 604, the device detects trigger conditions associated with various tasks and actions. In step 605, the device can begin communicating with other field agent devices, a device federation, and / or the central device management system based on the role assigned in the scouting application. For example, a device may begin accepting tasks assigned to it by a master of the federation (begin to a field agent device sending data to a central device management system). In some embodiments, the scout application can specify trigger conditions associated with actions and tasks. For example, a data transmission task can be triggered by a timestamp or by the detection of an anomalous condition in the data.In step 606, the device executes a task, which may be a coordinator task, a messenger task, an inspector task, or a guard task. In step 607, if a cloning trigger is detected, the device may send a copy of the scout application to be executed by another device. In some embodiments, the device may further transfer authority and role to another device in step 607. For example, if the device detects a stress condition (e.g., unstable power supply or data connection), the device may transfer its tasks and roles to another device in the network that can complete the commands of the scout application. In step 608, if a termination trigger is detected, the device may terminate execution of the scout application. In some embodiments, the termination trigger may include the completion of one or more tasks or the detection of a change in a specified condition. In some embodiments, the device may delete the scout application from its memory in step 608.
[0055] 7, a network management system according to some embodiments is shown comprising a network management system 710 having access to a context model 740, a device database 750, a context pattern database, a scout application database 770, and a network operation system 720. The network management system 710 is further configured to communicate with a plurality of operating cell devices 780.
[0056] The network management system 710 includes a control circuit 712, a memory 717, and a network interface device 713. The control circuit 712 may be comprised of one or more of a processor, a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), an application-specific integrated circuit (ASIC), etc., and may be configured to execute computer-readable instructions stored in a computer-readable storage memory 717. The computer-readable storage memory 717 may include a volatile memory and / or a non-volatile memory and stores thereon computer-readable instructions that, when executed by the control circuit 712, cause the control circuit 712 to detect an alert state in a network of devices, determine context data based on an ML learned model, determine a formation plan based on the context data, and cause a plurality of scouting applications to perform tasks in an operational cell of the network of devices to affect a state of the network. In some embodiments, the network management system 710 may be implemented using multiple memory devices and / or processors at a central location or geographically distributed across multiple locations. In some embodiments, the computer-executable instructions cause the control circuitry 712 of the network management system 710 to execute one or more modules described herein with reference to Figures 2-4 and / or to perform one or more steps described herein with reference to Figure 5.
[0057] The network interface device 713 may include a data port, a wired or wireless network adapter, etc. In some embodiments, the network interface device 713 may communicate with the operating cell devices 780 and the network operation system 720 over a network, such as a local network, a private network, or the Internet. In some embodiments, the network management system 710 may further access one or more of the context model 740, the device database 750, the context pattern database, and the scout application database 770 via the network interface device 713.
[0058] The operating cell device 780 may include a processor-based device in the network of devices with control circuitry and memory. The operating cell is generally configured to receive and execute one or more scout applications configured by the network management system 710 to perform tasks specified by the network management system 710. In some embodiments, the tasks cause the operating cell device 780 to collect data, send requests / instructions, or change the configuration of another component or subsystem of the network of devices. In some embodiments, the tasks cause the operating cell to report collected data or other status information to the network management system 710. In some embodiments, the tasks cause the operating cell to form federations with and communicate with other operating cells based on roles associated with the executed scout application. In some embodiments, the control circuitry executing the scout application stored on a computer-readable memory device may cause the operating cell device 780 to perform one or more steps described herein with reference to FIG. 6. The network management system 710 is configured to communicate, directly or through another network device, with multiple devices in the network that may function as operating cell devices for the network management system 710.
[0059] The network operation system 720 may constitute an operation system of the network of devices that aggregates data from various sources and coordinates commands and requests with the network of devices during normal operation. For example, in a power grid, the network operation system 720 may include the power grid operations data center 110 and the enterprise data center 120 described with reference to FIG. 1. In some embodiments, the network operation system 720 may be an existing control system of the device network with which the network management system 710 communicates, providing additional functionality of automatic reconfiguration of the network of devices in response to alert conditions through tasks performed in the operational cells. In some embodiments, the network management system 710 may further communicate with other cloud services or network data sources, such as third-party data services, for alert detection and context determination.
[0060] The context model 740 includes a pattern recognition model trained based on an ML learning algorithm, such as a neural network algorithm. In some embodiments, the context model is trained using historical signals from devices in the network as input and historical context data as classifications. In some embodiments, the historical context data may be entered manually, derived from device signals from earlier or later time points, or sourced from another data source, such as outage reports, news services, etc. The network management system 710 can use the context model to determine context data related to the current state of the network of devices.
[0061] The device database 750 includes parameter and / or status information associated with devices in the network. In some embodiments, the device database 750 may store information such as a unique identifier, device type identifier, geographic location, IP address, memory capacity, processing capacity, bandwidth capacity, operating system, etc. associated with devices in the device network, including devices that may be used as operating cell devices of the network management system 710. In some embodiments, the network management system 710 may use the information stored in the device database 750 to select devices for performing specific tasks and scout applications. In some embodiments, the network management system 710 may further configure the scout application based on the information associated with the device such that the scout application is suitable for execution on the selected device.
[0062] The context pattern database 760 is comprised of context patterns, each associated with a different problem domain. In some embodiments, each context pattern may be associated with one or more vital attributes and associated severity degrees that may be matched to context data associated with a current network condition. In some embodiments, the context pattern may be further associated with a geographic region. In some embodiments, the context pattern may identify a target network condition or response associated with the problem domain. In some embodiments, the network management system 710 may use the context patterns in the context pattern database 760 to determine formation plans and / or state machines for affecting the state of the network of devices in response to the problem domain and / or alert condition associated with the context pattern.
[0063] The scout application database 770 includes scout applications and / or application components (e.g., application DNA). In some embodiments, the scout application may include code for performing one or more tasks based on one or more trigger conditions. In some embodiments, the scout application may be configured based on the roles assigned to the application and / or operating cell, and the configuration and constraints of the operating cell. In some embodiments, the scout application database 770 may store previously configured applications, and the network management system 710 may retrieve existing applications for deployment in the operating cell based on a formation plan. In some embodiments, the network management system 710 may configure new applications for deployment using application components in the scout application database 770. In some embodiments, newly formed scout applications may be first tested in a simulation or incubation environment prior to deployment. In some embodiments, successfully simulated or executed scout applications may be stored in the scout application database 770 for future use.
[0064] In some embodiments, one or more of the context model 740, device database 750, context pattern database 760, and scout application database 770 may be implemented on the same memory storage device or devices. In some embodiments, one or more of the context model 740, device database 750, context pattern database 760, and scout application database 770 may be implemented on memory 717 of the network management system 710 or may be accessed over a network by the network management system 710. In some embodiments, one or more of the context model 740, device database 750, context pattern database 760, scout application database 770, network operation system 720 may be accessed and updated by multiple network management systems 710, e.g., a distributed network management system serving different geographic regions.
[0065]
[0046] Referring now to Figure 8, a flow diagram of an artificer module is shown according to some embodiments. In some embodiments, one or more steps of Figure 8 may be performed by a processor-based device or system, such as an adaptive grid management system described herein.
[0066] In step 801, the artificer module receives context data associated with an alert condition from a decision support module. The context data may be determined by the decision support module based on signals received from a plurality of devices on the network of devices. In step 802, the system forms a context meta object. In some embodiments, the context data includes a plurality of vital attributes, each associated with a severity and a timestamp. In some embodiments, the context meta object is determined based on a machine learning model trained based on historical context data associated with devices on the network of devices. In some embodiments, the context meta object is determined based on a stress assessment module that determines possible underlaying causes based on a deterministic algorithm. In some embodiments, the context meta object includes a plurality of likely contexts ranked based on probabilities determined by the machine learning model.
[0067] In step 803, the system determines a number of context abstraction panels based on parsing the context meta object into a number of context groups and assigning an abstraction problem domain to each context group based on a machine learning algorithm. In some embodiments, the context abstraction panels comprise panel positioning, operation classification, functional composition, and / or a severity index.
[0068] In step 805, the system determines interdependencies of one or more abstraction panels based on a machine learning model trained on the context history archive database, the context meta object further including composite abstraction panel data with associated relevance indexes. In step 806, the system forms a provisional cognitive state machine, and in step 807, validates the provisional cognitive state machine with a simulation engine. Based on the results of the validation, in step 808, one or more states of the provisional cognitive state machine can be calibrated. In step 809, the system forms a cognitive state machine based on the context meta object for execution by a cognitive state machine operator to affect operation of one or more devices on the network via a network adapter, the cognitive state machine including a plurality of states each defining at least one task for execution by at least one network device. In some embodiments, the cognitive state machine defines multiple states, an initial state, trigger conditions for each state, a final state with a corresponding performance index, and recalibration conditions for one or more states.
[0069] 9A, 9B, 10A, 10B, and 11, flow diagrams are shown according to some embodiments. In some embodiments, one or more steps of FIG. 9A, 9B, 10A, 10B, and 11 may be performed by a processor-based device or system, such as an adaptive power grid management system described herein. The variables in FIG. 9A, 9B, 10A, and 10B may be defined as follows:
[0070] for-ID: reference frame-ID Van: Vital Attributes (Va1~n) sdn:Severity(sd1,n) ti: Recorded time of the event ga: Power Grid Artificer ga-authentication key: Power grid artificer authentication key ca: Context Abstraction ca-id1, n: Context abstraction ID (1 to n) apd: abstraction problem domain apd-idn.:Abstract problem domain ID (1, n) Cn: Context 1, n pd: panel domain Hn: History pattern scn,n:Subcontext(1,n) for-idh(1,n): for-id(1,n) is a series of past for-ids with similar contexts / variations related to for-id ppn: Panel position relative to other panels OC: Panel-related operational classification fcn: function composition padn: Panel Abstraction Domain Ri1, n.: Relevance index Ccap: Composite contextual abstraction panel sn:state(1,n) an.:action(1,n) tn:transition Sf(1,0): Self-formation (1=on, 0=off) - state with self-formation CaCSMbp: Context-Aware State Machine-Based Production CaCSMp: Context-Aware State Machine - Provisional
[0071] 9A and 9B, a pre-formation workflow is illustrated. A grid-coordinated alert engine 951 detects an alert condition based on data from a grid operations data center and a grid-wide mind sensing system. An action stress frame constructor 952 constructs an action stress frame (including an action stress frame ID, a pattern plate, and a time (t)) based on network device data related to the detected alert. The action stress frame is passed to a decision support (ds) gatekeeper 953 for validation. If the action stress frame passes validation, the action stress frame is passed to a ds grid formation analysis and machine learning module 954. The ds grid formation analysis and machine learning module 954 sends a pattern recognition request to a learning engine 955 that is trained based on past action stress frame data. The learning engine 955 is configured to output a pattern mapping to the ds grid formation analysis engine and machine learning module 954 in response to the pattern recognition request. The ds grid formation analysis engine and machine learning module 954 then provides the analysis data and the context data to a ds context model builder 956. The ds context model builder 956 applies correlation analysis and context inference to the data received from the ds grid formation analysis engine and machine learning module 954 to determine context or reference frame information including a frame of reference ID (for-ID), vital attributes (van), severity degrees (sdn), and time (tn). The reference frame data is passed to the ds gatekeeper 953 for validation. If the reference frame information passes validation, the reference frame data (e.g., for-id, (va1, sd1), (va2, sd2), .... (van, sdn), ti) is passed to the grid artificer module 960.In some embodiments, the power grid artificer module 960 may constitute one or more of the power grid artificer modules 210, 315, and 432 described herein.
[0072] In step 961, the power grid artificer module 960 creates a power grid artificer (GA) authentication key. In step 962, the power grid artificer module 960 validates the frame of reference data against historical patterns. In step 963, the power grid artificer module 960 calculates a misalignment factor for the context. If the misalignment is within an acceptable range (e.g., a predefined threshold), the output of the power grid artificer module 960 is passed to the GA control and command module 970 and the context construction engine 980. If a deviation beyond the threshold is detected, an accuracy analysis request is generated and sent to the DS gatekeeper 953. The accuracy analysis request includes for-id, (attributes), tx1, ga authentication key, and misalignment factor. The DS gatekeeper 953 then sends a system state variance analysis request from the initial time to the grid-connected alert engine 951, which in turn sends an accuracy analysis request including the system state variance analysis result from t(x) to t(x1) to the grid-wide decision support module 966. In step 967, the system then calculates a confidence for the recommendation. If the confidence is not within an acceptable range (e.g., below a predetermined threshold), the process returns to the DS gatekeeper 953 for validation. If the confidence is within an acceptable range (e.g., above a predetermined threshold), the output of the grid artificer module 960 is passed to the GA control and command module 970 and the context configuration engine 980.
[0073] The GA control and command module 970 sends a formation alert (for-id, authentication key) to the output of the grid artificer module 960 (e.g., (for-id, (a1, sd1), (a2, sd2), ..., (an, sdn), t(y)). In some embodiments, the grid-wide federation command module 975 may be comprised of one or more of the federation command modules 230, 344, and 436. After authenticating the information, the federation command module 975 sends a notification to a grid wide federation manager, a grid wide foresight manager, a grid scouts command, and a grid edge federated transaction manager 977.
[0074] The output of the power grid artificer module 960 (e.g., (for-id, (a1, sd1), (a2, sd2), .an, sdn), t(y)) is also provided to the context configuration engine 980. In some embodiments, the context configuration engine 980 may configure the context configuration engine 211. In step 901, the context configuration engine 980 sends the received reference data (e.g., (for-id, (a1, sd1), (a2, sd2), .an, sdn), t(y)) to the learning engine 983. In step 902, the frame of reference data ((for-id, (a, sd1), (a2, sd2), .an, sdn), t(y)) is sent to a context inference module 985 that utilizes machine learning. In step 903, the for-id, (an, sdn) is sent to a stress force evaluation module 984, which determines the underlying cause of (an, sdn) based on a deterministic algorithm in step 904. In step 905, (an, sdn) is sent to a context history archive 987, which provides (anhn, sdn) in step 906. The stress force evaluation module 984 then provides (for-id, (anhn, sdn) + (dc1...dcn)) to a context inference module 985 in step 907. The context historical archive 987 also provides pattern extraction to the context inference module 985 in step 907a. In step 908, the context inference module determines the most probable context and rank.In step 909, the context inference module 985 outputs the context to a context meta object builder 986, which sends the context meta object ((for-id, ((c1, pd1, (sc1.1, ... sc1.n,)), (c2, pd2, (sc2.1, ... sc1.n,)), ..., ((cn, pdn, (sc2.1, ... sc1.n)))), t(y))+(for-id, (a1, sd1), (a2, sd2), ..., (an, sdn), t(y)))+((for-id, (for-idh1, n))) to the context construct engine 980 in step 911. The meta object (id) is also stored in the context history archive 987 for model training in step 910.
[0075] In step 912, the output of the power grid artificer module 960 is sent to the power grid artificer parser 981 which in step 913 parses the context meta objects from the one or more learning engines, determines coloration and grouping, creates context abstraction IDs (ca-id1, n,) for each context group, and assigns an abstraction problem domain ID (apd-idn). In step 914, the parsed data ((for-id, (ca-id1(c1, ...cn), ad-idn), .... (ca-idn(c1, ...cn), ad-idn), t(y)) is passed to a context abstraction panel builder 982, which creates individual abstraction panels with unique IDs for the corresponding correlated contexts and related problem domains, ranks the abstraction panels based on the probability provided by the learning engine, and determines the positioning of the panels relative to each other in step 915. In step 916, the context abstraction panel builder 982 passes the context abstraction panel data (e.g., cap-id, (panel positioning, operation classification, function configuration, severity index) to the context configuration engine 980. In step 917, the cap-id, abstraction panel attributes, and ga authentication key are sent from the context construction engine 980 to the power grid artificer module 960 for CaCSM creation.
[0076] 10A and 10B, the CaCSM creation stage of the creation workflow is shown. In step 917, the context construction engine 980 initiates CaCSM creation by sending for-id, (cap-id1, n, (panel position (ppn,), operation classification (ocn), functional configuration (fcn), severity (sdn), padn), ga-authentication key) to the power grid artificer module 960. In step 917a, the ga gatekeeper 1003 performs the authentication (ga-authentication key). In step 917b, if the authentication is rejected, a rejection with the cap-id, ga-authentication key is sent back to the context construction engine 980. If the authentication is accepted, in step 917c, the ga gatekeeper notifies the power grid artificer module 960.
[0077] In step 918, the power grid artificer module 960 establishes abstraction panel (AP) interdependencies based on the AP interdependency analyzer 1004 that maps the abstraction panels and associated contexts, and determines the level of context interdependence between the abstraction panels and associated indexes. The cap-ids 1, n, ((ppn,), (ocn), (fcn), (sdn), padn), and meta objects (id) output by the AP interdependency analyzer 1004 are sent to the ds correlation engine 1005 that verifies the relevance of the correlation. The information is then sent to a learning engine 965 that is trained on data from the context history archive 987 and outputs a highly dependent context abstraction panel (cap) (cap-id1,n, relevance index (ri1,n)), meta object (id)) to a composite cap builder 969. The composite cap builder 969 outputs the composite cap (Ccap-id) and relevance index (rlh,m,l) to a context meta object builder 986. The composite cap builder 969 builds a meta object (e.g., ((for-id, ((c1, pd1, (sc1.1, ... sc1.n,))), (c2, pd2, (sc2.1, ... sc1.n,)), ... , ((cn, (sc2.1, ... sc1.n)), pdn)), t(y)) + (for-id, (a1, sd1), (a2, sd2), ... , (an, sdn), t(y))) + ((for-id, (for-idh1,n ))) + composite cap (Ccap-id, relevance index (rih,m,l)) and sends the ((for-id, (Ccap-id, relevance index (rih,m,l)) meta object (id) to the grid metric module 960 in step 919.
[0078] In step 920, (for-id, (Ccap-id, relevance index (rih, m,)tx)) is passed to the CSM builder 1010, which extracts the cap-id with the highest relevance index, and forwards the for-id, (context ID, ri(high), attributes) to the learning engine 1011. In some embodiments, the CSM builder 1010 may include a CSM builder 214. The learning engine 1011 determines the best matching CaCSM pattern from the state machine history archive 1013, and sends the CaCSM template and the for-id, (context ID, attributes) to a provisional CaCSM builder 1012, which determines the parameters of the state machine, such as the number of states, initial states, triggers, functional compositions, self-forming triggers, and final states with corresponding performance indexes, based on the context ID and attributes stored in the state machine history archive 1013. The provisional CaCSM template is then sent to the context state machine meta object construction tool 1014 and the meta object ID is stored in the state machine history archive 1013 for use in training the learning engine 1011 .
[0079] In step 921, the context state machine meta object builder 1014 sends for-id, ((context id, attributes), CaCSMp-id, ((s1, a1, t1, sf(1,0)), .... (sn, an, tn, sf(1,0)), meta object(id) to the CSM builder 1010. In step 922, the CSM builder 1010 sends a CaCSM validate request with for-id and ty to the power grid artificer module 960, which in step 923 sends Validate(CaCSMp-id,), meta object(id) to the simulation engine 1020. In some embodiments, the simulation engine 1020 can include the simulation engine 204. The simulation scenarios are stored in a simulation scenarios historical archive. In step 923a, the simulation engine forwards Validate(CaCSMp-id,), meta object(id) as a query to the SM history archive 1013. In step 924, for-id, ((context id, attributes), CaCSMp-id, ((s1, a1, t1, sf(1, 0)), ... (sn, an, tn, sf(1, 0)), meta object(id) is retrieved from the state machine history archive 1013. In step 925, a CaCSM state calibration request including (sn, an, tn, sf(1, 0)) is retrieved from the state machine history archive 1013. request) is sent to the learning engine 1021, and in step 925a, the calibrated states (sn, an, tn, sf(1, 0)) are sent back to the simulation engine 204, which runs the mock CaCSM in simulation mode and calibrates each state as necessary. In step 926, the simulation engine 1020 validates the states (sn, an, tn, sf(1, 0)) and forwards the validated states to the production CaCSM builder 1022.The production CaCSM builder 1022 is configured to build, for each CaCSMbp-id, the number of states, initial states, trigger, functional composition, self-forming trigger, and final state with corresponding performance index, in step 927. In step 928, the CaCSMbp-id, (attributes), are stored in the state machine history archive 1013. In step 929a, the calibrated for-id, ((context id, attributes), CaCSM p-id, ((s1, a1, t1, sf(1, 0)), ... (sn, an, tn, sf(1, 0)), meta object (id) are provided to the simulation engine 1020. In some embodiments, the simulation and calibration steps may be repeated on the temporary CaCSM. In step 929, the simulation engine 1020 sends an operation-ready notification to the grid artificer module 960 with the (CaCSM p-id, ) and meta object (id). In step 930, the grid artificer module 960 signals the initiation of formation and operation to the CSM builder 1010. The initiation of formation and operation, including for-id, ((CaCSM bp-id), meta object (id), ga authentication key), is referred to as the operation phase of the formation workflow. The process is sent to the CSM operator 1040 to start the process (stage).
[0080] Next, referring to FIG. 11, the operation stage of the formation workflow is shown. In step 931, the CSM builder 1010 sends the for-id, ((CaCSMbp-id), meta object (id), ga-authentication key) to the CSM operator 1040. In step 932, the CSM operator 1040 sends an authentication request (ga-authentication key) to the ga gatekeeper 1003 for verification. The verified for-id is forwarded to the power grid artificer module 960 in step 932a, and the rejected for-id is sent back to the CSM builder 1010. In step 932b, the power grid artificer module 960 activates the CaCSMbp-id and sends the data to the CSM operator 1040. In step 933, the power grid artificer also sends the for-id, meta object (id), and context (id) to the command logistics builder 1023 which in step 933a creates a preparation action list and sends the command logistics data to the grid-wide coordination command module 975.
[0081] The CaCSM 1030 receives r(1, (sf=1, 0, (sf=0)) from the CSM operator and outputs the corresponding (t, r(1, 0), f) in each state to the grid-wide coordination manager, the preview manager, the scout command, and / or the edge coordination task manager. The CaCSM 1030 is further configured to send a CaCSM state calibration request to the learning engine 1045 to calibrate the state (sn, an, tn, sf(1, 0). The CaCSM 1030 also provides the state operation results to the CSM operator 1040.
[0082] The grid-wide coordination command module 975 is configured to communicate with the CSM operator 1040 and execute state machine command execution within the operation loop. The grid-wide coordination command module 975 communicates with the grid-wide coordination manager, the foresight manager, the scout command, and the edge coordination task manager based on the logistics preparation notification (Context(id), Logistics(id)). The grid-wide coordination command module 975 is further configured to communicate with the power grid artificer module 960 to respond to status requests with status reports when the state machine is executed.
[0083] In some embodiments, an adaptive power grid management system is provided. The system includes a network device database, a network adapter, and a processor. The processor is configured to train a context model with a machine learning algorithm using historical device signals and historical context information related to the network of devices, receive signals from the network of devices, determine context data related to a current state of the network of devices based on the context model, and determine a formation plan based on the context data, the formation plan including tasks to be performed by one or more devices in the device network, configure one or more scouting applications based on the formation plan and device information stored in the network device database, and cause the one or more scouting applications to be executed by the one or more devices in the device network via the network adapter.
[0084] In some embodiments, a decision support system for network management is provided. The system comprises a network device database and a processor coupled to the network device database. The processor is configured to execute the decision support module to detect an alert condition based on signals from a network of devices, determine context data associated with the alert condition, provide the context data to an artificer module, receive an artificer plan from the artificer module, perform a simulation of the artificer plan and current network conditions, and configure a plurality of scout applications based on the simulation, the plurality of scout applications for execution on a plurality of devices on the network. In some embodiments, the alert condition is detected based on determining whether the signals from the network of devices correspond to an abnormal condition that exceeds a tolerance band. In some embodiments, detecting the alert condition includes training an alert model based on historical signals and the context data from the network of devices with a learning engine. In some embodiments, detecting the alert condition includes aggregating the signals from the network into stress frames and performing pattern recognition analysis on the stress frames based on the historical stress frame data. In some embodiments, the processor is further configured to determine a set of selected vital attributes each associated with a severity associated with the alert condition, the context data including the set of vital attributes and the associated severity.
[0085] In some embodiments, an artificer system for device network management is provided. The system includes a context configuration database and a processor coupled to the context configuration database. The processor is configured to execute the artificer module and receive context data associated with an alert condition from a decision support module, the context data including a plurality of vital attributes and determined based on signals received from a plurality of devices on the network, and for each vital attribute of the context data, match a context pattern in the context pattern database to determine a context abstraction panel and determine a formation plan for execution by the decision support module. In some embodiments, the system is further configured to determine a misalignment between the context data and subsequent context data and request new context data from the decision support module when a misalignment occurs. In some embodiments, the context abstraction panel comprises one or more of a panel positioning, an operation classification, a functional composition, and a severity index.
[0086] In some embodiments, a decision support system for device network management is provided. In some embodiments, the system comprises a scout application database and a processor coupled to the scout database. The processor is configured to execute the decision support module, receive an artificer module, derive operational and functional attributes based on a simulation of the artificer plan, determine scout application configurations for a plurality of scout applications based on the operational and functional attributes using a cognitive state machine, and configure one or more scout applications for execution on one or more devices of the network based on the attributes. In some embodiments, the scout application configurations are determined based on selecting at least one stored scout application configuration from the scout database based on the artificer plan. In some embodiments, the processor is further configured to perform incubation and training of the newly configured scout application prior to transmitting the newly configured scout application to devices on the network.
[0087] In some embodiments, an operating cell device for power grid management is provided. The device includes a network adapter configured to communicate with a device network management system and a processor coupled to the network adapter, the processor configured to receive a scout application having an associated role assignment from a server via the network adapter, authenticate the scout application, automatically execute the scout application upon authentication, and communicate with a plurality of devices in the device network management system or network via the network adapter based on the role assignment of the scout application. In some embodiments, the scout application is assigned a role of messenger and configured for ingestion, transformation, and storage of data associated with one or more of locality, persons, assets, and applications associated with the cell device. In some embodiments, the scout application is assigned a role of coordinator configured to provide coordination and management of one or more other cell devices in a federation. In some embodiments, the scout application is assigned a role of inspector and configured to collect data associated with operation of the cell device or a federation including the cell device. In some embodiments, the Scout application is assigned a guard role and is configured to prevent a physical threat, a cyber threat, and functional misalignment. In some embodiments, the Scout application running on a cellular device is configured to install and run another instance of the Scout application on another cellular device.In some embodiments, the scout application is configured to automatically terminate in response to an exit condition being met. In some embodiments, the scout application is configured to perform its assigned role without communicating with a server or another cell device.
[0088] In some embodiments, a coordinated command system for device network management is provided. The system includes a network device database and a processor. The processor is configured to execute the coordinated command module and form a federation including a plurality of cell devices, each executing at least one of a plurality of scout applications associated with a role in the federation, based on the formation plan and the device database, cause the plurality of scout applications to execute on the plurality of cell devices, and send a command to one of the plurality of cell devices based on a role associated with the scout application executed by the cell device. In some embodiments, the processor is further configured to simulate the formation plan with the decision support module before configuring the plurality of scout applications.
[0089] Further aspects of the present disclosure are provided by the subject matter of the following clauses. [Embodiment 1] 1. An adaptive power grid management system comprising: a network device database; a network adapter; and a processor coupled to the network device database and the network adapter, the processor configured to: learn a context model with a machine learning algorithm using historical device signals and historical context information related to the network of devices; receive signals from the network of devices; and determine context data related to a current state of the network of devices based on the context model; determine a formation plan based on the context data, the formation plan including tasks to be performed by one or more devices in the device network; configure one or more scouting applications based on the formation plan and device information stored in the network device database; and cause the one or more scouting applications to be executed by the one or more devices in the device network via the network adapter. [Embodiment 2] 13. The adaptive power grid management system of any preceding embodiment, wherein the one or more devices include a field agent device associated with one or more of a power plant, a solar power plant, a wind power plant, a digital substation, a microgrid controller, and an electric vehicle charging station. [Embodiment 3] 13. An adaptive power grid management system as in any preceding embodiment, wherein the context data includes a set of selected significant attributes, each associated with a severity level. [Embodiment 4] 13. The adaptive power grid management system of any preceding embodiment, wherein the context data includes a control area identifier, a timing identifier, and an alert condition identifier. [Embodiment 5] 11. The adaptive power grid management system of any of the preceding embodiments, wherein the formation plan is determined based on matching the context pattern with context patterns in a context database, separating the context data into context abstraction panels, each corresponding to a separate problem domain perspective, and determining a task and one or more devices associated with each context abstraction panel. [Embodiment 6] 13. The adaptive power grid management system of any preceding embodiment, wherein the formation plan is determined based on a state machine including a list of steps and trigger conditions determined based on the context data. [Embodiment 7] 13. An adaptive power grid management system as in any preceding embodiment, wherein configuring the one or more scout applications includes selecting a stored scout application that matches the configuration plan. [Embodiment 8] 13. An adaptive power grid management system as in any preceding embodiment, wherein the one or more scout applications are each assigned a role and configured based on that role. [Embodiment 9] 13. The adaptive power grid management system of any preceding embodiment, wherein configuring the one or more scout applications includes generating a new scout application using stored application components based on the configuration plan; and testing the new scout application prior to deployment. [Embodiment 10] 13. An adaptive power grid management system as in any preceding embodiment, wherein the processor is further configured to simulate the formation plan based on updated signals from the network of devices before causing the one or more scouting applications to be executed by one or more devices in the network of devices. [Embodiment 11] 1. A method for adaptively managing a power grid, comprising: training a context model using a plurality of historical device signals and historical context information associated with a network of a plurality of devices by a machine learning algorithm executed on a processor; receiving a signal from said network of a plurality of devices; determining, by the processor, context data relating to a current state of the network of a plurality of devices based on the context model; determining, by the processor, an operational plan based on the context data, the operational plan including tasks to be performed by one or more devices in the network of devices; determining, by the processor, a configuration plan based on context data, the configuration plan including tasks to be performed by one or more devices in the network of devices; configuring, by the processor, one or more scout applications based on the formation plan and device information stored in the network device database; causing the one or more scout applications to be executed by one or more devices in the network of devices. [Embodiment 12] 13. The method of adaptively managing a power grid of any preceding embodiment, wherein the one or more devices include a field agent device associated with one or more of a power plant, a solar power plant, a wind power plant, a digital substation, a microgrid controller, and an electric vehicle charging station. [Embodiment 13] 13. The method of adaptive management of an electric power grid of any preceding embodiment, wherein the context data includes a set of selected significant attributes, each associated with a severity degree. [Embodiment 14] 13. The method of adaptive management of an electric power grid of any preceding embodiment, wherein the context data includes a control area identifier, a timing identifier, and an alert condition identifier. [Embodiment 15] 13. The method of adaptively managing a power grid of any preceding embodiment, wherein the formation plan is determined based on matching the context data with context patterns in a context pattern database, separating the context data into context abstraction panels, each corresponding to a separate problem domain perspective, and determining a task and one or more devices associated with each context abstraction panel. [Embodiment 16] 13. The method of any preceding embodiment, wherein the formation plan is determined based on a state machine including a list of steps and trigger conditions determined based on the context data. [Embodiment 17] 13. The method of adaptive management of an electric power grid of any preceding embodiment, wherein configuring the one or more scout applications comprises selecting a stored scout application that matches the configuration plan. [Embodiment 18] 13. The method of adaptive management of a power grid of any preceding embodiment, wherein the one or more scout applications are each assigned a role and configured based on the role. [Embodiment 19] 13. The method of adaptive management of an electric power grid of any preceding embodiment, wherein configuring the one or more scout applications includes generating a new scout application using stored application components based on the configuration plan; and testing the new scout application prior to deployment. [Embodiment 20] 13. The method of adaptively managing a power grid of any preceding embodiment, further comprising simulating the formation plan based on updated signals from the network of multiple devices prior to causing one or more scouting applications to be executed by one or more devices in the network of multiple devices. [Embodiment 21] 1. An adaptive power grid management system comprising: a context history archive database; a network adapter; and a processor coupled to the context history archive database and the network adapter, the processor comprising: receiving contextual data associated with the alert condition from a decision support module, the contextual data being determined based on signals received from a plurality of devices on a network of devices; forming a context meta object by matching the context data with historical context data stored in a context history archive database; forming a cognitive state machine based on the context meta object for execution by a cognitive state machine operator to affect operation of one or more devices on a network via said network adapter; configured to execute an artificer module that executes The adaptive power grid management system, wherein the cognitive state machine includes a plurality of states, each state defining at least one task to be performed by at least one network device. [Embodiment 22] 13. The adaptive power grid management system of any preceding embodiment, wherein the context data includes a plurality of vital attributes, each associated with a severity, and a timestamp. [Embodiment 23] 13. The adaptive power grid management system of any preceding embodiment, wherein the context meta object is determined based on a machine learning model trained based on historical context data associated with devices on the network of devices. [Embodiment 24] 13. The adaptive power grid management system of any preceding embodiment, wherein the context meta object is determined based on a stress force assessment module that determines likely root causes based on a deterministic algorithm. [Embodiment 25] 13. The adaptive power grid management system of any preceding embodiment, wherein the context meta object includes a plurality of likely contexts ranked based on a probability determined by a machine learning model. [Embodiment 26] 13. The adaptive power grid management system of any preceding embodiment, wherein the processor further forms the cognitive state machine based on determining a plurality of context abstraction panels by parsing the context meta object into a plurality of context groups and assigning an abstraction problem domain to each context group based on a machine learning algorithm. [Embodiment 27] 13. The adaptive power grid management system of any preceding embodiment, wherein each context abstraction panel includes one or more of a panel location, an operation classification, a functional configuration, and a severity index. [Embodiment 28] 11. The adaptive power grid management system of any preceding embodiment, wherein the context meta object is further determined based on determining interdependencies of one or more abstracted panels based on a machine learning model trained on a context history archive database, and the context meta object further includes composite abstracted panel data having an associated relevance index. [Embodiment 29] 13. The adaptive power grid management system of any preceding embodiment, wherein the cognitive state machine forms a provisional cognitive state machine, the provisional cognitive state machine being determined based on validating with a simulation engine. [Embodiment 30] 13. The adaptive power grid management system of any preceding embodiment, wherein the cognitive state machine defines a number of states, an initial state, a trigger condition for each state, a final state with a corresponding performance metric, and a recalibration condition for one or more states. [Embodiment 31] 1. A method for adaptive power grid management, comprising: receiving, at a processor executing an artificer module, from a decision support module, contextual data relating to an alert condition, the contextual data being determined based on signals received from a plurality of devices on a network of a plurality of devices; forming, at the processor, a context meta object by matching the context data with historical context data stored in a context history archive database; forming, by a processor, a cognitive state machine based on the context meta object that is executed by a cognitive state machine operator to affect operation of one or more devices on the network via a network adapter; wherein the cognitive state machine includes a plurality of states that define at least one task to be performed by at least one network device. [Embodiment 32] 13. An adaptive power grid management method as in any preceding embodiment, wherein the context data includes a plurality of vital attributes, each associated with a severity, and a timestamp. [Embodiment 33] 13. The adaptive power grid management method of any preceding embodiment, wherein the context meta object is determined based on a machine learning model trained based on historical context data associated with devices on the network of devices. [Embodiment 34] 13. An adaptive power grid management method as in any preceding embodiment, wherein the context meta object is determined based on a stress force assessment module that determines likely root causes based on a deterministic algorithm. [Embodiment 35] 13. The method of any preceding embodiment, wherein the context meta object includes a plurality of likely contexts ranked based on probabilities determined by a machine learning model. [Embodiment 36] 13. The adaptive power grid management method of any preceding embodiment, wherein forming the cognitive state machine is further based on determining a plurality of context abstraction panels based on parsing the context meta object into a plurality of context groups and assigning an abstraction problem domain to each context group based on a machine learning algorithm. [Embodiment 37] 13. An adaptive power grid management method as in any preceding embodiment, wherein each context abstraction panel includes one or more of a panel location, an operation classification, a functional configuration, and a severity index. [Embodiment 38] 11. The adaptive power grid management method of any preceding embodiment, wherein the context meta object is further determined based on determining interdependencies of one or more abstracted panels based on a machine learning model trained on a context history archive database, and the context meta object further includes composite abstracted panel data having an associated relevance index. [Embodiment 39] 13. The method of adaptive power grid management as recited in any preceding embodiment, wherein the cognitive state machine forms a provisional cognitive state machine, the provisional cognitive state machine being determined based on validating with a simulation engine. [Embodiment 40] 13. An adaptive power grid management method as in any preceding embodiment, wherein the cognitive state machine defines a number of states, an initial state, trigger conditions for each state, and a final state with a corresponding performance metric.
[0090] Those skilled in the art will recognize that numerous other modifications, variations and combinations may be made to the above-described embodiments without departing from the scope of the present invention, and such modifications, variations and combinations are considered to be within the scope of the inventive concept. [Explanation of symbols]
[0091] 110: Power grid operation data center 120: Enterprise data center 130: Cloud service 140: Adaptive power grid management system 150: System operator 160: DR / DER / EV aggregator 170: Field agent device 180: Virtual power plant 200: Adaptive power grid management system 201: Alert correlation engine 202: Decision support module 203: Learning engine 204: Simulation engine 205: Power grid management database 210: Power grid artificer module 211: Context composition engine 212: Context abstraction plane database 213: Cognitive state machine operator / CSM operator 214: CSM builder 215: CaCSM 220: Grid wide eye module 230: Power grid cooperation command module 240: Cooperation edge transaction manager 250: Scout incubator 251: Scout launch manager 260: Power grid operation data center 270:Operation Cell 272:Distributed Context Sensing 300:Adaptive Grid Management System 310:Grid Wide Mind Module 315:Grid Artificer Module 320:Federated Grid Data Fabric 340:Grid Wide Command Hub 342:Grid Wide Eye Module 344:Grid Wide Coordination Command Module 381:Grid Operational Data Center 382:Enterprise Data Center and Cloud Services 390:Federation 391, 392:Operation Cell 410:Grid Operational Data Center 420:Enterprise Data Center 421:Cloud Services 425:Grid System Operator 426:Market Capability System 430:Adaptive Grid Management System 431:Grid Wide Mind Module 433:Grid Wide Foresight Manager 434:Grid Wide Formation Manager 435:Grid Wide Eye Module 436:Grid Wide Coordination Command 437:Coordination Edge Transaction Manager 438:Scout Command 439: Grid-wide command hub 440: Distributed operating cell 450: Adaptive field agent device 461: Plant 462: Controller 463: Sensor 464: Device465: Asset 466: Electric vehicle 710: Network management system 712: Control circuit 713: Network interface device 717: Memory 720: Network operation system 740: Context model 750: Device database 760: Context pattern database 770: Scout application database 780: Operating cell device 951: Grid cooperation alert engine 952: Action stress frame constructor 953: Decision support gatekeeper 954: ds grid formation analysis and machine learning module 955: Learning engine 956: ds context model builder 960: Grid artificer module 965: Learning engine 966: Grid-wide decision support 967: Context history archive 969: Composite CAP builder 970: ga control and command module 975: Cooperation command module 977: Grid-wide cooperation manager, Grid-wide foresight manager, Grid scout command and Grid-edge cooperation transaction manager 980: Context configuration engine 981: Power grid artificer parser 982: Context abstraction panel builder 983: Learning engine 984: Stress force assessment module 985: Context inference module 986: Context meta object builder 987: Context history archive 1004: AP interdependence analyzer 1005: DS correlation engine 1010: CSM builder 1011: Learning engine 1012: Provisional CaCSM builder 1013: State machine history archive 1014: Context state machine meta object construction 1020: Simulation engine 1021: Learning engine 1022: Production CaCSM builder 1023: Command logistics builder 1025: Simulation scenario history archive 1030: CaCSM 1040: CSM operator 1045: Learning engine
Claims
1. 1. An adaptive power grid management system, comprising: A network device database; A network adapter; A processor coupled to the network device database and to the network adapter, the processor comprising: training a context model with a machine learning algorithm using historical device signals and historical context information associated with the network of multiple devices; receiving a signal from said network of a plurality of devices; determining context data relating to a current state of the network of a plurality of devices based on the context model; determining a configuration plan based on context data, the configuration plan including tasks to be performed by one or more devices in the network of devices; configuring one or more scout applications based on the formation plan and device information stored in the network device database; running, via the network adapter, the one or more scout applications on one or more devices in the network of devices; 2. An adaptive power grid management system configured to:
2. The system of claim 1 , wherein the one or more devices comprise a field agent apparatus associated with one or more of a power plant, a solar power plant, a wind power plant, a digital substation, a microgrid controller, and an electric vehicle charging station.
3. The system of claim 1 , wherein the contextual data comprises a set of selected significant attributes, each associated with a severity level.
4. The system of claim 1 , wherein the context data includes a control domain identifier, a timing identifier, and an alert condition identifier.
5. The formation plan is Matching the context data with context patterns in a context pattern database; Separating the context data into context abstraction panels, each corresponding to a different problem domain perspective; The system of claim 1 , wherein the context abstraction panel is determined based on determining the task and the one or more devices associated with each context abstraction panel.
6. The system of claim 1 , wherein the formulation plan is determined based on a state machine including a list of steps and trigger conditions determined based on the context data.
7. The system of claim 1 , wherein configuring the one or more scout applications comprises selecting a saved scout application that matches the formation plan.
8. The system of claim 1 , wherein the one or more scout applications are each assigned a role and configured based on the role.
9. Configuring the one or more scout applications includes: generating a new scout application using the stored application components based on the formation plan; and testing the new scout application prior to deployment.
10. 2. The system of claim 1, wherein the processor is further configured to simulate the formation plan based on updated signals from the network of multiple devices before causing the one or more scout applications to be executed by the one or more devices in the network of multiple devices.
11. 1. A method for adaptively managing a power grid, comprising: training a context model using historical device signals and historical context information associated with a network of multiple devices with a machine learning algorithm executing on a processor; receiving a signal from said network of a plurality of devices; determining, at the processor, context data relating to a current state of the network of a plurality of devices based on the context model; determining, at the processor, an operational plan based on the context data, the operational plan including tasks to be performed by one or more devices in the network of devices; configuring, by the processor, one or more scout applications based on the configuration plan and device information stored in the network device database; causing one or more scout applications to be executed on one or more devices in the network of devices.
12. 12. The method of claim 11, wherein the one or more devices include a field agent device associated with one or more of a power plant, a solar power plant, a wind power plant, a digital substation, a microgrid controller, and an electric vehicle charging station.
13. The method of claim 11 , wherein the contextual data comprises a set of selected important attributes, each associated with a severity level.
14. The method of claim 11 , wherein the context data includes a control domain identifier, a timing identifier, and an alert condition identifier.
15. The formation plan is matching said context data with context patterns in a context pattern database; separating the context data into context abstraction panels, each corresponding to a different problem domain perspective; determining the task and the one or more devices associated with each context abstraction panel; The method of claim 11 , wherein the step is determined based on
16. The method of claim 11 , wherein the formulation plan is determined based on a state machine including a list of steps and trigger conditions determined based on the context data.
17. The method of claim 11 , wherein the step of configuring one or more scout applications includes selecting a stored scout application that matches the formation plan.
18. The method of claim 11 , wherein the one or more scout applications are each assigned a role and are configured based on the role.
19. configuring the one or more scout applications, generating a new Scout application using the stored application components based on the formation plan; Testing the new Scout application before deployment; The method of claim 11 , comprising:
20. 12. The method of claim 11, further comprising simulating a formation plan based on updated signals from the network of a plurality of devices prior to executing the one or more scout applications by one or more devices in the network of a plurality of devices.
Citation Information
Patent Citations
Multi-modal, multi-disciplinary feature discovery to detect cyber threats in electric power grid
US20180262525A1
