Multi-asset and multi-market energy management system

Abstract

A computer-implemented method (100) for optimizing energy management behind-the-meter in a multi-asset environment, the method comprising: receiving (110), by a data processing unit, real-time and historical data from a plurality of energy assets, the energy assets including at least one of batteries, electric vehicle (EV) charging stations, solar inverters, or controllable loads; determining (120) a power distribution schedule by processing the received data in accordance with a predefined multi-variable control model, wherein the model accounts for: time-dependent grid capacity constraints, forecasted energy demand and supply, and market-based pricing signals; transmitting (130) the computed power distribution schedule to a local execution unit; receiving and implementing (140) the power distribution schedule at the local execution unit by adjusting the operation of the energy assets; monitoring (150), in real-time, electrical parameters including voltage, current, and power, at the grid connection point and across the energy assets; overriding (160) the received power distribution schedule when local conditions indicate a risk of exceeding contractual grid capacity limits; transmitting (170) real-time operational data from the local execution unit to the data processing unit; wherein the method regulates energy consumption and generation behind-the-meter based on dynamically updated power distribution schedules, ensuring compliance with contractual grid capacity constraints, maximizing self-consumption of locally generated energy, and selectively adjusting asset operation in response to market-based energy pricing conditions.

Description

[0001] Title: Multi-asset and multi-market energy management system

[0002] Description:

[0003] Technical field

[0004] The present invention generally relates to the field of energy management systems, and more particularly to the field of computer-implemented optimization of behind-the-meter energy distribution in multi-asset environments, including energy storage, electric vehicle (EV) charging, and grid interaction.

[0005] Background

[0006] Energy management in distributed systems has become an increasingly complex challenge due to the rapid integration of renewable energy sources, electric vehicles (EVs), and flexible loads into the electrical grid. Traditionally, energy distribution has been managed through centralized control mechanisms that rely on static grid models and predefined operational parameters. These models are often designed to accommodate predictable and relatively stable energy consumption patterns. However, with the growing adoption of behind-the-meter energy assets such as battery storage, photovoltaic (PV) systems, and dynamic EV charging stations, conventional grid management strategies are becoming increasingly inefficient. The current state of the art includes systems that employ peak shaving techniques and demand response strategies to mitigate grid overloads and stabilize voltage and frequency fluctuations. These solutions often rely on aggregators and energy market participation to balance distributed loads across a network. However, such aggregation is often too simplistic to manage complex multi-asset scenarios, such as a single battery system being shared at a portfolio level across multiple sites, each with its own local consumption patterns and constraints.

[0007] Despite these advancements, existing methods suffer from several critical disadvantages. One primary limitation is that traditional peak shaving and demand response mechanisms operate on fixed schedules or simple threshold-based control schemes, which do not dynamically adapt to real-time grid conditions or marketfluctuations. For instance, upon receiving a real-time imbalance market signal, they cannot make a sophisticated decision on whether to strategically deviate from their schedule to capture value or to maintain perfect balance to avoid penalties This results in inefficiencies where valuable energy resources are either underutilized or excessively constrained, leading to economic losses and suboptimal grid stability. Moreover, current systems lack a comprehensive approach to integrating contractually defined grid capacity constraints, which can vary based on time-dependent agreements, grid congestion levels, or market signals. Without real-time adaptation to these constraints, distributed energy resources (DERs) often fail to optimize their charging and discharging cycles in a way that benefits both the grid and the asset owners. A typical example of this problem is an industrial user who has batteries and electric vehicle chargers on-site, and who participates in both a dynamic capacity contract (CBC) and the imbalance market. Traditional systems cannot properly manage and optimize all these different assets together, especially when the market rules and grid limits are complex and connected. This often causes extra costs and missed chances to earn money.

[0008] Another major disadvantage of the existing systems is the lack of a hierarchical control framework that efficiently coordinates multiple levels of decision-making. Current methods typically apply either centralized cloud-based optimization, which suffers from latency and limited responsiveness to local grid disturbances, or decentralized control at the asset level, which lacks holistic oversight of market trends and grid-wide constraints. This disconnect results in inefficient dispatch of energy storage and generation resources, leading to increased reliance on expensive grid reinforcement measures such as infrastructure upgrades and additional power generation capacity. Additionally, conventional approaches do not sufficiently integrate multi-variable optimization models that consider a combination of physical grid constraints, predictive market dynamics, and real-time power flow data, further limiting the effectiveness of existing solutions.

[0009] It is therefore a goal of the present invention to provide a method and system for optimizing behind-the-meter energy management in a multi-asset environment by dynamically adapting energy distribution schedules in response to real-time grid constraints, market conditions, and forecasted demand. This approach enables a more efficient allocation of distributed energy resources while ensuring compliance withcontractually defined grid capacity limits, thereby overcoming the above-mentioned disadvantages of the prior art at least in part.

[0010] Summary

[0011] One aspect of the present invention relates to a computer-implemented method for optimizing energy management behind-the-meter in a multi-asset environment, the method comprising:

[0012] receiving, by a data processing unit, real-time and historical data from a plurality of energy assets, the energy assets including at least one of batteries, electric vehicle (EV) charging stations, solar inverters, or controllable loads;

[0013] determining a power distribution schedule by processing the received data in accordance with a predefined multi-variable control model, wherein the model accounts for:

[0014] - time-dependent grid capacity constraints,

[0015] - forecasted energy demand and supply, and

[0016] - market-based pricing signals;

[0017] transmitting the computed power distribution schedule to a local execution unit; receiving and implementing the power distribution schedule at the local execution unit by adjusting the operation of the energy assets;

[0018] monitoring, in real-time, electrical parameters including voltage, current, and power, at the grid connection point and across the energy assets;

[0019] overriding the received power distribution schedule when local conditions indicate a risk of exceeding contractual grid capacity limits;

[0020] - transmitting real-time operational data from the local execution unit to the data processing unit;

[0021] wherein the method regulates energy consumption and generation behind-the-meter based on dynamically updated power distribution schedules, ensuring compliance with contractual grid capacity constraints, maximizing self-consumption of locally generated energy, and selectively adjusting asset operation in response to market-based energy pricing conditions.Another aspect of the present invention relates to a system for optimizing energy management behind-the-meter in a multi-asset environment, the system comprising:

[0022] a data processing unit, configured to:

[0023] receive real-time and historical data from a plurality of energy assets, the energy assets including at least one of batteries, electric vehicle (EV) charging stations, solar inverters, or controllable loads;

[0024] determine a power distribution schedule by processing the received data in accordance with a predefined multi-variable control model, wherein the model accounts for:

[0025] time-dependent grid capacity constraints,

[0026] forecasted energy demand and supply, and market-based pricing signals;

[0027] transmit the computed power distribution schedule to a local execution unit;

[0028] a local execution unit, configured to:

[0029] receive and implement the power distribution schedule by adjusting the operation of the energy assets;

[0030] continuously monitor electrical parameters, including voltage, current, and power, at the grid connection point and across the energy assets;

[0031] override the received power distribution schedule when local conditions indicate a risk of exceeding contractual grid capacity limits;

[0032] transmit real-time operational data to the data processing unit; wherein the system regulates energy consumption and generation behind-the-meter based on dynamically updated power distribution schedules, ensuring compliance with contractual grid capacity constraints, maximizing self-consumption of locally generated energy, and selectively adjusting asset operation in response to market-based energy pricing conditions.

[0033] Another aspect of the present invention relates to a computer program product comprising instructions which, when executed by a data processing unit of a system for optimizing energy management behind-the-meter in a multi-asset environment, cause the data processing unit to perform the method of an aspect of the present disclosure.Another aspect of the present invention relates to a non-transitory computer-readable storage medium storing a computer program comprising instructions which, when executed by a processor, cause the processor to perform the method of an aspect of the present disclosure.

[0034] Another aspect of the present invention relates to a computer-readable data carrier comprising computer-executable instructions which, when executed by a data processing unit, cause the system to execute the method of an aspect of the present disclosure, wherein the data carrier is at least one of:

[0035] a semiconductor memory,

[0036] a magnetic storage device,

[0037] an optical disc, or

[0038] a network-accessible data storage system.

[0039] One aspect of the present invention relates to a computer-implemented method for optimizing energy management behind-the-meter in a multi-asset environment. Energy management refers to the control, monitoring, and optimization of energy flows within a system comprising energy generation, storage, and consumption units. Behind-the-meter denotes that the energy assets involved operate within a local energy distribution environment, typically on the consumer side of an electricity meter, rather than as part of a centrally controlled grid infrastructure. A multi-asset environment includes various controllable energy resources such as energy storage systems, electric vehicle charging stations, solar inverters, and other flexible loads, each contributing to the optimization process in different ways.

[0040] The method comprises receiving, by a data processing unit, real-time and historical data from a plurality of energy assets, the energy assets including at least one of batteries, electric vehicle charging stations, solar inverters, or controllable loads. A data processing unit refers to a computational system capable of executing algorithms to analyze, predict, and optimize energy management processes. Realtime data includes instantaneous measurements of electrical parameters such as power flow, voltage, and current, while historical data provides insights into past consumption patterns, generation profiles, and market conditions. An effect of this arrangement is that decisions can be based on continuously updated information rather than relying on static, preconfigured schedules, thereby improving responsiveness to dynamic conditions.The method further comprises determining a power distribution schedule by processing the received data in accordance with a predefined multi-variable control model, wherein the model accounts for time-dependent grid capacity constraints, forecasted energy demand and supply, and market-based pricing signals. A power distribution schedule is a computationally determined plan that specifies how energy assets should be controlled over a given time period to optimize their operation. A multi-variable control model refers to a structured algorithm that simultaneously considers multiple influencing factors to derive an optimized control strategy. Timedependent grid capacity constraints define the permissible power exchange with the grid as a function of time, often dictated by contractual agreements, regulatory limitations, or grid congestion considerations. Forecasted energy demand and supply predict the future power requirements and generation potential based on historical trends, weather conditions, and consumption behavior. Market-based pricing signals include data from energy trading platforms that provide price forecasts for electricity markets, allowing for informed decisions regarding energy storage and dispatch. An effect of this arrangement is that energy asset control is optimized in a way that minimizes costs, maximizes efficiency, and ensures compliance with contractual grid constraints, while dynamically responding to fluctuating supply and demand conditions.

[0041] The method further comprises transmitting the computed power distribution schedule to a local execution unit. Transmission in this context refers to the electronic communication of computed schedules from the data processing unit to a control system responsible for implementing the schedule at the physical asset level. A local execution unit denotes a processing system that directly interfaces with the energy assets to enforce computed schedules by adjusting their operation accordingly. An effect of this arrangement is that the optimized control strategy derived at the central processing level can be effectively executed at the operational level, ensuring that distributed energy resources operate in accordance with the determined plan.

[0042] The method further comprises receiving and implementing the power distribution schedule at the local execution unit by adjusting the operation of the energy assets. Implementation refers to the process of executing the prescribed control actions to regulate energy storage, consumption, and generation. Adjusting the operation of energy assets includes actions such as modifying charging anddischarging cycles of batteries, increasing or decreasing power consumption of flexible loads, and curtailing or maximizing solar power feed-in based on real-time grid conditions. An effect of this arrangement is that energy resources can dynamically respond to computed optimization strategies, ensuring that the system operates within desired efficiency, cost, and regulatory parameters.

[0043] The method further comprises monitoring, in real-time, electrical parameters including voltage, current, and power, at the grid connection point and across the energy assets. Monitoring refers to the continuous acquisition and analysis of electrical measurements to assess the state of the energy system. Electrical parameters such as voltage, current, and power provide fundamental insights into the operational status of assets and the interaction between distributed energy resources and the grid. An effect of this arrangement is that deviations from expected operational conditions can be identified immediately, allowing for corrective actions to maintain system stability and avoid potential grid violations.

[0044] The method further comprises overriding the received power distribution schedule when local conditions indicate a risk of exceeding contractual grid capacity limits. Overriding refers to the ability of the local execution unit to modify or disregard the originally transmitted control schedule if real-time measurements suggest that executing the schedule as planned would lead to non-compliance with grid constraints. Contractual grid capacity limits define the maximum permissible energy exchange between the local system and the grid, and exceeding these limits may result in financial penalties or grid instability. An effect of this arrangement is that the energy management system maintains compliance with regulatory and contractual requirements while dynamically adapting to unforeseen fluctuations in demand, generation, or external constraints.

[0045] The method further comprises transmitting real-time operational data from the local execution unit to the data processing unit. This transmission ensures continuous feedback between the execution layer and the computational layer, allowing the data processing unit to refine future power distribution schedules based on observed deviations, system performance metrics, and external influencing factors. An effect of this arrangement is that the control model can continuously improve its predictive accuracy and decision-making efficiency through an iterative data-driven approach.The method regulates energy consumption and generation behind-the-meter based on dynamically updated power distribution schedules, ensuring compliance with contractual grid capacity constraints, maximizing self-consumption of locally generated energy, and selectively adjusting asset operation in response to marketbased energy pricing conditions. By dynamically integrating real-time monitoring, predictive scheduling, and adaptive control mechanisms, this method enhances the efficiency and cost-effectiveness of distributed energy resource management, while mitigating risks associated with grid constraints and market volatility.

[0046] In an example, the computer-implemented method may include determining the power distribution schedule, wherein the determination comprises acquiring real-time and historical data, wherein the acquired data includes active and reactive power measurements, energy consumption and generation profiles, contractually defined time-dependent grid capacity limits, and market price signals for energy trading. It may be provided that real-time and historical data acquisition enables a more accurate prediction of future energy demand and supply by incorporating time-dependent trends and variations. Active and reactive power measurements allow the system to distinguish between real power consumption and power losses due to inductive or capacitive loads, which improves the precision of energy dispatch calculations. By considering contractually defined time-dependent grid capacity limits, the system can ensure compliance with predefined constraints, thereby avoiding penalties or excessive grid usage fees. Market price signals provide an additional layer of optimization by allowing energy assets to be dispatched in response to dynamic price fluctuations, leading to cost savings and improved financial efficiency.

[0047] In an example, the step of determining the power distribution schedule may comprise computing forecasted energy demand and generation profiles using a timeseries regression model trained on historical and real-time data. It may be provided that time-series regression modeling allows for the identification of patterns in energy consumption and generation, which improves the accuracy of forecasts. Training the model on both historical and real-time data ensures that it continuously adapts to changing conditions, enabling more effective energy scheduling. An effect of this feature is that the system can anticipate fluctuations in demand and generation, allowing for proactive energy asset control and reducing the likelihood of inefficiencies due to incorrect forecasting.In an example, the method may comprise executing a hierarchical control sequence, comprising generating a first-level power distribution model that determines a base load allocation plan for each asset by solving a constrained load balancing equation, based on the forecasted demand and contractually defined grid capacity limits. It may be provided that a hierarchical control sequence enables a structured decision-making process in which different levels of optimization are performed at different time horizons, leading to a more adaptable and responsive energy management system. The base load allocation plan provides a primary structure for distributing energy resources while ensuring compliance with grid capacity constraints. By solving a constrained load balancing equation, the system ensures that energy assets are allocated in a mathematically optimal way, reducing overall system losses and enhancing efficiency.

[0048] In an example, the method may comprise applying a second-level real-time correction model, operating on a rolling time horizon, to dynamically adjust asset operation based on short-term deviations from forecasted demand, market price variations exceeding a predefined threshold, and localized grid constraints detected in real-time. It may be provided that real-time correction allows the system to react to sudden changes in energy supply, demand, or market conditions, improving adaptability and efficiency. A rolling time horizon ensures that decisions are continuously updated based on the latest available data, rather than relying on outdated predictions. Adjusting asset operation based on market price variations exceeding a predefined threshold allows for opportunistic trading decisions that maximize financial benefits. Incorporating localized grid constraints ensures that system operations do not negatively impact local grid stability, improving reliability.

[0049] In an example, the method may comprise transmitting the computed power distribution schedule to the local execution unit. It may be provided that this transmission ensures that optimized control decisions made at the data processing level are effectively communicated to the execution layer, enabling real-time implementation. An effect of this arrangement is that distributed energy resources can be dynamically adjusted in accordance with the most recent optimization results, improving system efficiency and compliance with grid constraints.

[0050] In an example, the method may comprise monitoring real-time electrical parameters by measuring, by a power flow monitoring circuit, voltage, current, andpower levels at the grid connection point and at each controllable energy asset. It may be provided that continuous monitoring allows for immediate detection of anomalies, preventing system failures and inefficiencies. Voltage and current measurements ensure that energy assets operate within safe and efficient electrical limits, reducing equipment wear and improving reliability. Monitoring power levels at the grid connection point provides insight into total energy exchange with the grid, allowing for compliance with contractual constraints.

[0051] In an example, the method may comprise processing the measured data using a real-time control logic module, comprising a load prioritization submodule that enforces asset-specific power constraints to prevent exceeding contractual grid limits. It may be provided that load prioritization ensures that critical loads receive power before non-essential loads, improving overall system efficiency. Asset-specific power constraints allow for differentiated treatment of various energy resources, leading to better resource allocation and utilization.

[0052] In an example, the method may comprise a deviation correction submodule that overrides the received power distribution schedule when real-time power deviations exceed a predefined tolerance window. It may be provided that deviation correction allows for immediate adjustments to unexpected variations in energy demand or supply, ensuring continuous compliance with grid constraints. The predefined tolerance window ensures that minor fluctuations do not trigger unnecessary interventions, balancing responsiveness with system stability.

[0053] In an example, the method may comprise transmitting the measured data to the data processing unit and receiving updated control schedules. It may be provided that this feedback loop allows for continuous refinement of optimization strategies based on observed system behavior, improving long-term efficiency and reliability. An effect of this feature is that the system becomes progressively more accurate in forecasting and decision-making, reducing energy waste and improving financial performance.

[0054] In an example, the method may comprise analyzing market trading conditions by retrieving day-ahead market prices, intraday trading signals, and balancing market incentives. It may be provided that integrating market data into the energy management process allows for cost-optimized scheduling decisions. Retrieving day-ahead prices ensures that energy assets can be scheduled for optimal participation in energy markets. Intraday trading signals provide real-time updates on marketfluctuations, allowing for adjustments that take advantage of short-term pricing opportunities. Balancing market incentives encourage participation in grid stability mechanisms, leading to financial rewards for providing grid-supportive services. In an example, the method may comprise incorporating a market response function that adjusts the power distribution schedule to prioritize energy asset operation based on maximization of arbitrage opportunities, minimization of on-peak electricity costs, and avoidance of imbalance penalties. It may be provided that market-driven scheduling improves the financial viability of behind-the-meter energy assets. Maximizing arbitrage opportunities allows for strategic buying and selling of energy to generate revenue. Minimizing on-peak electricity costs reduces operational expenses by shifting consumption to lower-priced periods. Avoiding imbalance penalties ensures that the system does not incur financial losses due to deviations from scheduled energy dispatch.

[0055] In an example, the method may comprise dynamically modifying the schedule when market fluctuations exceed a predefined risk threshold. It may be provided that this arrangement ensures that the system does not expose itself to excessive market risk while still allowing for opportunistic trading. The predefined risk threshold balances responsiveness with financial stability, ensuring that modifications are made only when they provide a tangible benefit. An effect of this feature is that the system operates in a financially optimized manner while mitigating exposure to volatility.

[0056] In an example, the method may execute a hierarchical control sequence to determine the power distribution schedule. This sequence comprises a first-level model that generates a long-term, strategic base load allocation plan, for instance for the next 24-48 hours, based on forecasted demand and supply. This is complemented by a second-level real-time correction model, operating on a short, rolling time horizon, which dynamically adjusts asset operation in response to immediate deviations from the forecast, market price fluctuations, or localized grid events. This dual-horizon approach ensures the system combines long-term strategic efficiency with tactical, real-time responsiveness.

[0057] In an example, the method provides for contract-aware optimization, wherein the contractually defined grid capacity constraints are not static but are dynamically adjusted. This adjustment is based on specific contractual terms such as timedependent power purchase agreements (PPAs), real-time demand responsecontracts, or capacity-based tariffs like a Capaciteitsbeperkende Contract (CBC). The power distribution schedule is automatically recalculated whenever a contract modification signal is received from a utility or grid operator, ensuring the system maintains strict compliance with complex energy contracts to avoid penalties and capitalize on financial incentives.

[0058] In an example, the method includes a critical fail-safe mechanism through an autonomous override logic at the local execution unit. This logic empowers the local unit to override the centrally computed power distribution schedule if its own real-time monitoring detects an imminent risk of exceeding a contractual grid limit, for instance due to an unforeseen event on-site. This low-latency, local intervention ensures grid compliance and system resilience, overcoming the inherent communication delays and potential failure points of a purely centralized control architecture.

[0059] In an example, the method may incorporate a market response function that adjusts the power distribution schedule to prioritize energy asset operation based on financial signals. This function actively analyzes market trading conditions, including day-ahead and intraday prices, to reprioritize asset schedules for the maximization of arbitrage opportunities, the minimization of on-peak electricity costs, and the avoidance of imbalance penalties. This transforms the system from a simple operational controller into a financially-optimized agent that actively seeks to generate revenue and reduce costs.

[0060] In an example, the method is designed to be resilient against imperfect data conditions commonly found in real-world deployments. The data acquisition module performs robust data engineering, including imputing missing data points to handle incomplete information and applying validation logic to detect and filter erroneous or anomalous measurements. The system's architecture inherently mitigates the impact of delayed data by leveraging the local execution unit for time-critical decisions. This ensures high data integrity for the optimization models, preventing faulty data from causing suboptimal or unstable asset control and enhancing overall system reliability.

[0061] The accompanying drawings are included to provide a further understanding of the invention. These are incorporated in and constitute a part of the application. The drawings further illustrate embodiments of the invention and together with the description serve to explain the principle of the invention .Brief description of the drawings

[0062] Figure 1 is a flowchart illustrating a computer-implemented method for optimizing energy management behind-the-meter, according to an aspect of the present invention.

[0063] Figure 2 is a schematic block diagram of a system for optimizing energy management behind-the-meter, according to an aspect of the present invention.

[0064] Figure 3 is a schematic block diagram showing the functional modules of the data processing unit of the system of Figure 2, according to an aspect of the present invention.

[0065] Detailed description of the drawings

[0066] Figure 1 shows a flowchart illustrating the steps of the computer implemented method (100) for optimizing behind-the-meter energy management. This method represents a different approach from traditional, static energy management by establishing a dynamic, closed-loop control system. The method (100) is designed to be iterative, continuously refining its own performance by learning from real-world outcomes and new insights.

[0067] The method starts with step (110), in which data is received. The data is realtime and historical data from a plurality of energy assets. This step is performed by a central data processing unit and involves the aggregation of both real-time and historical data from a plurality of interconnected energy assets. Real-time data refers to high-frequency measurements capturing the immediate operational state of the system. For example, for a battery energy storage system (BESS), this includes parameters such as its State of Charge (SoC), its operational status (e.g., 'charging', 'discharging', 'standby'), its State of Health (SoH) as a percentage of its original capacity, and its instantaneous active and reactive power flow. For an EV charging station, this could for example be its status ('available', 'occupied', 'charging') and its current power draw. Such type of data is often collected via industrial communication protocols like MODBUS TCP, but the skilled person will appreciate that this is a mere example and that other types of communication standards may also be applicable. Historical data, conversely, comprises vast archives of these measurements, whichare invaluable for identifying patterns and training predictive models. For instance, historical consumption data from a facility can reveal predictable peak demand periods, while historical solar generation data, correlated with weather data, can be used to build an accurate production forecast model.

[0068] Next, in the following step (120), a power distribution schedule is determined. This is an important step of the method (100). The schedule is not a typical simple set of on / off commands but may be understood to be a detailed, time-indexed plan of power setpoints for each controllable asset over a future horizon (e.g., the next 24-48 hours or typically day ahead, intraday and imbalance horizon). A practical example of a schedule could be: "Battery: charge at 10 kW from 02:00 to 04:00 to leverage low overnight prices; discharge at 15 kW from 17:00 to 19:00 to shave the evening peak. EV charging station 3: modulate charging power to align with solar production, ensuring the vehicle is 80% charged by 07:00." Such a schedule is the output of a multi-variable control model, which is a sophisticated optimization algorithm. The algorithm's primary objective is to find an optimal balance between often-conflicting goals, subject to a set of hard constraints. A key innovation is the integration of timedependent grid capacity constraints. Unlike a fixed grid connection, many commercial users have contracts, for example in the Netherlands known as “Capaciteitsbeperkende Contracten (CBC)” or non-firm ATOs (Afname Transport Overeenkomst), where the allowed import / export power varies by time of day or is subject to signals from the grid operator. The model explicitly incorporates these variable limits into its optimization. Other crucial inputs include forecasted energy demand and supply and market-based pricing signals from various energy markets (e.g., day-ahead, intraday, and imbalance markets). The optimization algorithm weighs these inputs according to a clear hierarchy of priorities: 1) never exceeding the grid capacity limits (highest priority), 2) maximizing the efficient use of locally generated solar energy, and 3) generating revenue from trading on energy markets (lowest priority).

[0069] Once determined, the schedule is transmitted (130) to a local execution unit. This transmission occurs over a secure network connection, packaging the schedule into a machine-readable format like JSON to ensure robust and interoperable communication between the central and local units..At the local site, the schedule is received and implemented (140). The local execution unit acts as a protocol translator, converting the high-level setpoints from the schedule into low-level, hardware-specific commands. For instance, it will use MODBUS TCP write commands to set the active / reactive power setpoints for a battery inverter or to limit the maximum power output of a solar inverter.

[0070] Simultaneously and continuously, the monitoring (150) step is performed. The local execution unit uses power meters and direct communication with assets to monitor electrical parameters like voltage, current, and power at the grid connection point and at each individual asset. This provides the necessary ground truth to verify that the system is performing as instructed and to detect anomalies or safety issues.

[0071] One feature of the system to ensure system robustness is the overriding (160) step. This enables the local execution unit with the autonomy to deviate from the centrally-provided schedule when local conditions demand it. This is not a simple error state but a designed feature for resilience. A typical example is when the local unit's real-time monitoring detects that the total site consumption is rapidly approaching the contractual grid limit, perhaps due to an unforeseen start-up of a large machine. Even if the central schedule instructs an asset to draw more power, the local unit will override this instruction and instead curtail a flexible load to prevent a costly penalty for exceeding the grid capacity. This local, high-speed intervention (often at a subsecond or millisecond level) is something a centralized, cloud-only system cannot achieve due to communication latency.

[0072] Finally, the feedback loop is closed by transmitting (170) operational data back to the central unit. This is not just raw data, but contextualized information, including any deviations from the plan and the reasons for any overrides. This rich feedback allows the central data processing unit to continuously learn and improve its models, making future schedules more accurate and resilient.

[0073] Figure 2 provides a schematic block diagram shown by way of example, of a system (200) designed to implement the method (100). The figure highlights the hybrid architecture, which is key to the invention. The system is composed of two distinct but tightly integrated levels: a centralized intelligence level and a decentralized execution level.

[0074] The centralized level is embodied by the data processing unit (DPU) (210). In a typical implementation, this is a scalable software platform hosted in a cloudenvironment. The DPU (210) serves as the system's "central computational hub" responsible for all computationally intensive tasks. These include storing and processing vast amounts of historical data, training and running sophisticated machine learning models for forecasting, and executing complex optimization algorithms (e.g., mixed-integer linear-programming (MILP) and satisfiability-modulo-theories (SMT) solvers) to generate the power distribution schedules. The DPU operates with a strategic, long-term perspective, planning for horizons of hours to days. Its role is analogous to that of a central command center, which has the overview and processing power to make complex, data-driven strategic decisions.

[0075] The decentralized level is embodied by the local execution unit (LEU) (220). This is a physical, ruggedized hardware device, such as an industrial PC or an embedded gateway, installed on-premise at the site where the energy assets are located. The LEU (220) acts as the system's "local agent" or "on-site commander" or “local gateway”. Its function is primarily operational and tactical. It receives the strategic plans (the schedule) from the DPU (210) and is responsible for their flawless execution. This involves translating the high-level schedule into low-level, real-time control signals using asset-specific protocols like MODBUS TCP or CAN-bus. Crucially, the LEU (220) is equipped with its own processing capabilities to perform continuous, high-frequency monitoring and to execute the autonomous override logic. This local intelligence ensures that the system can react to immediate local events in milliseconds and can maintain safe, stable operation even if the connection to the central DPU is temporarily lost. This hybrid architecture overcomes the key limitations of prior art: it avoids the latency and responsiveness issues of purely cloud-based systems, while also avoiding the lack of holistic oversight of purely decentralized systems.

[0076] The system (200) is capable of managing from a single energy asset but more typically, a plurality of energy assets (230), illustrated as (230') through (230""). This multi-asset environment is a key challenge the invention solves. The assets can be highly diverse and are not particularly limited to just one:

[0077] - A battery energy storage system (230'): The system can control its charge / discharge cycles to perform peak shaving, energy arbitrage, or provide grid services.- A solar PV installation (230"): The system controls the inverter to manage its power output, for instance, by curtailing production to avoid grid congestion or to optimize self-consumption.

[0078] - EV charging infrastructure (230'"): The system can implement smart charging strategies, dynamically adjusting the charging power of multiple vehicles based on grid capacity, solar availability, and user preferences.

[0079] - Other controllable loads (230""): This can include any thermal assets like heat pumps or industrial assets like furnaces, whose operation can be shifted in time to provide flexibility.

[0080] The communication architecture shown is deliberate. The link between the DPU (210) and LEU (220) may be any suitable connection but typically a WAN connection, which is secured and managed for reliability. The link between the LEU (220) and the assets (230) is a local, high-speed network, ensuring the low latency required for realtime control.

[0081] Figure 3 presents a detailed functional block diagram of the software architecture of the data processing unit (210). The blocks represent logical software modules that work in concert to deliver the system's core intelligence.

[0082] The data acquisition module (211) is the system's primary interface to the outside world. It is a highly robust and versatile module designed to ingest data from a wide array of sources. In practice, it contains a library of connectors for different protocols and APIs. For instance, it can pull data from electricity meters and industrial assets via MODBUS TCP, communicate with EV chargers using OCPP, and connect to the cloud backends of various battery and inverter manufacturers via their proprietary REST APIs. This module is also responsible for crucial data engineering tasks: parsing different data formats, handling time zone conversions, imputing missing data points, and transforming all incoming data into a standardized internal format for consistent processing by downstream modules.

[0083] The predictive modeling module (212) functions as the system's forecasting engine. It leverages the clean, standardized data from the acquisition module (211) to generate accurate predictions about future conditions. This module heavily relies on machine learning. For example, to forecast a building's energy consumption, it might use a time-series regression model like an LSTM (Long Short-Term Memory) network, which is trained on historical consumption data as well as exogenous variables likeweather forecasts (temperature, solar irradiance) and calendar information (day of the week, holidays). The output is not a single value but a probabilistic forecast over a defined horizon (e.g., the next 48 hours), providing not just an expected value but also a measure of uncertainty.

[0084] The hierarchical control sequence generator (213) is the optimization core of the system. This module takes the forecasts from module (212), combines them with the system's constraints (e.g., grid capacity, asset limitations) and objectives (e.g., cost minimization), and calculates the optimal power distribution schedule. The term "hierarchical" is key, referring to a multi-stage optimization process that mirrors different planning horizons. A first-level power distribution model, often running once a day (e.g., at 09:00), creates a long-term strategic plan for the next day. This is where the system decides on its day-ahead market strategy. A second-level real-time correction model operates on a much shorter, rolling horizon (e.g., every 15 minutes). This model uses a Model Predictive Control (MPC) approach to continuously refine the long-term plan based on the very latest real-time measurements and updated forecasts, ensuring the system can adeptly respond to unforeseen deviations.

[0085] The schedule transmission module (214) is the system's "dispatcher." It takes the complex schedule object generated by the optimizer (213), serializes it into an efficient data format (like JSON), and manages its secure and reliable transmission to the correct local execution unit.

[0086] While physically located on the LEU, modules (215), (216), and (217) are shown to illustrate the complete control loop. The power flow monitoring circuit (215) represents the physical sensors providing the ground-truth data. The real-time control logic module (216) contains the local intelligence. Its load prioritization submodule ensures that during an override event, critical loads are never curtailed. Its deviation correction submodule implements the logic for when and how to override the central schedule, using predefined tolerance windows to avoid overreacting to minor fluctuations.

[0087] Finally, the bidirectional communication interface (217) represents the comprehensive communication infrastructure that enables the seamless flow of data between all these modules and between the DPU and the LEU, forming the backbone of this dynamic, adaptive, and intelligent energy management system.

Claims

CLAIMS1. A computer-implemented method (100) for optimizing energy management behind-the-meter in a multi-asset environment, the method comprising:receiving (110), by a data processing unit, real-time and historical data from a plurality of energy assets, the energy assets including at least one of batteries, electric vehicle (EV) charging stations, solar inverters, or controllable loads;determining (120) a power distribution schedule by processing the received data in accordance with a predefined multi-variable control model, wherein the model accounts for:- time-dependent grid capacity constraints,- forecasted energy demand and supply, and- market-based pricing signals;transmitting (130) the computed power distribution schedule to a local execution unit;receiving and implementing (140) the power distribution schedule at the local execution unit by adjusting the operation of the energy assets;monitoring (150), in real-time, electrical parameters including voltage, current, and power, at the grid connection point and across the energy assets;overriding (160) the received power distribution schedule when local conditions indicate a risk of exceeding contractual grid capacity limits;transmitting (170) real-time operational data from the local execution unit to the data processing unit;wherein the method regulates energy consumption and generation behind-the-meter based on dynamically updated power distribution schedules, ensuring compliance with contractual grid capacity constraints, maximizing self-consumption of locally generated energy, and selectively adjusting asset operation in response to market-based energy pricing conditions.

2. The computer-implemented method (100) of claim 1, wherein:the step of determining (120) the power distribution schedule comprises:- acquiring real-time and historical data, wherein the acquired data includes:- active and reactive power measurements,- energy consumption and generation profiles,- contractually defined time-dependent grid capacity limits, and- market price signals for energy trading;- computing forecasted energy demand and generation profiles using a timeseries regression model trained on historical and real-time data;- executing a hierarchical control sequence, comprising:- generating a first-level power distribution model that determines a base load allocation plan for each asset by solving a constrained load balancing equation, based on the forecasted demand and contractually defined grid capacity limits;- applying a second-level real-time correction model, operating on a rolling time horizon, to dynamically adjust asset operation based on:- short-term deviations from forecasted demand, - market price variations exceeding a predefined threshold, and - localized grid constraints detected in real-time;- transmitting the computed power distribution schedule to the local execution unit;the step of monitoring (105) real-time electrical parameters comprises:- measuring, by a power flow monitoring circuit, voltage, current, and power levels at the grid connection point and at each controllable energy asset;- processing the measured data using a real-time control logic module, comprising:- a load prioritization submodule that enforces asset-specific power constraints to prevent exceeding contractual grid limits;- a deviation correction submodule that overrides the received power distribution schedule when real-time power deviations exceed a predefined tolerance window;- transmitting the measured data to the data processing unit and receiving updated control schedules;wherein the method dynamically optimizes energy flows behind-the-meter by executing a hierarchical control process, ensuring compliance with contractually defined grid capacity limits, maximizing self-consumption of locally generated energy, and selectively engaging in market-based trading when cost-beneficial conditions are detected.

3. The computer-implemented method (100) of claim 2, wherein the step of determining (120) the power distribution schedule further comprises:- analyzing market trading conditions by retrieving day-ahead market prices, intraday trading signals, and balancing market incentives;- incorporating a market response function that adjusts the power distribution schedule to prioritize energy asset operation based on:- maximization of arbitrage opportunities,- minimization of on-peak electricity costs, and- avoidance of imbalance penalties;- dynamically modifying the schedule when market fluctuations exceed a predefined risk threshold.

4. The computer-implemented method (100) of claim 2 or 3, wherein the local execution unit further comprises:a load classification module that categorizes energy assets into:- critical loads (e.g., industrial machinery, refrigeration),- flexible loads (e.g., HVAC systems, EV charging), and- storage assets (e.g., batteries, supercapacitors);a priority dispatch model that adjusts power consumption based on load classification, ensuring:- uninterrupted operation of critical loads,- dynamic curtailment of flexible loads during grid constraints, and- optimized charging / discharging cycles for storage assets based on forecasted demand and market pricing.

5. The computer-implemented method (100) of claim 2, wherein the real-time control logic module further comprises:a grid congestion prediction submodule, configured to:- retrieve grid congestion signals from a distribution system operator (DSO), - monitor voltage fluctuations and reactive power imbalances, and- compute probabilistic congestion forecasts based on past grid performance data;a congestion mitigation protocol, configured to:- reduce power draw from the grid during projected congestion periods, - increase self-consumption of locally generated energy, and- trigger preemptive load shedding for non-essential assets.

6. The computer-implemented method (100) of claim 2, wherein the contractually defined grid capacity constraints are dynamically adjusted based on:time-dependent power purchase agreements (PPAs) that specify variable import / export limits;real-time demand response contracts that allow for temporary grid flexibility in exchange for financial incentives;capacity-based tariffs that penalize peak demand exceedances while rewarding off-peak load shifting;wherein the power distribution schedule is recalculated whenever a contract modification signal is received from a utility or grid operator.

7. The computer-implemented method (100) of claim 2, wherein the local execution unit further comprises:a fault detection module that continuously monitors for:- power surges or undervoltage events,- abnormal asset behavior (e.g., excessive cycling of a battery system), and - loss of communication with the data processing unit;a failover control protocol, configured to:- isolate faulty assets to prevent cascading failures,- revert to a predefined emergency energy profile in case of control loss, and - log and transmit error diagnostics upon fault resolution.

8. A system (200) for optimizing energy management behind-the-meter in a multiasset (230’, 230”, 230”’, 230””) environment, the system comprising:a data processing unit (210), configured to:receive real-time and historical data from a plurality of energy assets(230’, 230”, 230’”, 230””), the energy assets including at least one of batteries, electric vehicle (EV) charging stations, solar inverters, or controllable loads;determine a power distribution schedule by processing the received data in accordance with a predefined multi-variable control model, wherein the model accounts for:time-dependent grid capacity constraints,forecasted energy demand and supply, and market-based pricing signals;transmit the computed power distribution schedule to a local execution unit (220);a local execution unit (220), configured to:receive and implement the power distribution schedule by adjusting the operation of the energy assets (230’, 230”, 230”’, 230””);continuously monitor electrical parameters, including voltage, current, and power, at the grid connection point and across the energy assets (230’, 230”, 230’”, 230””);override the received power distribution schedule when local conditions indicate a risk of exceeding contractual grid capacity limits;transmit real-time operational data to the data processing unit (210); wherein the system regulates energy consumption and generation behind-the-meter based on dynamically updated power distribution schedules, ensuring compliance with contractual grid capacity constraints, maximizing self-consumption of locally generated energy, and selectively adjusting asset operation in response to market-based energy pricing conditions.

9. The system (200) of claim 8, wherein the data processing unit (210) comprises:a data acquisition module (211) configured to receive real-time and historical data from the energy assets (230’, 230”, 230’”, 230””), wherein the data includes:active and reactive power measurements,energy consumption and generation profiles,contractually defined time-dependent grid capacity limits, and market price signals for energy trading;a predictive modeling module (212) configured to compute forecasted energy demand and generation profiles using a time-series regression model trained on historical and real-time data;a hierarchical control sequence generator (213), implemented as a multilayered control process, comprising:a first-level power distribution model configured to determine a base load allocation plan for each asset by solving a constrained load balancing equation, based on the forecasted demand and contractually defined grid capacity limits;a second-level real-time correction model, operating on a rolling time horizon, configured to dynamically adjust asset operation based on:short-term deviations from forecasted demand,market price variations exceeding a predefined threshold, and localized grid constraints detected in real-time;a schedule transmission module (214) for communicating the computed power distribution schedule to the local execution unit; and wherein the local execution unit comprises:a power flow monitoring circuit (215) configured to measure voltage, current, and power levels at the grid connection point and at each controllable energy asset (230’, 230”, 230’”, 230””);a real-time control logic module (216) comprising:a load prioritization submodule that enforces asset-specific power constraints to prevent exceeding contractual grid limits,a deviation correction submodule that overrides the received power distribution schedule when real-time power deviations exceed a predefined tolerance window;a bidirectional communication interface (217) for transmitting real-time operational data to the data processing unit and receiving updated control schedules;wherein the system dynamically optimizes energy flows behind-the-meter by executing a hierarchical control process, ensuring compliance with contractually defined grid capacity limits, maximizing self-consumption of locally generated energy, and selectively engaging in market-based trading when cost-beneficial conditions are detected.

10. A computer program product comprising instructions which, when executed by a data processing unit (210) of a system (200) for optimizing energy managementbehind-the-meter in a multi-asset environment, cause the data processing unit to perform the method of claim 8.

11. A non-transitory computer-readable storage medium storing a computer program comprising instructions which, when executed by a processor, cause the processor to perform the method of claim 1.

12. A computer-readable data carrier comprising computer-executable instructions which, when executed by a data processing unit, cause the system to execute the method of claim 1, wherein the data carrier is at least one of:a semiconductor memory,a magnetic storage device,an optical disc, ora network-accessible data storage system.

Images