Optimized regulation and control method and system for interaction between electricity-hydrogen-heat distributed energy supply system and power grid

Abstract

The invention discloses an optimized regulation and control method and system for interaction between an electricity-hydrogen-heat distributed energy supply system and a power grid, and belongs to the technical field of electric power system control, and the method comprises the steps: generating an operation data set and a power grid parameter set; mapping the time sequence data to equipment accumulated damage by using a health assessment model, calculating to obtain a loss degree parameter, and generating a health degree index; constructing a multi-target game model, and constructing a comprehensive revenue function including economy, reliability and health penalty terms; determining a Nash equilibrium point as an optimization strategy; and packaging the optimization strategy into a regulation and control instruction with an execution timestamp, issuing and executing the regulation and control instruction, monitoring an actual response state, comparing the actual response state with a target output quantity to generate a report, and correcting a loss accumulation weight in the health assessment model and a comprehensive revenue function parameter in the multi-target game model in real time based on deviation. According to the method, the complexity and uncertainty in the operation of the multi-energy system can be effectively handled, and the self-adaptability, stability and operation efficiency of the system are improved.

Description

Technical Field

[0001] This invention belongs to the field of power system control technology, and in particular relates to an optimized control method and system for the interaction between an electric hydrogen-thermal distributed energy supply system and the power grid. Background Technology

[0002] The distributed energy supply system (EDS) is a comprehensive energy system integrating multiple energy production, conversion, and storage units. It typically includes photovoltaic (PV) systems, hydrogen fuel cells, electrolyzers, hydrogen storage devices, and heat pumps, enabling the coordinated supply and local consumption of multiple energy forms such as electricity, hydrogen, and heat. As an effective supplement to traditional centralized power grids, this type of system plays an increasingly important role in improving renewable energy utilization, ensuring regional energy supply reliability, and enhancing overall energy efficiency, making it a key component in building new power systems.

[0003] Currently, centralized optimization scheduling is commonly used for the optimization and control of such distributed energy supply systems. This typically relies on a central controller, which is responsible for collecting operating status data from all energy units in the system and load demand information from the power grid. Then, based on an established global system model, it uses linear programming or intelligent optimization algorithms to perform unified calculations to seek the scheduling strategy that minimizes the overall operating cost of the system or optimizes a single performance indicator. Finally, the calculated control commands are sent to each energy unit for execution.

[0004] However, existing centralized control schemes have inherent limitations. First, their control models often prioritize short-term economic benefits, lacking effective quantification and consideration of the health status and long-term wear and tear of energy unit equipment. This can lead to optimization strategies accelerating the physical aging of critical equipment while pursuing immediate economic gains. Second, this highly centralized control architecture places extremely high demands on the computing power and communication bandwidth of the central controller, and its response speed is limited, making it difficult to flexibly adapt to the rapid dynamic changes in distributed energy output and load demand. The system's robustness and scalability are also affected by the risk of single points of failure. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this invention is to provide an optimized control method and system for the interaction between an electric hydrogen-thermal distributed energy supply system and the power grid. This method employs a multi-objective dynamic game model that treats each energy unit as an independent participant and uses the equipment health index as the core endogenous variable in the game decision-making process. This enables the synergistic optimization of system economy, reliability, and equipment health lifespan, thereby improving the adaptive capability and overall control effect of the energy supply system.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An optimized control method for the interaction between an electro-hydrogen-thermal distributed energy supply system and the power grid, the method comprising: Acquire real-time operating data of each energy unit and real-time parameters of the power grid, and generate an operating dataset and a power grid parameter set; Time series data is extracted from the running dataset, and the time series data is mapped to the cumulative damage of the equipment using a health assessment model. The wear and tear parameter is calculated, and a health index that is negatively correlated with the wear and tear is generated. A multi-objective game model is constructed based on the health index and the power grid parameter set. Each energy unit and the power grid interface are independent game participants. A comprehensive payoff function is constructed that includes economic efficiency, reliability and a health penalty term that is negatively correlated with the health index. The distributed iterative algorithm drives each participant to independently update the strategy based on the gradient of the comprehensive benefit function, and determines whether the strategy update amount is less than the convergence threshold to determine the Nash equilibrium point as the optimization strategy. The optimization strategy is encapsulated as a control command with an execution timestamp and issued for execution. The actual response status is monitored and compared with the target output to generate a report. Based on the deviation, the cumulative loss weight in the health assessment model and the comprehensive payoff function parameters in the multi-objective game model are corrected in real time.

[0007] As a preferred embodiment of the present invention, for a multi-objective game model, a scheduling cycle timer is started, and at the beginning of each scheduling cycle, real-time running data and real-time parameters are acquired. Using the latest acquired running dataset and power grid parameter set, update the health index and reconstruct the multi-objective game model.

[0008] As a preferred embodiment of the present invention, generating the operating dataset and the power grid parameter set includes: The output power, efficiency, and temperature parameters of each energy unit are collected to form a subset of energy operation data; Collect frequency and load demand parameters of the power grid to form a subset of power grid parameters; The energy operation data subset and the power grid parameter subset are timestamped and merged to generate the operation dataset and the power grid parameter set.

[0009] As a preferred embodiment of the present invention, the calculation of the wear degree parameter and the generation of a health index negatively correlated with the wear degree include: Historical and current operating data of each energy unit are extracted from the operating dataset to form time series data; By mapping time series data to a health assessment model of cumulative equipment damage, the loss parameters of each energy unit are calculated. The loss parameter is normalized to generate a health index that is negatively correlated with the loss parameter.

[0010] As a preferred embodiment of the present invention, constructing a multi-objective game model includes: Identify the set of participants, including photovoltaic units, hydrogen fuel cell units, heat pump units, and grid interfaces; Based on the health index of each participant and the load demand of the power grid parameter set, the selectable range of energy output is dynamically determined to form a strategy space; Construct a comprehensive benefit function that includes three benefit sub-items: economy, reliability, and health.

[0011] As a preferred embodiment of the present invention, the construction of the comprehensive benefit function includes: Establish a sub-function to measure the economic benefits of fuel consumption and electricity purchase and sale costs; Establish a reliability benefit subfunction to measure grid frequency deviation and load satisfaction; Establish a health penalty term as a sub-function of health benefit; The economic benefit subfunction, reliability benefit subfunction, and health benefit subfunction are weighted and summed to form a comprehensive benefit function.

[0012] As a preferred embodiment of the present invention, driving each participant to independently update the strategy based on the gradient of the comprehensive benefit function, and determining the Nash equilibrium point as the optimization strategy includes: Initialize the initial strategies of each participant based on the health index and the power grid parameter set; The distributed iterative algorithm drives each participant to independently update its own policy based on the gradient of the comprehensive reward function. It determines whether the policy update amount of all participants is less than the preset convergence threshold. If so, the current policy combination is determined to be the Nash equilibrium point, and the optimized policy is output.

[0013] As a preferred embodiment of the present invention, encapsulating the optimization strategy into a control instruction with an execution timestamp and issuing it for execution includes: Analyze and optimize the strategy to extract the target output for each energy unit; The target output quantity and execution timestamp are encapsulated to generate control instructions; Control commands are sent to the execution controllers of each energy unit via a communication interface.

[0014] As a preferred embodiment of the present invention, the cumulative loss weight in the deviation real-time correction health assessment model and the comprehensive payoff function parameters in the game model include: After issuing control instructions, monitor the actual response status of each energy unit, compare the actual response status with the target output in the control instructions, and generate a response status report. Based on the deviations presented in the response status report, adjust the internal parameters in the health assessment model and the multi-objective game model; The adjusted model parameters are used for the next round of optimization and control to improve the model's accuracy and adaptive performance.

[0015] An optimized control system for interaction between an electro-hydrogen-thermal distributed energy supply system and the power grid, used to implement the above method, includes: The data acquisition module is used to acquire real-time operating data of each energy unit and real-time parameters of the power grid, and generate operating datasets and power grid parameter sets. The health calculation module is used to extract time series data based on the running dataset, map the time series data to the cumulative damage of the equipment using a health assessment model, calculate the wear degree parameter, and generate a health index that is negatively correlated with the wear degree. The game modeling module is used to construct a multi-objective game model based on the health index and the power grid parameter set. It takes each energy unit and the power grid interface as independent game participants and constructs a comprehensive payoff function that includes economic efficiency, reliability and health penalty terms that are negatively correlated with the health index. The equilibrium solution module is used to drive each participant to independently update the policy according to the gradient of the comprehensive benefit function through a distributed iterative algorithm, and to determine whether the policy update amount is less than the convergence threshold, and to determine the Nash equilibrium point as the optimization policy. The instruction generation and correction module is used to encapsulate the optimization strategy into control instructions with execution timestamps, issue them for execution, monitor the actual response status and compare it with the target output to generate a report, and correct the cumulative loss weight in the health assessment model and the comprehensive payoff function parameters in the game model in real time based on the deviation.

[0016] The present invention has the following advantages: This invention achieves deep synergistic optimization among system economy, operational reliability, and equipment health lifespan by constructing a multi-objective game model with equipment health as the core parameter. This method does not simply weigh multiple objectives; instead, it uses an embedded health penalty mechanism to ensure that actions protecting equipment health have direct economic rationality in game decision-making. This naturally improves the long-term stability and economy of energy supply while extending the lifespan of critical equipment within the system.

[0017] The decentralized game-theoretic decision-making mechanism employed in this invention enhances the system's adaptability and robustness. Because it eliminates the need for a single central controller for global computation and command distribution, it can respond more quickly to changes in local operating conditions, such as sudden changes in the health status of individual energy units or instantaneous load fluctuations. Each energy unit, as an independent participant, can spontaneously adjust its strategy and quickly converge to a new operational equilibrium point, improving its stability and self-healing capabilities in complex and volatile environments.

[0018] This invention achieves continuous optimization throughout the entire lifecycle through periodic closed-loop control and model self-calibration. This method not only adapts to dynamic changes in the external environment but also, by monitoring the deviation between actual response and model predictions, corrects the internal health assessment and game theory model parameters, demonstrating its learning and self-evolution capabilities. This ensures that the effectiveness of the optimization control does not diminish over time, thereby guaranteeing the long-term, efficient, and reliable operation of the entire distributed energy supply system and maximizing the lifecycle value of the assets. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an optimized control method for the interaction between an electro-hydrogen-thermal distributed energy supply system and the power grid, according to the present invention. Figure 2 This is a graph showing the changes in the power and health index of the hydrogen fuel cell in Embodiment 1 of the present invention. Figure 2 (a) in the figure is a graph showing the power variation of a hydrogen fuel cell; Figure 2 (b) in the figure is a curve showing the change in the health index; Figure 3 This is a scatter plot showing the deviation between the actual output and the command of the heat pump unit in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the structure of an optimized control system for interaction between an electric hydrogen-thermal distributed energy supply system and the power grid, according to the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0021] Example 1: As Figure 1 As shown, an optimized control method for the interaction between an electro-hydrogen-thermal distributed energy supply system and the power grid specifically includes: Acquire real-time operating data of each energy unit and real-time parameters of the power grid, and generate an operating dataset and a power grid parameter set; Time series data is extracted from the running dataset, and the time series data is mapped to the cumulative damage of the equipment using a health assessment model. The wear and tear parameter is calculated, and a health index that is negatively correlated with the wear and tear is generated. A multi-objective game model is constructed based on the health index and the power grid parameter set. Each energy unit and the power grid interface are independent game participants. A comprehensive payoff function is constructed that includes economic efficiency, reliability and a health penalty term that is negatively correlated with the health index. The distributed iterative algorithm drives each participant to independently update the strategy based on the gradient of the comprehensive benefit function, and determines whether the strategy update amount is less than the convergence threshold to determine the Nash equilibrium point as the optimization strategy. The optimization strategy is encapsulated into a control command with an execution timestamp and issued for execution. The actual response status is monitored and compared with the target output to generate a report. Based on the deviation, the cumulative loss weight in the health assessment model and the comprehensive payoff function parameters in the game model are corrected in real time.

[0022] By acquiring real-time data, a snapshot of the instantaneous state of the energy system and power grid is obtained. Instead of directly using operational data for optimization, an intermediate step is taken to abstract and quantify complex data reflecting equipment operating history and conditions into a single-dimensional health index. Subsequently, the traditional centralized optimization problem is reconstructed into a distributed multi-objective game model, where each energy unit and power grid interface becomes a participant with independent decision-making power. The essence of the game model lies in the design of the strategy space and the comprehensive payoff function. The range of optional strategies for each participant is dynamically constrained by their own health index, and their pursued payoffs not only include traditional economic and reliability objectives but also creatively introduce a health penalty term negatively correlated with the health index. Finally, by solving for the Nash equilibrium point of this game model, a stable and mutually optimal strategy combination reached by all participants after comprehensively considering economy, reliability, and their own health is obtained. This strategy is then transformed into precise control commands for physical equipment.

[0023] For the multi-objective game model, a scheduling cycle timer is started, and at the beginning of each scheduling cycle, real-time running data and real-time parameters are acquired. A time-driven cyclic execution mechanism is introduced to achieve continuous rolling optimization of the electric-hydrogen-thermal distributed energy supply system. A scheduling cycle timer needs to be set and started. The timer's period, i.e., the scheduling cycle, is preset based on the response characteristics and load change rate, for example, 15 minutes or 30 minutes. At the beginning of each scheduling cycle, the scheduling cycle timer generates a trigger signal. This signal first activates data acquisition, causing it to immediately execute the acquisition of real-time operating data from each energy unit and real-time grid parameters, thereby generating the latest operating dataset and grid parameter set for the current moment.

[0024] Using the latest acquired running dataset and power grid parameter set, update the health index and reconstruct the multi-objective game model.

[0025] Using the latest operational dataset, health calculations are invoked to recalculate the health index of each energy unit, reflecting changes in the operating status and accumulated losses of the equipment during the most recent scheduling cycle. The updated health index, along with the latest grid parameter set, is then fed into the game model. Based on these latest inputs, the multi-objective game model is reconstructed. This reconstruction process is dynamic, manifested in two aspects: first, the strategy space of each participant is adjusted according to its latest health index; the energy output range of units with lower health may be narrowed to avoid overload operation that exacerbates their losses; second, the comprehensive payoff function of the multi-objective game model is also updated, particularly the health penalty term, which is adjusted according to the new health index, while the economic and reliability sub-terms are also updated due to changes in grid parameters. After model reconstruction, the Nash equilibrium point is solved, generating and issuing new control commands. The entire process from data acquisition to command issuance is repeated in each scheduling cycle, forming a closed-loop, real-time rolling optimization process.

[0026] Acquire real-time operating data of each energy unit and real-time parameters of the power grid, and generate operating datasets and power grid parameter sets, including: The output power, efficiency, and temperature parameters of each energy unit are collected to form a subset of energy operation data; For each energy unit in the distributed energy supply system (electro-hydrogen-thermal), such as photovoltaic units, hydrogen fuel cell units, and heat pump units, data is collected through a pre-deployed sensor network located inside or nearby. The collected physical quantities include output power reflecting the energy conversion state, real-time efficiency measuring conversion performance, and temperature parameters of key equipment components characterizing operating load and loss status. These multi-data points collected at the same time are integrated and timestamped to form a subset of energy operation data.

[0027] Collect frequency and load demand parameters of the power grid to form a subset of power grid parameters; Key operating parameters of the power grid are collected in real time through monitoring equipment located at a common coupling point connected to the grid. These mainly include grid frequency, which reflects grid stability, and real-time load demand parameters, which characterize the supply-demand relationship. This grid-side data is also appended with precise timestamps, forming a subset of grid parameters.

[0028] The energy operation data subset and the power grid parameter subset are timestamped and merged to generate the operation dataset and the power grid parameter set.

[0029] Because the sampling periods and communication delays of sensors and grid monitoring equipment in different energy units may differ, directly using these raw data subsets can lead to data misalignment in the time dimension. To address this issue, timestamp alignment and merging are performed. First, a unified scheduling time base is established. Then, each data record in the energy operation data subset and grid parameter subset is mapped to the nearest base time point based on its timestamp. For data missing at a certain base time point, methods such as preserving previous values ​​or linear interpolation are used to fill in the gaps, ultimately generating a fully time-synchronized operation dataset and grid parameter set.

[0030] Time series data is extracted from the operational dataset, and a health assessment model is used to map the time series data to cumulative equipment damage. This calculates the wear and tear parameters and generates health indices negatively correlated with the wear and tear, including: Historical and current operating data of each energy unit are extracted from the operating dataset to form time series data; Key operational data for each individual energy unit, including historical data and data from the current scheduling cycle, are extracted. These data points primarily consist of output power, efficiency, and temperature parameters. Arranged chronologically, these data points form a time-series dataset that comprehensively records the unit's operational trajectory.

[0031] By mapping time series data to a health assessment model of cumulative equipment damage, the loss parameters of each energy unit are calculated. The time-series data is processed using a health assessment model to quantify the cumulative damage to the equipment. A health assessment model is a pre-established mathematical or physical model whose core function is to map the various stresses experienced by the equipment during operation into irreversible physical wear and tear. By inputting the time-series data into this model, it calculates and outputs comprehensive wear and tear parameters. These parameters monotonically increase, reflecting the total cumulative wear and tear on the equipment from its initial state to the current moment.

[0032] The loss parameter is normalized to generate a health index that is negatively correlated with the loss parameter.

[0033] To make the wear and tear parameters more universal and comparable, normalization is required to generate the final health index. The normalization operation converts absolute wear and tear values ​​into a relative health status indicator. The calculation method is as follows: ; H represents the health index, which typically ranges from 0 to 1. 1 indicates that the equipment is in brand new or ideal health condition, while 0 indicates that the equipment is ready for scrap or requires immediate overhaul. It is the current cumulative wear and tear parameter calculated through a health assessment model. This is a preset maximum wear threshold, representing the total wear corresponding to the end of the equipment's design life. It is typically provided by the equipment manufacturer or calibrated using offline aging test data. This formula ensures that the health index H and the wear parameter are related. There is a clear negative correlation between them; that is, the more wear and tear a device accumulates, the lower its health index becomes.

[0034] A multi-objective game model is constructed based on the health index and the power grid parameter set, with each energy unit and the power grid interface as independent game participants, including: Identify the set of participants, including photovoltaic units, hydrogen fuel cell units, heat pump units, and grid interfaces; The participants in the game are defined as independent energy decision-making units, including photovoltaic units, hydrogen fuel cell units, and heat pump units on the distributed power generation side, as well as the grid interface, which serves as the energy exchange medium between the distributed energy supply system (electricity, hydrogen, and heat) and the main power grid. Designating these units as participants implies that each possesses independent decision-making capabilities and pursues the maximization of its own gains within the game framework.

[0035] Based on the health index of each participant and the load demand of the power grid parameter set, the selectable range of energy output is dynamically determined to form a strategy space; A strategy space is constructed for each participant, which is the set of actions each participant can choose. For each energy unit and grid interface, the strategy is the amount of energy output or exchange in the next scheduling cycle. This strategy space is not fixed but dynamically formed. The boundaries of each participant's strategy, especially the upper limit of energy output, are directly constrained by its own health index. The lower the health index of an energy unit, the lower the upper limit of its selectable energy output will be. This is a protective constraint to prevent equipment from being overused when its health is poor. At the same time, the total load demand in the grid parameter set also jointly determines the effective domain of all participants' strategy combinations, that is, the sum of the energy outputs of all participants must meet the total load demand.

[0036] Construct a comprehensive benefit function that includes three benefit sub-items: economy, reliability, and health.

[0037] A comprehensive benefit function is constructed for each participant. This comprehensive benefit function is a holistic indicator consisting of three benefit sub-items, corresponding to economic efficiency, reliability, and health. The economic benefit sub-item quantifies the economic benefits or costs obtained by the participant through energy production or exchange; the reliability benefit sub-item reflects the contribution or impact of the participant's strategy on stable operation; and the health benefit sub-item is a penalty term negatively correlated with the health index. When the strategy chosen by the participant causes significant potential damage to its own health, this item will give a negative benefit, thus reflecting the consideration of equipment lifespan depletion in decision-making.

[0038] Constructing a comprehensive benefit function that includes three benefit components: economy, reliability, and health, involves: Establish a sub-function to measure the economic benefits of fuel consumption and electricity purchase and sale costs; The economic benefit sub-function aims to quantify the direct economic impact of participants' strategies. For fuel-consuming energy units such as hydrogen fuel cells, this function calculates the corresponding fuel consumption cost based on the output power; for grid interfaces, it calculates the electricity purchase cost or electricity sales revenue based on the electricity exchanged with the grid and real-time time-of-use pricing. The goal of this sub-function is to maximize economic benefits or minimize operating costs.

[0039] Establish a reliability benefit subfunction to measure grid frequency deviation and load satisfaction; A reliability benefit subfunction is established to evaluate the contribution of participant strategies to the overall energy supply system and grid stability. It is primarily measured by two indicators: first, load sufficiency, which is the deviation between the total output power of all energy units and the actual load demand of the system; the smaller the deviation, the higher the reliability benefit. Second, grid frequency deviation, which is the impact of participants' energy injection or absorption behavior on grid frequency stability; strategies that contribute to frequency stability will achieve higher reliability benefits.

[0040] Establish a health penalty term as a sub-function of health benefit; The health benefit sub-function directly reflects the concern for the long-term health of the equipment and is constructed as a health penalty term. This function links the output power strategy selected by an energy unit with its own health index. If a strategy causes the equipment to operate under higher stress, thereby accelerating its wear and tear, then the health benefit sub-function value corresponding to that strategy will be a large negative number, forming an effective penalty.

[0041] The economic benefit subfunction, reliability benefit subfunction, and health benefit subfunction are weighted and summed to form a comprehensive benefit function.

[0042] To integrate these three sub-functions with different physical meanings and dimensions into a unified evaluation standard, a weighted summation is used to construct the final comprehensive benefit function. Before summing, each benefit sub-function needs to be normalized, mapping its values ​​to a unified, dimensionless interval, such as 0 to 1, thus resolving the issue of direct calculations based on different dimensions. The formula for the comprehensive benefit function is as follows: ; Here, U represents the overall benefit of a particular participant. , and These are the normalized values ​​of the economic benefit subfunction, the reliability benefit subfunction, and the health benefit subfunction, respectively, all of which are functions of the participant's strategy. , and These are pre-set weighting coefficients, representing the system's emphasis on economy, reliability, and health, respectively, with a sum of 1. These weighting coefficients can be dynamically adjusted by the operator according to operational objectives at different times. For example, during peak load periods or grid disturbances, the reliability weight can be increased. When the overall health of equipment is low, the weight of health can be increased. .

[0043] A distributed iterative algorithm is used to drive each participant to independently update their policy based on the gradient of the comprehensive reward function. The algorithm then determines whether the policy update amount is less than the convergence threshold and identifies the Nash equilibrium point as the optimal policy. Initialize the initial strategies of each participant based on the health index and the power grid parameter set; Before the iteration begins, an initial strategy needs to be set for each participant, namely each energy unit and grid interface. That is, the initial energy output is reasonably estimated based on the latest health index and grid parameter set. For example, the initial load can be allocated proportionally according to the health status and rated power of each unit.

[0044] Through a distributed iterative algorithm, each participant is driven to independently update its own strategy based on the gradient of the comprehensive return function; After initialization, a distributed iterative algorithm is initiated. In each iteration, each participant updates their decision independently. Each participant calculates the gradient of its own overall reward function with respect to its own policy, based on the policies of all other participants in the previous iteration. This gradient indicates the direction in which the participant adjusts its policy to maximize its overall reward. Subsequently, each participant fine-tunes its current policy along this gradient direction, multiplying by a preset step size factor. The update rule is as follows: ; in, It is the new strategy of participant i in the next iteration. It is the current strategy of participant i. It is the step size factor that controls the iteration speed and stability. Participant Comprehensive return function Regarding one's own strategy The gradient is calculated by differentiating the comprehensive reward function. After each update, the new strategy also needs to be applied. Projecting back into the strategy space constrained by conditions such as the health index ensures the effectiveness of the strategy.

[0045] Determine whether the policy update amount of all participants is less than the preset convergence threshold. If yes, determine that the current policy combination is the Nash equilibrium point and output the optimized policy.

[0046] After all participants complete their policy updates in parallel, the system enters the convergence assessment phase. The system calculates the policy update amount for all participants from t to t+1 and determines if the update amount is less than a preset convergence threshold. If the policy changes for all participants are negligible, it indicates a stable state has been reached. This policy combination is considered the Nash equilibrium point of the multi-objective game model, the iteration process terminates, and this set of policies is output as the final optimized policy. If the convergence condition is not met, the system proceeds to the next iteration, repeating the policy update and assessment process.

[0047] Encapsulating optimization strategies into control instructions with execution timestamps and issuing them for execution includes: Analyze and optimize the strategy to extract the target output for each energy unit; The obtained Nash equilibrium point, i.e., the optimal strategy, is analyzed. The optimal strategy is a set containing the optimal actions of all participants. First, it is necessary to extract the specific target output corresponding to each energy unit from it. For example, the electrical power that the photovoltaic unit should output, and the electrical and thermal power that the hydrogen fuel cell unit should output, etc., are identified from the strategy set.

[0048] The target output quantity and execution timestamp are encapsulated to generate control instructions; The extracted purely digital target output is encapsulated into a structured information package to generate control instructions. To ensure the coordinated action of each unit, a clear execution timestamp must be written into the instruction during encapsulation. The execution timestamp is usually set to the start time of the next scheduling cycle, serving as a unified action command that instructs all energy units to begin executing the new output target at the same time.

[0049] Control commands are sent to the execution controllers of each energy unit via a communication interface.

[0050] These encapsulated control commands are then sent out via a communication interface. Through a pre-defined industrial Ethernet, wireless network, or other reliable communication interface, each control command is precisely sent to the local execution controller of the corresponding energy unit. The execution controller is the direct control unit of the energy unit, such as the digital signal processor of an inverter or the control board of a fuel cell. Upon receiving the control command, it parses the command and adjusts the internal operating parameters of the equipment when the execution timestamp arrives, ensuring that the output accurately tracks the target output quantity in the command.

[0051] The system monitors the actual response status and compares it with the target output to generate a report. Based on the deviation, it adjusts the cumulative loss weights in the health assessment model and the comprehensive payoff function parameters in the game theory model in real time, including: After issuing control instructions, monitor the actual response status of each energy unit, compare the actual response status with the target output in the control instructions, and generate a response status report. After the control commands are issued and executed, the actual operating status of each energy unit is continuously monitored, and real-time response status data such as output power and efficiency are collected through sensors. Then, these real-time collected response statuses are compared one by one with the target output quantities contained in the previously issued control commands to accurately calculate the deviation of each energy unit during command execution. This deviation data is organized and structured to form a response status report. This report not only records the magnitude of the deviation but also includes the time of occurrence and the corresponding equipment operating conditions, forming the original basis for model correction.

[0052] Based on the deviations presented in the response status report, adjust the internal parameters in the health assessment model and the multi-objective game model; Based on the systematic or persistent deviations presented in the response status report, a model parameter adjustment procedure is initiated. The adjustment is divided into two levels: First, adjusting the health assessment model. If an energy unit exhibits a significant decrease in efficiency or an abnormal increase in temperature before reaching its theoretical output limit, resulting in actual output lower than the target output, this indicates that the actual health degradation rate of this energy unit may be faster than the original model predicts. This deviation information will be used to correct the internal parameters in the health assessment model, such as increasing the weight of cumulative losses under specific operating conditions, so that the model can more accurately reflect the actual degradation process of the equipment. Second, adjusting the multi-objective game model. Response deviations may also stem from inaccuracies in the game model's description of the physical system, such as cost coefficients or efficiency curves used in the comprehensive payoff function that do not match reality. Deviation data will be used to reverse-correct these internal parameters in the multi-objective game model through identification and other algorithms, making the model more representative of the true characteristics of the physical entity.

[0053] The adjusted model parameters are used for the next round of optimization and control to improve the model's accuracy and adaptive performance.

[0054] Using these adjusted and optimized model parameters, a new round of optimization and control calculations will be performed in the next scheduling cycle, thus starting a new optimization cycle based on a more accurate model.

[0055] To verify the feasibility of this invention in practice, it was applied to a distributed energy supply system of electricity, hydrogen, and heat in a "green energy ecological industrial park." This industrial park is equipped with photovoltaic (PV) units, hydrogen fuel cell units, and heat pump units, and is connected to the main power grid via a grid interface, aiming to achieve a high proportion of renewable energy consumption and comprehensive and efficient energy utilization. Traditional dispatching methods often focus on instantaneous economic efficiency, which can easily lead to overload operation of some equipment, accelerating its aging and wear, and affecting long-term operational stability and economic efficiency. The industrial park hopes to adopt the method of this invention to ensure energy supply reliability and economic efficiency while incorporating equipment health management into optimized control, thereby maximizing comprehensive benefits throughout the entire life cycle.

[0056] In this embodiment, the central controller of the industrial park deploys the optimized control system of this invention. The optimized control is executed cyclically with a 15-minute scheduling cycle. At the beginning of each cycle, real-time operating data such as output power, efficiency, and temperature of each energy unit, as well as real-time parameters of the power grid interface, are acquired, and a timestamp-aligned operating dataset is generated. Based on historical and current operating data, a health index for each unit is calculated using a preset health assessment model. Subsequently, with each energy unit and the power grid interface as participants, a multi-objective game model including economic efficiency, reliability, and health penalty terms is constructed based on the latest health index and power grid parameters. The Nash equilibrium point of the model is solved using a distributed iterative algorithm to obtain the optimal energy output strategy for each participant. Finally, the optimized strategy is transformed into a control command with an execution timestamp and issued to the execution controller of each unit.

[0057] To verify the effectiveness of the method in this embodiment, operational data from April 10th to April 16th of a certain year was selected for analysis and compared with a simulated control group using the traditional economic optimal scheduling method.

[0058] In practice, the effectiveness of the method in this embodiment is demonstrated in the following typical scenarios: Scenario 1: Health Protection of Hydrogen Fuel Cells. After several days of continuous high-load operation, the health index of the hydrogen fuel cell was monitored and decreased from an initial 0.95 to 0.82. When a new round of optimization began, based on the lower health index, the strategy space of the hydrogen fuel cell was dynamically constrained, reducing the maximum allowable output power from 100% to 85% of the rated power. Simultaneously, the weight of the health penalty term in the comprehensive revenue function was increased. When solving for the Nash equilibrium point, even though the fuel cell's power generation cost was slightly lower than the cost of purchasing electricity from the grid, the optimization strategy ultimately chose to transfer part of the load to the grid interface and photovoltaic units, allowing only the fuel cell to bear the rated power output. In contrast, the traditional method in the control group would drive the fuel cell to operate at high load to pursue the lowest cost.

[0059] This embodiment of the method avoids the continuous exploitation of equipment in poor condition by making minor economic concessions, thus extending its service life. Figure 2 The changes in power output and equipment health index of the hydrogen fuel cell were compared between the method of this embodiment and the conventional method over an operating cycle. The method of this embodiment dynamically adjusts its power output after detecting a decline in equipment health, while the conventional method continues to operate at high load, thus clearly demonstrating the advantages of the method of this embodiment in protecting equipment health.

[0060] Scenario 2: Responding to Grid Frequency Disturbances. At 14:30 on April 12th, a brief drop in grid frequency was detected, falling below the normal threshold. This was automatically identified as a reliability-priority event. During the reconstruction of the game theory model, the weight of the reliability payoff sub-function was temporarily increased from the usual 0.3 to 0.6, while the economic payoff weight was reduced. Each participant iteratively solved the problem under the guidance of the new comprehensive payoff function. The final Nash equilibrium strategy showed that the output power of the hydrogen fuel cell and grid interface was increased to quickly support the grid frequency, while the frequency-insensitive heat pump load was appropriately reduced. The response helped the regional grid frequency stabilize within 2 minutes. This demonstrates that the method in this embodiment can flexibly adjust the optimization objective according to changes in the external environment, prioritizing the reliable and stable operation of the system and the grid at critical moments.

[0061] Scenario 3: Adaptive Correction of the Model. In the initial stage of operation, it was found that the actual output power of the heat pump unit was consistently lower than the target value of the control command, especially in the high-power output range. After the control command was issued, this deviation was recorded and a response status report was generated. Based on this report, the self-correction mechanism was triggered. Analysis indicated that the actual aging and loss rate of the heat pump exceeded the initial setting of the health assessment model. Therefore, the internal parameters of the model were automatically adjusted, increasing the loss accumulation coefficient under high-temperature operating conditions. After adjustment, the calculated health index of the heat pump was closer to reality. In subsequent optimized control, the target values ​​of the commands issued to the heat pump became more accurate, the deviation between predicted and actual output decreased, and the control accuracy and overall predictability of the system were improved. Figure 3 Each scatter point represents a control cycle, and the color depth reflects the magnitude of the deviation between the control command and the actual output. This visually demonstrates that in the initial stage of operation, the model exhibits deviations, but after self-correction, the actual output gradually approaches the command value, and the deviation significantly decreases. This verifies that the method in this embodiment possesses adaptive correction capabilities.

[0062] Example 2: As Figure 4 As shown, an optimized control system for interaction between an electro-hydrogen-thermal distributed energy supply system and the power grid is used to implement the method in Example 1, comprising: The data acquisition module is used to acquire real-time operating data of each energy unit and real-time parameters of the power grid, and generate operating datasets and power grid parameter sets. The health calculation module is used to extract time series data based on the running dataset, map the time series data to the cumulative damage of the equipment using a health assessment model, calculate the wear degree parameter, and generate a health index that is negatively correlated with the wear degree. The game modeling module is used to construct a multi-objective game model based on the health index and the power grid parameter set. It takes each energy unit and the power grid interface as independent game participants and constructs a comprehensive payoff function that includes economic efficiency, reliability and health penalty terms that are negatively correlated with the health index. The equilibrium solution module is used to drive each participant to independently update the policy according to the gradient of the comprehensive benefit function through a distributed iterative algorithm, and to determine whether the policy update amount is less than the convergence threshold, and to determine the Nash equilibrium point as the optimization policy. The instruction generation and correction module is used to encapsulate the optimization strategy into control instructions with execution timestamps, issue them for execution, monitor the actual response status and compare it with the target output to generate a report, and correct the cumulative loss weight in the health assessment model and the comprehensive payoff function parameters in the game model in real time based on the deviation.

Claims

1. An optimal control method for interaction between an electric-hydrogen-thermal distributed energy system and a power grid, characterized in that, The method comprises: acquiring real-time operation data of each energy unit and real-time parameters of the power grid to generate an operation data set and a power grid parameter set; extracting time series data based on the operation data set, mapping the time series data to device cumulative damage using a health assessment model, calculating a loss degree parameter, and generating a health degree index negatively correlated with the loss degree parameter; constructing a multi-objective game model according to the health degree index and the power grid parameter set, taking each energy unit and the power grid interface as an independent game participant, and constructing a comprehensive benefit function including economy, reliability, and a health penalty term negatively correlated with the health degree index; driving each participant to independently update the strategy according to the gradient of the comprehensive benefit function, and determining a Nash equilibrium point as the optimization strategy by judging whether the strategy update amount is less than a convergence threshold; packaging the optimization strategy into a control instruction with an execution timestamp, executing the control instruction, monitoring the actual response state, comparing the actual response state with the target output, generating a report, and real-time correcting the loss accumulation weight in the health assessment model and the comprehensive benefit function parameter in the multi-objective game model based on the deviation.

2. The method of claim 1, wherein the method further comprises: For the multi-objective game model, start a scheduling cycle timer, and at the beginning of each scheduling cycle, trigger the acquisition of real-time operation data and real-time parameters; use the latest acquired operation data set and power grid parameter set to update the health degree index and reconstruct the multi-objective game model.

3. The method of claim 1, wherein the method further comprises: Generating the operation data set and the power grid parameter set comprises: collecting the output power, efficiency and temperature parameters of each energy unit to form an energy operation data subset; collecting the frequency and load demand parameters of the power grid to form a power grid parameter subset; timestamp alignment and merging of the energy operation data subset and the power grid parameter subset to generate the operation data set and the power grid parameter set.

4. The method of claim 1, wherein the method further comprises: Calculating the loss degree parameter and generating the health degree index negatively correlated with the loss degree parameter comprises: extracting historical and current operation data of each energy unit from the operation data set to form time series data; using the health assessment model that maps the time series data to device cumulative damage to calculate the loss degree parameter of each energy unit; normalizing the loss degree parameter to generate a health degree index negatively correlated with the loss degree parameter.

5. The method of claim 1, wherein the method further comprises: Constructing a multi-objective game model comprises: determining a participant set including photovoltaic units, hydrogen fuel cell units, heat pump units and power grid interfaces; dynamically determining the selectable range of energy output based on the health degree index of each participant and the load demand in the power grid parameter set to form a strategy space; constructing a comprehensive benefit function including economy, reliability and health degree.

6. The method of claim 5, wherein the method further comprises: Constructing a comprehensive benefit function comprises: establishing an economy benefit sub-function for measuring fuel consumption and electricity purchase and sale cost; establishing a reliability benefit sub-function for measuring power grid frequency deviation and load satisfaction degree; establishing a health penalty term as a health degree benefit sub-function; weighting and summing the economy benefit sub-function, the reliability benefit sub-function and the health degree benefit sub-function to form the comprehensive benefit function.

7. The method of claim 1, wherein the method further comprises: Driving each participant to independently update the strategy according to the gradient of the comprehensive benefit function, and determining a Nash equilibrium point as the optimization strategy comprises: initializing the initial strategy of each participant based on the health degree index and the power grid parameter set; By means of a distributed iterative algorithm, each participant is driven to independently update his own strategy according to the gradient of the comprehensive benefit function, and it is judged whether the strategy update amount of all participants is less than a preset convergence threshold, if so, the current strategy combination is determined as a Nash equilibrium point, and the optimized strategy is output.

8. The method of claim 1, wherein the method further comprises: The optimized strategy is encapsulated as a regulation and control instruction with an execution timestamp, and the regulation and control instruction is executed, including: Analyzing the optimized strategy and extracting the target output amount corresponding to each energy unit; Encapsulating the target output amount and the execution timestamp to generate a regulation and control instruction; Sending the regulation and control instruction to the execution controller of each energy unit through a communication interface.

9. The method of claim 1, wherein the method further comprises: Based on the deviation, the loss accumulation weight in the health assessment model and the comprehensive benefit function parameter in the game model are real-time corrected, including: After issuing the regulation and control instruction, the actual response state of each energy unit is monitored, and the actual response state is compared with the target output amount in the regulation and control instruction to generate a response state report; Based on the deviation presented in the response state report, the internal parameters in the health assessment model and the multi-objective game model are adjusted; Using the adjusted model parameters for the next round of optimization and control, the accuracy and adaptive performance of the model are improved.

10. An optimal regulation system for interaction between an electro-hydro-thermal distributed energy supply system and a power grid, for implementing the method of any one of claims 1-9, characterized in that, Including: The data acquisition module is used to acquire real-time operation data of each energy unit and real-time parameters of the power grid, generate an operation data set and a power grid parameter set; The health calculation module is used to extract time series data based on the operation data set, map the time series data to the cumulative damage of the equipment using the health assessment model, calculate the loss degree parameter, and generate a health degree index negatively correlated with the loss degree; The game modeling module is used to construct a multi-objective game model according to the health degree index and the power grid parameter set, taking each energy unit and power grid interface as an independent game participant, and constructing a comprehensive benefit function including economy, reliability and health penalty term negatively correlated with the health degree index; The equilibrium solving module is used to drive each participant to independently update the strategy according to the gradient of the comprehensive benefit function by means of a distributed iterative algorithm, and to judge whether the strategy update amount is less than the convergence threshold, and to determine the Nash equilibrium point as the optimized strategy; The instruction generation and correction module is used to encapsulate the optimized strategy as a regulation and control instruction with an execution timestamp, execute the regulation and control instruction, monitor the actual response state and compare it with the target output amount to generate a report, and real-time correct the loss accumulation weight in the health assessment model and the comprehensive benefit function parameter in the game model based on the deviation.

Images