A source network load storage capacity planning method and system based on multi-agent game

Abstract

The application relates to a source network load storage capacity planning method and system based on multi-agent game, which comprises the following steps: constructing a multi-agent game behavior data set, analyzing strategy dynamic interaction by using a game theory model, obtaining strategy preference and interaction mode of each agent, constructing a source network load storage collaborative planning model, determining an initial capacity configuration scheme of each agent, extracting the benefit distribution proportion of each agent from the initial capacity configuration scheme, adjusting the capacity configuration by using a multi-objective optimization algorithm, obtaining an optimized capacity planning scheme, analyzing the influence of peak demand fluctuation and power transmission capacity limitation on the scheme by using a Monte Carlo simulation method, obtaining a dynamic adaptability index of the scheme, adjusting the game theory model parameters, obtaining updated strategy preference and interaction mode, adjusting the capacity configuration through iterative calculation, and generating a new capacity planning scheme. Compared with the prior art, the application has the advantages of comprehensively considering multi-agent strategy dynamic interaction and improving the feasibility of a planning scheme.

Description

Technical Field

[0001] This invention belongs to the field of smart grid technology and relates to an energy system capacity planning method, particularly to a source-grid-load-storage capacity planning method and system based on multi-agent game theory. Background Technology

[0002] Capacity planning for energy systems is a key area for achieving low-carbon transformation and efficient operation, directly impacting the stability and economy of energy supply. In the context of integrated power generation, grid, load, and energy storage, coordinated planning of power sources, grids, loads, and energy storage is crucial, requiring a balance of interests among multiple parties to ensure maximum overall system benefits. Existing technologies have disclosed various methods for coordinated planning of power generation, grid, load, and energy storage. For example, Chinese patent application CN120433182A discloses a method for coordinated planning of power generation, grid, load, and energy storage, including: obtaining the structural, operational, and economic parameters of the power system; dividing extreme scenarios and establishing a frequency response model; setting frequency security constraints based on the extreme scenarios; constructing a three-layer optimization planning model, including an objective function and constraints, with the third layer including frequency security constraints; and solving the model to determine the planning scheme. However, existing methods have significant shortcomings in handling dynamic interactions and complex decision-making among multiple stakeholders, making it difficult to comprehensively capture the impact of changes in the strategies of each stakeholder on the system, especially when facing uncertainties, lacking effective dynamic analysis tools. This often results in planning schemes that cannot adapt to complex scenarios in actual operation, affecting the reliability and economy of the system.

[0003] The core challenge lies in accurately characterizing the strategic interactions among multiple stakeholders and, based on this, assessing the feasibility of various planning schemes. First, the decisions of each stakeholder in a power-grid-load-storage system influence each other. For example, power producers may want to expand capacity to increase revenue, while load producers are more focused on cost control. This conflict of interest makes it difficult to reach a consensus on planning schemes. Second, the complexity of this dynamic game further exacerbates the difficulty of assessing the feasibility of schemes. For instance, in a regional energy plan, power companies may prioritize low-cost thermal power expansion, while energy storage companies tend to increase storage capacity to cope with peak demand fluctuations. However, grid companies may oppose both schemes due to transmission capacity limitations. Under this multi-party game, the lack of a unified analytical framework to comprehensively assess the feasibility of different capacity configurations leads to inefficient planning processes and suboptimal results.

[0004] Therefore, how to construct an analytical framework that can comprehensively consider the dynamic interaction of multiple agents' strategies and systematically evaluate the feasibility of various planning schemes has become a key issue in the research of source-grid-load-storage capacity game planning methods. Summary of the Invention

[0005] The purpose of this invention is to provide a source-grid-load-storage capacity planning method and system based on multi-agent game theory, which comprehensively considers the dynamic interaction of multi-agent strategies and has high feasibility of planning schemes.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] This invention provides a source-grid-load-storage capacity planning method based on multi-agent game theory, comprising the following steps:

[0008] A multi-agent game behavior dataset is constructed, and a game theory model is used to analyze the dynamic interaction of strategies based on the multi-agent game behavior dataset to obtain the strategy preferences and interaction patterns of each agent.

[0009] Based on the strategy preferences and interaction patterns of each entity, a source-grid-load-storage collaborative planning model is constructed. Solving this source-grid-load-storage collaborative planning model determines the initial capacity configuration scheme of each entity.

[0010] The benefit distribution ratio of each entity is extracted from the initial capacity configuration scheme, and the capacity configuration is adjusted using a multi-objective optimization algorithm to obtain an optimized capacity planning scheme.

[0011] Based on the optimized capacity planning scheme, the Monte Carlo simulation method is used to analyze the impact of peak demand fluctuations and transmission capacity limitations on the scheme, and the dynamic adaptability index of the scheme is obtained.

[0012] The game theory model parameters are adjusted based on the dynamic adaptability index to obtain updated strategy preferences and interaction patterns. Based on the updated strategy preferences and interaction patterns, the capacity configuration is adjusted through iterative calculation to generate a new capacity planning scheme.

[0013] Furthermore, the multi-agent game behavior dataset is constructed based on historical decision-making data and real-time operational data of the source-grid-load-storage entities.

[0014] Furthermore, the strategy preferences and interaction patterns of each subject are obtained through the following steps:

[0015] The multi-agent game behavior dataset is cleaned to generate a structured behavior dataset;

[0016] Based on the structured behavior dataset, a clustering algorithm is used to divide the subject behavior patterns and determine the behavior characteristics of each subject;

[0017] Based on the behavioral characteristics of each subject, a game theory model is used to analyze multi-subject games, obtain the dynamics of strategy interaction, and use time series analysis to extract the changing trend of subject strategies over time to determine strategy preferences. If the deviation between the strategy preferences and historical behavior patterns exceeds a preset behavior threshold, the game theory model parameters are adjusted through regression analysis to obtain optimized strategy preferences.

[0018] Based on the behavioral characteristics of each subject, the association rule mining method is used to analyze the interaction patterns between subjects and predict the future interaction strategy choices between subjects.

[0019] The specific selection of the prediction strategy for future interactions between the subjects is as follows:

[0020] The stability and regularity of interaction patterns are extracted, and Markov chains are used to predict the future strategy choices of the subjects, thus obtaining the prediction results.

[0021] Furthermore, the constructed source-grid-load-storage collaborative planning model is as follows:

[0022]

[0023]

[0024] In the formula, This represents the first cost coefficient. The second cost coefficient, The third cost coefficient, Indicates the installed capacity of wind power. Indicates photovoltaic installation capacity. Indicates energy storage capacity. Indicates the maximum capacity of the transmission line. Indicates peak load demand. This indicates the upper limit of wind power installation capacity. Indicates the upper limit of photovoltaic installation capacity. This indicates the upper limit of the installed energy storage capacity.

[0025] Furthermore, the process of adjusting the capacity configuration using a multi-objective optimization algorithm to obtain the optimized capacity planning scheme is as follows:

[0026] A multi-objective optimization model is constructed to obtain the interest distribution set for conflict balance. If the distribution ratio of any subject in the interest distribution set is lower than a preset threshold, the distribution ratio is adjusted by linear interpolation to obtain the adjusted interest distribution set.

[0027] Based on the adjusted set of benefits, a genetic algorithm is used to optimize the capacity configuration parameters according to the initial capacity configuration scheme, resulting in a preliminary optimized capacity planning scheme.

[0028] Capacity configuration features are extracted from the preliminary optimized capacity planning scheme, and cluster analysis is used to classify the capacity configuration patterns among the subjects, resulting in a set of classified configuration patterns.

[0029] Based on the set of classified configuration patterns, a time series forecasting method is used to analyze the dynamic trend of capacity configuration and obtain the predicted capacity configuration results.

[0030] If the deviation between the predicted capacity configuration result and the initial capacity configuration scheme exceeds a preset threshold, the capacity configuration parameters are adjusted to obtain the final capacity planning scheme, which is then used as the optimized capacity planning scheme.

[0031] Furthermore, the multi-objective optimization model that yields the interest allocation set for conflict equilibrium is as follows:

[0032]

[0033]

[0034]

[0035]

[0036] In the formula, Represents the storage-side payment function. Represents the load-side payment function. Indicates the source-side payment function. This indicates market clearing constraints during peak hours. This indicates market clearing constraints during periods of low activity. This represents a linear inverse demand function for prices. Indicates the installed capacity of wind power. Indicates photovoltaic installation capacity. Indicates energy storage capacity. , , These represent the first policy variable, the second policy variable, and the third policy variable, respectively. Indicates peak electricity price, Indicates off-peak electricity price. Indicates peak demand. Indicating low demand, This indicates the unit charging cost of energy storage. Indicates the maximum transferable load. , , , These represent the first, second, third, and fourth inverse demand functions, respectively. Indicates the maximum capacity of the transmission line. Represents the payment function. Other subject equilibrium strategies are represented by W, PV, B, and L, which represent wind power aggregate, photovoltaic aggregate, energy storage aggregate, and load aggregate, respectively.

[0037] Furthermore, the dynamic adaptability index of the proposed scheme is obtained through the following steps:

[0038] Uncertainty factor data are extracted from the optimized capacity planning scheme to obtain a standardized uncertainty dataset;

[0039] Based on the standardized uncertainty dataset, multiple sets of peak demand fluctuation scenarios are generated using the Monte Carlo simulation method to obtain a demand distribution feature set.

[0040] Peak demand fluctuation parameters are extracted from the demand distribution feature set, and combined with transmission capacity constraints, the fluctuation impact weight is calculated to obtain the fluctuation impact assessment results.

[0041] Based on the assessment results of the fluctuation impact, dynamic adaptive features are extracted to obtain dynamic adaptive indicators.

[0042] Furthermore, the method also includes:

[0043] Determine whether the new capacity planning scheme meets the overall system benefit requirements. If the overall system benefit does not reach the preset threshold, obtain real-time feedback from the source-grid-load-storage operation data, update the Monte Carlo simulation parameters, re-evaluate the dynamic adaptability index of the scheme, and obtain the final capacity planning scheme.

[0044] This invention also provides a source-grid-load-storage capacity planning system based on multi-agent game theory, comprising:

[0045] The data acquisition module is used to construct a multi-agent game behavior dataset. Based on the multi-agent game behavior dataset, a game theory model is used to analyze the dynamic interaction of strategies to obtain the strategy preferences and interaction patterns of each agent.

[0046] The game analysis module is used to construct a source-grid-load-storage collaborative planning model based on the strategy preferences and interaction patterns of each subject, and solve the source-grid-load-storage collaborative planning model to determine the initial capacity configuration scheme of each subject.

[0047] The collaborative planning module is used to extract the benefit distribution ratio of each subject from the initial capacity configuration scheme, and to adjust the capacity configuration using a multi-objective optimization algorithm to obtain an optimized capacity planning scheme.

[0048] The benefit optimization module is used to analyze the impact of peak demand fluctuations and transmission capacity limitations on the optimized capacity planning scheme using the Monte Carlo simulation method, and obtain the dynamic adaptability index of the scheme.

[0049] The dynamic evaluation module is used to adjust the parameters of the game theory model based on the dynamic adaptability index to obtain the updated strategy preferences and interaction patterns. Based on the updated strategy preferences and interaction patterns, the capacity configuration is adjusted through iterative calculation to generate a new capacity planning scheme.

[0050] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, characterized in that the processor executes the computer program to implement the steps of the method described above.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] This invention addresses the challenge of collaborative optimization among power system stakeholders (source, grid, load, and storage) in capacity planning due to conflicts of interest and uncertainties. It constructs a multi-stakeholder game behavior dataset, analyzes the strategy preferences and interaction patterns of each stakeholder using game theory models, incorporates transmission capacity constraints and peak demand fluctuations, and employs a Nash equilibrium algorithm to generate an initial capacity configuration scheme. A multi-objective optimization algorithm is then used to balance conflicting interests and adjust the capacity configuration. Monte Carlo simulation is further utilized to assess the impact of uncertainties and calculate dynamic adaptability indicators. If the indicators fail to reach a threshold, feedback data is used to dynamically adjust the game model parameters and re-optimize the capacity planning until the overall system efficiency requirements are met. This invention verifies the scheme's effectiveness through real-time data monitoring and dynamic analysis tools, comprehensively considering the dynamic interaction of multi-stakeholder strategies. Ultimately, it achieves high efficiency and adaptability in source-grid-load-storage collaborative planning, improves the reliability of the planning scheme, and thus enhances the overall efficiency and stability of the power system. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the method of the present invention.

[0054] Figure 2 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0055] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0056] Example 1

[0057] This embodiment provides a source-grid-load-storage capacity planning method based on multi-agent game theory, comprising the following steps: constructing a multi-agent game behavior dataset; analyzing the dynamic interaction of strategies using a game theory model based on the multi-agent game behavior dataset to obtain the strategy preferences and interaction patterns of each agent; constructing a source-grid-load-storage collaborative planning model based on the strategy preferences and interaction patterns of each agent; solving the source-grid-load-storage collaborative planning model to determine the initial capacity configuration scheme for each agent; extracting the benefit distribution ratio of each agent from the initial capacity configuration scheme; adjusting the capacity configuration using a multi-objective optimization algorithm to obtain an optimized capacity planning scheme; analyzing the impact of peak demand fluctuations and transmission capacity limitations on the scheme using Monte Carlo simulation based on the optimized capacity planning scheme to obtain the dynamic adaptability index of the scheme; adjusting the game theory model parameters based on the dynamic adaptability index to obtain updated strategy preferences and interaction patterns; and adjusting the capacity configuration through iterative calculation based on the updated strategy preferences and interaction patterns to generate a new capacity planning scheme. This method verifies the implementation effect of the scheme through real-time data monitoring and dynamic analysis tools, ultimately achieving high efficiency and adaptability of source-grid-load-storage collaborative planning, and improving the overall efficiency and stability of the power system.

[0058] like Figure 1 As shown, the specific steps of the above method are described below:

[0059] S101. Obtain historical decision-making data and real-time operation data from the source, grid, load, and storage entities to construct a multi-entity game behavior dataset. Use game theory models to analyze the dynamic interaction of strategies and obtain the strategy preferences and interaction patterns of each entity.

[0060] Historical decision-making data and real-time operational data are obtained from the source-grid-load-storage entities. Data cleaning techniques are used to remove missing and outlier values, resulting in a structured behavioral dataset. Based on this dataset, clustering algorithms are employed to segment the entities' behavioral patterns and determine the behavioral characteristics of each entity. If the entity's behavioral characteristics meet preset game participation conditions, a Nash equilibrium model is used to analyze the multi-entity game and obtain the dynamics of strategy interactions. Based on these dynamics, time series analysis is used to extract the trends in entity strategies over time and determine strategy preferences. If the deviation between the strategy preferences and the behavioral patterns in the historical decision-making data exceeds a preset threshold, regression analysis is used to adjust the game theory model parameters to obtain optimized strategy preferences. Based on the optimized strategy preferences, association rule mining techniques are used to analyze the interaction patterns between entities and determine the stability and regularity of these patterns. Based on the stability of the interaction patterns, Markov chains are used to predict the future strategy choices of the entities, yielding prediction results.

[0061] Furthermore, as a specific implementation of this embodiment, historical decision-making data and real-time operational data are obtained from the power grid, load, and energy storage entities, involving the power grid, load, and energy storage entities. Data includes dispatch instructions recorded by the power grid company, user electricity consumption, and energy storage device charging and discharging records. Data cleaning techniques are used to handle missing and outlier values. If a user's data for a specific day is missing, it can be imputed using the average of adjacent days. Outliers, such as sudden power fluctuations of an energy storage device exceeding the physical range, are removed and replaced with trend values, resulting in a structured behavioral dataset, such as a table containing fields like timestamps, electricity consumption, and dispatch instructions.

[0062] In this embodiment, the K-means clustering algorithm is used to classify the main behavioral patterns. Taking the electricity load data of a certain area as an example, based on the daily electricity consumption curve, users are divided into three behavioral patterns: peak-type, off-peak-type, and off-peak-type. Peak-type users consume electricity in concentrated amounts during the day, characterized by high peak loads; off-peak-type users have evenly distributed electricity consumption; and off-peak-type users consume more electricity at night.

[0063] Furthermore, as a specific implementation of this embodiment, if the behavioral characteristics of the subjects meet the conditions for game participation, such as the strategy choices of peak-hour users and energy storage entities under fluctuating electricity prices, a Nash equilibrium model is used to analyze the multi-subject game. When users choose to reduce electricity consumption during periods of high electricity prices, energy storage entities choose to charge during periods of low electricity prices. Nash equilibrium analysis shows that when the strategies of both parties reach equilibrium, users reduce costs, and energy storage entities optimize profits. Based on the dynamics of strategy interaction, time series analysis is used to extract the trend of changes in subject strategies. Assuming 30 consecutive days of electricity consumption strategy data, analysis using the ARIMA model reveals that the trend of reducing electricity consumption is enhanced when electricity prices are high, and the strategy preference tends towards saving. If this preference deviates from historical behavioral patterns by more than a threshold (e.g., 20%), the game model parameters are adjusted through regression analysis, such as adjusting the user's sensitivity coefficient to electricity prices from 0.5 to 0.7, to obtain the optimized strategy preference and improve the model's predictive accuracy.

[0064] Furthermore, as a specific implementation of this embodiment, the Apriori algorithm is used to mine interaction patterns between entities based on optimized strategy preferences. Analysis reveals that the charging-consumption interaction pattern between peak-period users and energy storage entities is stable during low-electricity-price periods, occurring in 80% of cases. The stability of this interaction pattern is assessed using support and confidence scores, demonstrating strong regularity and potential for predicting market behavior. Based on the stability of the interaction pattern, a Markov chain is used to predict the future strategy choices of entities. Taking energy storage entities as an example, the probability of charging during low-electricity-price periods and discharging during high-electricity-price periods is analyzed, predicting a 70% probability of charging within the next 7 days. The prediction results provide a reference for grid dispatch, reducing peak load pressure and improving system stability.

[0065] S102. Based on strategy preferences and interaction patterns, construct a source-grid-load-storage collaborative planning model, incorporate transmission capacity constraints and peak demand fluctuation constraints, and use the Nash equilibrium algorithm to determine the initial capacity configuration scheme for each entity.

[0066] Furthermore, as a specific implementation of this embodiment, when obtaining strategy preference and interaction pattern data from the source-grid-load-storage entity, data can be collected through historical transaction records and real-time monitoring systems. The source side can obtain the power plant's output plan, the grid side collects grid dispatch data, the load side extracts user electricity consumption curves, and the storage side records energy storage charging and discharging strategies. When processing missing and outlier values ​​using data cleaning techniques, the mean interpolation method can be used to fill in missing electricity consumption data. For example, if a user's electricity consumption is missing for a certain period, it can be filled in using the mean of the preceding and following periods. Outliers are eliminated through standard deviation detection; for example, if a power plant's output suddenly increases by more than three times the standard deviation, it is considered an anomaly.

[0067] Furthermore, as a specific implementation of this embodiment, a linear programming method is used to optimize the initial capacity allocation in conjunction with transmission capacity constraints. The objective function is set as minimizing the total cost, and the constraints include the upper limit of transmission line capacity and the output range of each main entity. The maximum transmission capacity of a certain regional power grid is 500 MW, with wind power and photovoltaic outputs of 200 MW and 150 MW respectively, and energy storage capacity of 100 MW. Linear programming can solve for the output ratio of each main entity to obtain a preliminary capacity configuration scheme. Peak demand fluctuation constraints can be verified through historical peak load data. For example, if the peak load on a certain day is 450 MW, the scheme must meet this demand.

[0068] The preliminary capacity allocation scheme allocation model, namely the source-grid-load-storage coordinated planning model, is as follows:

[0069]

[0070]

[0071] In the formula, This represents the first cost coefficient. The second cost coefficient, The third cost coefficient, Indicates the installed capacity of wind power. Indicates photovoltaic installation capacity. Indicates energy storage capacity. Indicates the maximum capacity of the transmission line. Indicates peak load demand. This indicates the upper limit of wind power installation capacity. Indicates the upper limit of photovoltaic installation capacity. This indicates the upper limit of the installed energy storage capacity.

[0072] S103. Extract the interest distribution ratio of each subject from the initial capacity configuration scheme. For the interest conflict balancing problem, adopt a multi-objective optimization algorithm to adjust the capacity configuration and obtain the optimized capacity planning scheme.

[0073] The initial capacity allocation scheme yields the benefit distribution ratios of each entity. The Pareto optimal solution method is used to analyze the conflicts of interest among the entities, resulting in a benefit distribution set where conflicts are balanced. If any entity's distribution ratio in this set is lower than a preset threshold, linear interpolation is used to adjust the ratio, resulting in an adjusted benefit distribution scheme. Based on the adjusted scheme, a genetic algorithm is used to optimize the capacity allocation parameters, yielding a preliminary optimized capacity planning scheme. Capacity allocation features are extracted from the preliminary optimized scheme, and cluster analysis is used to classify capacity allocation patterns among the entities, resulting in a set of categorized configuration patterns. Based on these patterns, time series forecasting is used to analyze the dynamic trends of capacity allocation, obtaining predicted capacity allocation results. If the deviation between the predicted and initial capacity allocation schemes exceeds a preset threshold, the capacity allocation parameters are iteratively optimized to obtain the final capacity planning scheme.

[0074] Specifically, in the coordinated planning of power generation, grid, load, and storage, when obtaining the profit distribution ratio of each entity from the initial capacity configuration scheme, the proportion of each entity in the total revenue can be calculated by analyzing the power generation, energy storage, and electricity consumption data of each entity. Assuming a regional power grid includes renewable energy power generation companies, energy storage operators, and industrial users, the initial capacity configuration scheme shows that renewable energy power generation companies provide 60% of the electricity, energy storage operators undertake 30% of the peak-shaving task, and industrial users consume all the electricity. Through the revenue model, the profit distribution ratio for renewable energy power generation companies is 0.55, for energy storage operators it is 0.35, and for industrial users it is 0.1.

[0075] Furthermore, as a specific implementation of this embodiment, when processing the allocation ratio data using data standardization technology, the min-max standardization method is used to map the profit allocation ratio to the range of 0 to 1. Assuming the ratio for new energy power generation enterprises is 0.55, for energy storage operators 0.35, and for industrial users 0.1, the normalized values ​​after standardization may be 0.8, 0.5, and 0.1.

[0076] When analyzing conflicts of interest using the Pareto optimal solution method, a multi-objective optimization model is constructed to identify the set of interest allocations. New energy power generation companies want to increase the proportion to 0.6, while energy storage operators want to maintain at least 0.4. Through Pareto frontier analysis, a solution set {0.58, 0.38, 0.04} is obtained, ensuring maximum overall efficiency and minimum conflict between stakeholders. If the proportion of industrial users (0.04) in the interest allocation set is lower than the preset threshold of 0.1, it is adjusted using linear interpolation. Assuming the adjustment target is 0.12, a new allocation {0.56, 0.32, 0.12} is obtained by linearly scaling according to the proportions of other stakeholders.

[0077] The multi-objective optimization model for obtaining the interest distribution set in the conflict balance is as follows:

[0078]

[0079]

[0080]

[0081]

[0082] In the formula, Represents the storage-side payment function. Represents the load-side payment function. Indicates the source-side payment function. This indicates market clearing constraints during peak hours. This indicates market clearing constraints during periods of low activity. This represents a linear inverse demand function for prices. Indicates the installed capacity of wind power. Indicates photovoltaic installation capacity. Indicates energy storage capacity. , , These represent the first policy variable, the second policy variable, and the third policy variable, respectively. Indicates peak electricity price, Indicates off-peak electricity price. Indicates peak demand. Indicating low demand, This indicates the unit charging cost of energy storage. Indicates the maximum transferable load. , , , These represent the first, second, third, and fourth inverse demand functions, respectively. Indicates the maximum capacity of the transmission line. Represents the payment function. Other subject equilibrium strategies are represented by W, PV, B, and L, which represent wind power aggregate, photovoltaic aggregate, energy storage aggregate, and load aggregate, respectively.

[0083] Furthermore, as a specific implementation of this embodiment, when using a genetic algorithm to optimize capacity configuration parameters, an initial population is set up with multiple capacity configuration schemes, and the scheme with high fitness is selected iteratively. In the initial scheme, the new energy power generation capacity is 1000MW and the energy storage capacity is 300MW. After optimization by the genetic algorithm, schemes of 1050MW and 320MW may be obtained.

[0084] Furthermore, as a specific implementation of this embodiment, if the deviation between the prediction result and the initial plan exceeds a threshold, such as a 15% deviation between the predicted capacity of 1150MW and the initial 1000MW, the parameters are adjusted through iterative optimization. For example, the new energy power generation capacity is gradually increased to 1120MW, the benefit distribution is recalculated, and the final plan is obtained.

[0085] S104. Based on the optimized capacity planning scheme, obtain data on uncertain factors, use the Monte Carlo simulation method to analyze the impact of peak demand fluctuations and transmission capacity limitations on the scheme, and obtain the dynamic adaptability index of the scheme.

[0086] Uncertainty factor data are extracted from the optimized capacity planning scheme to obtain a standardized uncertainty dataset. Based on the standardized uncertainty dataset, multiple peak demand fluctuation scenarios are generated using the Monte Carlo simulation method to obtain a demand distribution feature set. Peak demand fluctuation parameters are extracted from the demand distribution feature set, and combined with transmission capacity constraints, fluctuation impact weights are calculated to obtain fluctuation impact assessment results. Dynamic adaptive features are extracted based on the fluctuation impact assessment results to obtain dynamic adaptive indicators.

[0087] Furthermore, as a specific implementation of this embodiment, when extracting uncertainty factor data from the optimized capacity planning scheme, attention should be paid to factors such as demand fluctuations, equipment failure rates, and changes in the external environment. For example, in a power system capacity planning scheme, uncertainties include seasonal fluctuations in user electricity demand, the probability of failure due to equipment aging, and the impact of policy changes on the access of new energy sources.

[0088] When generating peak demand fluctuation scenarios using the Monte Carlo simulation method, 1000 random scenarios can be generated based on historical electricity consumption data, assuming that the demand follows a normal distribution. In one implementation, the average peak demand in summer is 5000 MW, with a standard deviation of 500 MW. Simulation results show that demand exceeds 6000 MW in 10% of the scenarios. These scenarios form a demand distribution feature set, reflecting the probability distribution characteristics of peak demand. After extracting fluctuation parameters from the demand distribution feature set, and combining them with transmission capacity limitations, such as a maximum line transmission capacity of 5500 MW, the impact weight of fluctuations is calculated through linear regression analysis.

[0089] Furthermore, as a specific implementation of this embodiment, dynamic adaptive features, such as reserve capacity ratio and demand response speed, are extracted from the initially adjusted scheme, and their periodic changes are analyzed using time series decomposition technology. For example, the decomposition shows that the reserve capacity ratio fluctuates sinusoidally with the seasons, reaching 20% ​​during the summer peak and dropping to 10% during the winter trough. Based on this, cluster analysis can classify the adaptive patterns into two categories: high adaptability and low adaptability, with higher adaptability indicating more sufficient reserve capacity.

[0090] S105. If the dynamic adaptability index is lower than the preset threshold, then obtain feedback data from the dynamic analysis tool, adjust the game theory model parameters, recalculate the dynamic interaction of the strategy, and obtain the updated strategy preferences and interaction patterns.

[0091] If the dynamic adaptability index is lower than a preset threshold, feedback data is extracted from the dynamic analysis tool and processed using data standardization techniques to obtain a standardized feedback dataset. Key feature parameters are extracted from the standardized feedback dataset, and principal component analysis is used for dimensionality reduction to obtain a dimensionality-reduced feature set. Game theory model parameters are adjusted based on the dimensionality-reduced feature set, and the strategy interaction relationships are updated through Nash equilibrium calculation to obtain a preliminary strategy preference set. Interaction weight features are extracted from the preliminary strategy preference set, and cluster analysis is used to classify strategy interaction patterns to obtain a categorized interaction pattern set. If any pattern in the categorized interaction pattern set deviates from a preset stability threshold, the game theory model parameters are optimized using gradient descent to obtain an optimized strategy preference set. Dynamic interaction features are extracted from the optimized strategy preference set, and time series analysis is used to decompose the trend of interaction pattern changes to obtain a dynamic interaction trend set. Based on the dynamic interaction trend set, capacity planning parameters are adjusted through iterative calculation to obtain the final adaptive optimization scheme.

[0092] Furthermore, as a specific implementation of this embodiment, when extracting feedback data from the dynamic analysis tool, power grid operation status data is obtained through a real-time monitoring system.

[0093] Furthermore, as a specific implementation of this embodiment, when extracting dynamic interaction features from the optimization strategy preference set, attention is paid to the load response speed and system recovery time after strategy adjustment. For example, after optimization, the load response time of a power grid is shortened from 10 minutes to 5 minutes. When using time series analysis to decompose the interaction pattern change trend, the trend can be divided into long-term growth and short-term fluctuations. The long-term trend of a certain regional power grid shows that the load increases by 3% annually, while short-term fluctuations are affected by seasonal demand, with summer peaks being 20% ​​higher. Based on the dynamic interaction trend set, capacity planning parameters are adjusted through iterative calculation, such as increasing substation capacity by 100 MW, ultimately forming an adaptive optimization scheme to ensure stable system operation under peak load.

[0094] S106. Based on the updated strategy preferences and interaction patterns, rerun the multi-objective optimization algorithm, adjust the capacity configuration, generate a new capacity planning scheme, and determine whether it meets the overall system benefit requirements.

[0095] If any pattern in the classification and allocation pattern set deviates from a preset stability threshold, pattern feature data is obtained from the classification and allocation pattern set. This data is then processed using feature standardization techniques to obtain a standardized feature dataset. Key allocation features are extracted from the standardized feature dataset, and feature priorities are classified using a support vector machine method to obtain a priority feature set. Capacity configuration parameters are adjusted based on the priority feature set, and iterative calculation methods are used to optimize resource allocation ratios to obtain an optimized allocation scheme. Dynamic allocation features are extracted from the optimized allocation scheme, and time series decomposition methods are used to analyze allocation change trends to obtain an allocation trend set. If the allocation trend set shows a change trend exceeding a preset fluctuation threshold, abnormal allocation data is obtained from the allocation trend set. Anomaly detection techniques are used to process the abnormal allocation data to obtain an anomaly correction dataset. Capacity planning parameters are adjusted based on the anomaly correction dataset, and resource allocation is optimized using a balanced allocation algorithm to obtain a final capacity planning scheme. Benefit evaluation features are extracted from the final capacity planning scheme, and statistical analysis methods are used to determine the overall system benefit to obtain the benefit evaluation result.

[0096] Furthermore, as a specific implementation of this embodiment, the stability assessment of the classification allocation mode set is a core component of the dynamic resource allocation system. Suppose a company's data center needs to optimize server resource allocation. The classification allocation mode set includes multiple allocation modes, such as high-load mode and low-load mode. If a mode deviates from a preset stability threshold, for example, if the CPU utilization of a server consistently exceeds 80% under high-load mode, then mode characteristic data, such as request response time and memory usage, needs to be extracted from that mode.

[0097] Furthermore, as a specific implementation of this embodiment, after the optimized allocation scheme is generated, dynamic allocation features, such as the trend of resource utilization over time, can be extracted. Data on the utilization of a server over the past 24 hours shows peak fluctuations every 4 hours. By analyzing these fluctuations, if the trend exceeds a preset fluctuation threshold, for example, if the utilization fluctuation exceeds 15%, then abnormal allocation data needs to be extracted. Anomaly detection techniques, such as those based on isolated forests, can identify servers with abnormally high loads within a certain time period. For example, if a server's utilization suddenly surges to 95% at 2 AM, an anomaly correction dataset is generated, and the task allocation ratio of that server is adjusted, for example, by reducing the task load by 10%. After adjusting the capacity planning parameters based on the anomaly correction dataset, resource allocation can be optimized using a balanced allocation algorithm.

[0098] High-load tasks are redistributed to backup servers to ensure overall load balancing. The final capacity planning scheme may control the average utilization of the server cluster within the range of 60%-70%. Benefit evaluation characteristics extracted from this scheme include a reduction in system response time to 200ms and a 10% increase in user satisfaction.

[0099] S107. If the overall system benefit does not reach the preset threshold, real-time feedback is obtained from the source-grid-load-storage operation data to update the Monte Carlo simulation parameters, re-evaluate the dynamic adaptability index of the scheme, and obtain the final capacity planning scheme.

[0100] If the adaptability indicators of the optimized capacity planning scheme do not reach the preset threshold, real-time dynamic data is obtained from the source-grid-load-storage operation data. Data standardization techniques are used to process the dynamic data, resulting in a standardized dynamic dataset. Key operational features are extracted from the standardized dynamic dataset, and principal component analysis is used to reduce the dimensionality of these features, resulting in a dimensionality-reduced feature set. Monte Carlo simulation parameters are updated based on the dimensionality-reduced feature set, and Monte Carlo simulation is used to generate multi-scenario operational data, resulting in a scenario dataset. If the scenario dataset shows operational fluctuations exceeding the preset fluctuation threshold, abnormal operational data is obtained from the scenario dataset, and cluster analysis is used to process the abnormal data, resulting in an anomaly correction dataset. Dynamic planning parameters are adjusted based on the anomaly correction dataset, and dynamic planning is used to optimize resource allocation ratios, resulting in an optimized allocation scheme. Benefit evaluation features are extracted from the optimized allocation scheme, and statistical analysis methods are used to determine the overall system benefit, resulting in a benefit evaluation result. If the benefit evaluation result does not reach the preset benefit threshold, deviation data is obtained from the benefit evaluation result, and iterative optimization techniques are used to adjust the capacity planning parameters, resulting in the final capacity planning scheme.

[0101] S108. Based on the final capacity planning scheme, generate capacity configuration instructions for each entity, and use dynamic analysis tools to monitor real-time operating status, judge the effectiveness of the scheme, and obtain verification results of the overall system benefits.

[0102] Real-time dynamic data is obtained from the source-grid-load-storage operation data. Data standardization techniques are used to process this data, resulting in a standardized dynamic dataset. Key operational features are extracted from this dataset, and principal component analysis (PCA) is employed to reduce their dimensionality, yielding a dimensionality-reduced feature set. Dynamic analysis parameters are updated based on this feature set, and a real-time operational status dataset is generated using dynamic analysis tools. If the operational status dataset shows fluctuations exceeding a preset fluctuation threshold, abnormal operational data is extracted and processed using cluster analysis to obtain an anomaly correction dataset. Capacity configuration parameters are adjusted based on this correction dataset, and dynamic programming is used to optimize the capacity configuration ratios of each entity, resulting in optimized configuration instructions. Benefit verification features are extracted from these instructions and processed using statistical analysis to obtain benefit verification results. If the benefit verification results do not meet the preset benefit threshold, deviation data is obtained from these results, and iterative optimization techniques are used to adjust the capacity configuration parameters, resulting in the final optimized configuration instructions.

[0103] If the above methods are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0104] Example 2

[0105] like Figure 2 As shown, this embodiment provides a source-grid-load-storage capacity planning system based on multi-agent game theory, used to implement the source-grid-load-storage capacity planning method based on multi-agent game theory disclosed in Embodiment 1. The system includes:

[0106] Data acquisition module 1 is used to acquire historical decision-making data and real-time operation data from various entities in the source, grid, load and storage system, construct a multi-entity game behavior dataset, and use game theory models to analyze the dynamic interaction of strategies to obtain the strategy preferences and interaction patterns of each entity.

[0107] Game analysis module 2 is used to construct a source-grid-load-storage collaborative planning model based on the strategy preferences and interaction patterns of each subject, and to determine the initial capacity configuration scheme of each subject based on the source-grid-load-storage collaborative planning model using the Nash equilibrium algorithm.

[0108] Collaborative planning module 3 is used to extract the benefit distribution ratio of each subject from the initial capacity configuration scheme, and to adjust the capacity configuration using a multi-objective optimization algorithm to obtain an optimized capacity planning scheme.

[0109] The benefit optimization module 4 is used to analyze the impact of peak demand fluctuations and transmission capacity limitations on the optimized capacity planning scheme using the Monte Carlo simulation method, and obtain the dynamic adaptability index of the scheme.

[0110] The dynamic evaluation module 5 is used to adjust the parameters of the game theory model based on the dynamic adaptability index, obtain the updated strategy preferences and interaction patterns, and rerun the multi-objective optimization algorithm based on the updated strategy preferences and interaction patterns to adjust the capacity configuration and generate a new capacity planning scheme.

[0111] The rest is the same as in Example 1.

[0112] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0113] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0114] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for source network load storage capacity planning based on multi-agent game, characterized in that, The method comprises the following steps: constructing a multi-agent game behavior dataset, analyzing strategy dynamic interaction based on the multi-agent game behavior dataset using a game theory model, obtaining strategy preferences and interaction modes of each agent; constructing a source-grid-load-storage collaborative planning model based on the strategy preferences and interaction modes of each agent, solving the source-grid-load-storage collaborative planning model to determine an initial capacity configuration scheme of each agent; extracting the benefit distribution proportion of each agent from the initial capacity configuration scheme, adjusting the capacity configuration using a multi-objective optimization algorithm to obtain an optimized capacity planning scheme; analyzing the influence of peak demand fluctuation and power transmission capacity limitation on the scheme based on the optimized capacity planning scheme using a Monte Carlo simulation method to obtain a dynamic adaptability index of the scheme; adjusting the game theory model parameters based on the dynamic adaptability index to obtain updated strategy preferences and interaction modes, and adjusting the capacity configuration through iterative calculation based on the updated strategy preferences and interaction modes to generate a new capacity planning scheme; The dynamic adaptability index of the scheme is obtained through the following steps: extracting uncertainty factor data from the optimized capacity planning scheme to obtain a standardized uncertainty dataset; generating multiple peak demand fluctuation scenarios using a Monte Carlo simulation method based on the standardized uncertainty dataset to obtain a demand distribution feature set; extracting peak demand fluctuation parameters from the demand distribution feature set, combining with the power transmission capacity limitation condition to calculate fluctuation impact weights, and obtaining fluctuation impact evaluation results; extracting dynamic adaptability features based on the fluctuation impact evaluation results to obtain a dynamic adaptability index.

2. The multi-agent game based source network load storage capacity planning method according to claim 1, characterized in that, The multi-agent game behavior dataset is constructed based on historical decision data and real-time operation data of source-grid-load-storage agents.

3. The method of claim 1, wherein, The strategy preferences and interaction modes of each agent are obtained through the following steps: performing data cleaning on the multi-agent game behavior dataset to generate a structured behavior dataset; dividing agent behavior patterns based on the structured behavior dataset using a clustering algorithm to determine agent behavior characteristics; analyzing multi-agent games based on the agent behavior characteristics using a game theory model to obtain strategy interaction dynamics, and extracting the trend of agent strategies over time using time series analysis to determine strategy preferences, and if the deviation of the strategy preferences from the historical behavior patterns exceeds a preset behavior threshold, adjusting the game theory model parameters through regression analysis to obtain optimized strategy preferences; analyzing the interaction modes between agents based on the agent behavior characteristics using an association rule mining method, and predicting future interaction strategy selection between agents.

4. The method of claim 1, wherein, The constructed source-grid-load-storage collaborative planning model is: ， ， wherein, denotes a first cost coefficient, a second cost coefficient, a third cost coefficient, denotes the wind power installation capacity, denotes the photovoltaic installation capacity, denotes the energy storage power capacity, denotes the maximum transmission line capacity, denotes the peak load demand, denotes the upper limit of the wind power installation capacity, denotes the upper limit of the photovoltaic installation capacity, denotes the upper limit of the energy storage installation capacity.

5. The method of claim 1, wherein, The multi-objective optimization algorithm is used to adjust the capacity configuration to obtain an optimized capacity planning scheme, specifically: constructing a multi-objective optimization model, solving to obtain a conflict-balanced benefit distribution set, if the distribution proportion of any agent in the benefit distribution set is lower than a preset threshold, adjusting the distribution proportion through linear interpolation to obtain an adjusted benefit distribution set; based on the adjusted benefit distribution set, optimizing the capacity configuration parameters based on the initial capacity configuration scheme using a genetic algorithm to obtain a preliminary optimized capacity planning scheme; Capacity configuration features are extracted from the preliminarily optimized capacity planning scheme, clustering analysis technology is used to divide the capacity configuration mode between subjects, and a classified configuration mode set is obtained; According to the classified configuration mode set, the time series prediction method is used to analyze the dynamic change trend of the capacity configuration, and the predicted capacity configuration result is obtained; If the deviation of the predicted capacity configuration result from the initial capacity configuration scheme exceeds the preset threshold, the capacity configuration parameters are adjusted, and the final capacity planning scheme is obtained as the optimized capacity planning scheme.

6. The method of claim 5, wherein, The multi-objective optimization model for obtaining the conflict-balanced benefit allocation set is: ， ， ， ， wherein denotes the supply-side payment function, denotes the load-side payment function, denotes the source-side payment function, denotes the peak-period market clearing constraint, denotes the off-peak market clearing constraint, denotes the price-linear inverse demand function, denotes the wind power installation capacity, denotes the photovoltaic installation capacity, denotes the energy storage power capacity, , , denote the first, second, and third strategy variables, respectively, denotes the peak electricity price, denotes the off-peak electricity price, denotes the peak demand, denotes the off-peak demand, denotes the energy storage unit charging cost, denotes the maximum transferable load amount, , , , denote the first, second, third, and fourth inverse demand functions, respectively, denotes the transmission line maximum capacity, denotes the payment function, denotes the other principal equilibrium strategy, W, PV, B, and L denote the wind power set, the photovoltaic set, the energy storage set, and the load set, respectively.

7. The method of claim 1, wherein, The method further comprises: Judging whether the new capacity planning scheme meets the overall system benefit requirement, if the overall system benefit does not reach the preset threshold, obtaining real-time feedback from source network load storage operation data, updating Monte Carlo simulation parameters, re-evaluating the dynamic adaptability index of the scheme, and obtaining the final capacity planning scheme. 8.A source network load storage capacity planning system based on multi-agent game, characterized in that, Comprise: A data acquisition module is configured to construct a multi-agent game behavior dataset, analyze the dynamic interaction of strategies based on the multi-agent game behavior dataset using a game theory model, and obtain the strategy preference and interaction mode of each agent; A game analysis module is configured to construct a source network load storage collaborative planning model based on the strategy preference and interaction mode of each agent, solve the source network load storage collaborative planning model to determine the initial capacity configuration scheme of each agent; A collaborative planning module is configured to extract the benefit allocation ratio of each agent from the initial capacity configuration scheme, adjust the capacity configuration using a multi-objective optimization algorithm, and obtain an optimized capacity planning scheme; A benefit optimization module is configured to analyze the influence of peak demand fluctuation and power transmission capacity limitation on the scheme based on the optimized capacity planning scheme using a Monte Carlo simulation method, and obtain a dynamic adaptability index of the scheme; A dynamic evaluation module is configured to adjust the game theory model parameters based on the dynamic adaptability index, obtain updated strategy preference and interaction mode, and adjust the capacity configuration based on the updated strategy preference and interaction mode through iterative calculation to generate a new capacity planning scheme; The dynamic adaptability index of the scheme is obtained by the following steps: Uncertainty factor data is extracted from the optimized capacity planning scheme to obtain a standardized uncertainty dataset; According to the standardized uncertainty dataset, a Monte Carlo simulation method is used to generate multiple peak demand fluctuation scenarios to obtain a demand distribution feature set; Peak demand fluctuation parameters are extracted from the demand distribution feature set, and the fluctuation influence weight is calculated in combination with the power transmission capacity limitation condition to obtain a fluctuation influence evaluation result; The dynamic adaptability feature is extracted based on the fluctuation influence evaluation result to obtain the dynamic adaptability index.

9. A computer apparatus comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program, when executed by the processor, causes the processor to perform the method of any one of claims 1 to 8. The processor executes the computer program to implement the steps of the method of any one of claims 1-8.

Images