Cross-platform power transaction data interaction optimization method

Abstract

The invention discloses a cross-platform power transaction data interaction optimization method, and particularly relates to the technical field of power transaction data digital twinning. Simulation transaction data of market participants is generated based on an antagonistic neural network model; identifying a causal relationship among market variables from the simulation transaction data of the digital twin through a causal discovery algorithm, and constructing a causal graph; modeling spatial dependence and causal conduction paths among market participants by using a graph neural network model, and performing anti-factual reasoning in a single external impact scene in a digital twinborn body by intervening key variables in a causal graph; simulating dynamic game behaviors of market participants in an extreme weather parameter distribution scene; the power market failure assessment is carried out by combining the causal conduction path to obtain the power market failure assessment index, and early warning is carried out based on the power market failure assessment index, so that the adaptability and stability of the power market in extreme weather scenes are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twin technology for power trading data. More specifically, the present invention relates to a cross-platform power trading data interaction optimization method. Background Art

[0002] As the core platform for energy distribution and trading, the power market needs to operate stably under extreme weather scenarios to ensure power supply reliability and price reasonableness; power market regulatory authorities rely on data-driven risk assessment methods to predict market behavior and formulate response strategies. However, market fluctuations caused by extreme weather events (such as droughts, storms) are highly uncertain, and real transaction data is often scarce and difficult to directly use for analysis; digital twin technology simulates power market behavior through a virtual environment, providing a potential solution for assessing market failure risks, but existing methods have limitations in data generation, causal modeling, and dynamic simulation, restricting their application in complex scenarios.

[0003] Existing power market risk assessment methods are limited by the scarcity of real transaction data and are difficult to accurately simulate market behavior under extreme weather scenarios in a digital twin environment, resulting in difficulties in predicting market failure risks, including price fluctuation ranges, supply gap probabilities, and transmission congestion rates; in addition, existing methods mostly adopt static analysis, lack modeling of the dynamic game behavior of market participants, and the risk indicators are not quantified comprehensively enough to capture the systemic impacts of extreme scenarios, restricting the management efficiency of power market regulatory authorities. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a cross-platform power trading data interaction optimization method, which realizes the optimization of cross-platform power trading data interaction through the combination of digital twins, causal reasoning, and reinforcement learning to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solution: A cross-platform power trading data interaction optimization method, comprising the following steps: Step 1: Construct a digital twin of the power market Generate simulation transaction data of market participants, including quotes, power generation, and electricity demand, based on an adversarial neural network model. The generator of the adversarial neural network model takes a multi-dimensional random vector as input to generate simulation transaction data; Step 2: Causal discovery and counterfactual inquiry Identify the causal relationships between market variables from the simulation transaction data of the digital twin through a causal discovery algorithm, and construct a causal graph; use a graph neural network model to model the spatial dependence and causal conduction path among market participants, and through intervening in the key variables in the causal graph, conduct counterfactual reasoning in the digital twin under a single external shock scenario; Step 3: Market failure risk assessment Deploy a multi-agent reinforcement learning model in the digital twin, combine Monte Carlo simulation to generate extreme weather parameter distribution scenarios, and simulate the dynamic game behavior of market participants under extreme weather parameter distribution scenarios; combine the causal conduction path to conduct power market failure assessment, obtain power market failure assessment indicators, and issue early warnings based on the power market failure assessment indicators.

[0006] Preferably, in the construction of the digital twin, embed physical constraint terms into the training process of the adversarial neural network model. The physical constraint terms include the DC power flow equation and the dynamic line capacity constraint. Based on the DC power flow equation, balance the injection power of each node and the line power flow to ensure that the generated simulation transaction data conforms to the node energy conservation condition in the actual operation of the power grid; introduce a dynamic line capacity constraint function, compare the power value of each transmission line in the generated data with its maximum allowable capacity under the current working condition, and constrain the line power not to exceed the real-time capacity threshold through a penalty term; ensure that the generated simulation transaction data meets the transmission capacity limit and node energy balance.

[0007] Preferably, use a graph neural network model to model the spatial dependence and causal conduction path among market participants, and through intervening in the key variables in the causal graph, simulate the counterfactual scenario under external shocks in the digital twin, and predict market price and supply-demand changes, including the following steps: Construct a causal graph: Extract market variables and perform standardization processing; search for the causal graph structure through the Bayesian information criterion of the GES algorithm; embed prior knowledge of the power network topology to constrain the direction of causal edges; verify and determine key variables through Granger causality test and N-1 security criterion; Construct a graph structure based on the causal graph: Market participants are used as the nodes of the graph structure, and the trading relationships and geographical location relationships of market participants are used as the edges of the graph structure; the node features include historical price, supply-demand situation, geographical location, line impedance, and weather sensitivity; Learn the spatial dependence and causal conduction path between nodes through the attention mechanism; learn the spatial dependence and causal conduction path between nodes; Intervene in key variables through the do-calculus rule, simulate the counterfactual scenario under external shocks in the digital twin, and predict market price fluctuations, supply-demand balance, and transmission congestion risks.

[0008] Preferably, the multi-agent reinforcement learning model includes a graph attention network module and the policy network of each agent; The graph attention network module is used to generate node embeddings of market participants to support dynamic game simulation. The node embeddings capture the dynamic interactions in the electricity market through a dynamic attention mechanism, improving the quantitative accuracy of market failure risk assessment; A policy network is configured for each agent. The policy network is a decision-making model based on a deep neural network. It analyzes the simulation transaction data to obtain the agent state, including its own quotation, power generation or electricity demand. Taking the node embedding and the agent state as inputs, along with fuel price, weather conditions, generation cost, and market clearing price in the set of market variables, it outputs the action decision of the market participant. The action decision includes at least one of the magnitude of adjusting the quotation and the plan of changing the power generation or electricity demand; The policy network is obtained through reinforcement learning training based on the simulation transaction data and the causal conduction path in the digital twin. Each agent's decision-making behavior is optimized based on the agent's reward function to adapt to the market dynamics under the extreme weather parameter distribution scenario.

[0009] Preferably, the multi-agent reinforcement learning model is obtained in the following way: Obtain the simulation transaction data of the digital twin and construct a graph structure based on the causal graph; Adopt a graph attention network module to aggregate node neighbor information and capture spatial dependence; input it into the policy network of each agent to guide the action decision; the action decision of the agent is fed back to the graph neural network model to update the market state sequence; Use the digital twin as a simulation environment to provide a dynamic market state sequence; design a reward function for each agent; Initialize the parameters of the graph neural network model and the policy network; run multiple rounds of simulations in the digital twin. The agent takes actions according to the market state sequence output by the graph neural network model and obtains rewards; use the reinforcement learning algorithm to update the policy network parameters, and at the same time optimize the graph neural network model to improve the state prediction accuracy; introduce a competition and cooperation mechanism to balance the goals of multiple agents; Verify the consistency between the power market failure assessment indicators output by the multi-agent reinforcement learning model and the historical data or actual trends; deploy the multi-agent reinforcement learning model to the digital twin to evaluate the power market failure risk in real time.

[0010] Preferably, the operation process of market failure risk assessment includes the following steps: Step 301: Construct a multi-agent environment, define each market participant as an agent, and all agents form an agent set; output the state, action, and reward function of each agent; Step 302: Dynamic game simulation: Run a multi-agent reinforcement learning model in the digital twin, generate extreme weather parameter distribution scenarios in combination with Monte Carlo simulation, simulate the dynamic game behavior of market participants under the extreme weather parameter distribution scenarios, calculate node embeddings through a graph attention network in each round of loop, adjust actions by the agents, verify the physical feasibility of the power market state sequence, update the strategy, and record the market state sequence; Step 303: Quantify the power market failure evaluation indicators: Calculate the power market failure evaluation indicators based on the market state sequence set and the causal conduction path, including the price fluctuation range, the probability of supply gap, and the transmission congestion rate; Summarize the indicators of all scenarios to form a power market failure evaluation indicator set.

[0011] Preferably, the method further includes a power market trading optimization step: Define the power trading rule parameters and the multi-objective optimization function, adopt a model-agnostic meta-learning framework, and generate multiple groups of extreme weather scenarios in the digital twin; Solve the local optimal power trading rule parameters through an inner-loop double-layer game, and update the meta-parameters in the outer loop to improve the cross-scenario generalization ability of the optimal power trading rule parameters; Design a progressive constraint tightening mechanism, and accelerate the optimization process in combination with a graph neural network surrogate model, and output the Pareto optimal power trading rule parameter set to improve the trading stability of the power market under extreme weather.

[0012] Preferably, the operation process of the power market trading optimization includes: Step 401: Define the power trading rule parameters and the multi-objective optimization function. The power trading rule parameters of the power market include the price ceiling, the reserve capacity threshold, and the penalty coefficient. Set the multi-objective optimization function to minimize the deviation of the power market failure evaluation indicator from the preset value. Step 402: Generate multiple groups of extreme weather scenario sets in the digital twin. Each extreme weather scenario is generated by intervening in the key variables in the causal graph; Input the simulation trading data, the causal graph, and the power trading rule parameters and the multi-objective optimization function in Step 401; Divide the scenario set into a training set and a test set. The training set is used for inner-loop optimization, and the test set is used for outer-loop verification. The output is the extreme weather scenario set, and each group of scenarios includes scenario trading data and market variable data. Step 403: Solve the local optimal power trading rule parameters in the inner loop: In the inner loop, solve the local optimal power trading rule parameters of the training set scenario set through a double-layer game method; The input is the training set scenario set and the power trading rule parameters and the multi-objective optimization function; The processing process constructs a double-layer game model. The upper layer is optimized by the regulatory agency for the power trading rule parameters, and the lower layer is adjusted by the market participants based on the parameters to maximize the profit; After the iteration is completed, the output is the set of local optimal power trading rule parameters corresponding to each training set scenario. Step 404: Outer loop updates meta-parameters and accelerates optimization: In the outer loop, update the meta-parameters based on the test set scenario set to improve the cross-scenario generalization ability of the electricity trading rule parameters, and combine the graph neural network proxy model and the progressive constraint tightening mechanism to accelerate optimization; the inputs include the set of locally optimal electricity trading rule parameters, the test set scenario set, as well as the electricity trading rule parameters and the multi-objective optimization function. Update the meta-parameters through meta-learning methods, iterate and optimize until convergence, and the output is the Pareto optimal set of electricity trading rule parameters.

[0013] Preferably, the graph neural network proxy model is a machine learning model based on the graph neural network model, which is used to quickly predict the electricity market failure evaluation index under different electricity trading rule parameters; taking the electricity trading rule parameters as the input and outputting the electricity market failure evaluation index, and approximating and simulating the dynamic behavior of the electricity market by learning the interaction relationships and historical data among market participants.

[0014] Preferably, the specific implementation method of the progressive constraint tightening mechanism is: gradually narrow the adjustable range of the electricity trading rule parameters during the optimization process, and the constraint range decays exponentially with the increase of the optimization rounds; the meta-parameters include the initial value, adjustment step size and direction of the electricity trading rule parameters, which are extracted from the optimization experience of multiple extreme weather scenarios through meta-learning methods, so that the adjustment of the electricity trading rule parameters in the new scenario quickly converges to the Pareto optimal solution.

[0015] The technical effects and advantages of the present invention: (1) The cross-platform electricity trading data interaction optimization method provided by the present invention simulates the electricity market behavior through digital twins, evaluates the market failure risk under extreme weather scenarios, facilitates timely early warning, and provides support for the decision-making of the electricity market; generates the simulation trading data of market participants based on the adversarial neural network model, constructs a causal graph and conducts counterfactual reasoning through the causal discovery algorithm and the graph neural network model, and evaluates the electricity market failure evaluation index by combining the graph neural network model and multi-agent reinforcement learning with Monte Carlo simulation.

[0016] (2) The cross-platform electricity trading data interaction optimization method provided by the present invention optimizes the electricity trading rule parameters of the electricity market based on the electricity market failure evaluation index. Through the inner loop double-layer game and the outer loop meta-parameter update, combined with the progressive constraint tightening mechanism and the graph neural network proxy model, output the Pareto optimal rule parameter set, and improve the adaptability and stability of the electricity market under extreme weather scenarios. Brief Description of the Drawings

[0017] Figure 1 It is the cross-platform electricity trading data interaction optimization flow chart of the present invention.

[0018] Figure 2 This is the flowchart for optimizing power market transactions of the present invention. Detailed implementation manners

[0019] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0020] Meanwhile, it should be understood that, for the convenience of description, the dimensions of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0021] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way a limitation on the present application, its application, or its use.

[0022] Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and devices should be regarded as part of the specification.

[0023] Embodiment 1. Refer to Figure 1 the flowchart for optimizing cross-platform power trading data interaction, the present invention provides a cross-platform power trading data interaction optimization method as shown in Figure 1 the following, including the following steps: Step 1: Construct a digital twin of the power market Generate simulation trading data of market participants in the power market based on an adversarial neural network model, including quotes, power generation, and electricity demand. The generator of the adversarial neural network model takes a multi-dimensional random vector as input to generate simulation trading data, and the discriminator evaluates the distribution consistency between the simulation trading data and the real data, and outputs the authenticity probability value corresponding to the simulation trading data; Explanation: Provide power market variables and power network topology information to build the basis for the digital twin; Step 2: Causal discovery and counterfactual inquiry Identify the causal relationships between market variables from the simulation transaction data of the digital twin through a causal discovery algorithm (PC algorithm or GES algorithm), and construct a causal graph; use a graph neural network model to model the spatial dependence and causal conduction path among market participants, and through intervening in the key variables in the causal graph, conduct counterfactual reasoning in the digital twin under a single external shock scenario; predict the price changes and supply-demand balance in the electricity market; market variables such as fuel price, weather condition, generation cost, market clearing price, etc.; provide an explanation, offer a structured expression of the causal relationship, clarify the conduction mechanism among key variables, and provide a theoretical framework for risk assessment; Step 3: Market failure risk assessment Deploy a multi-agent reinforcement learning model in the digital twin, combine with Monte Carlo simulation to generate extreme weather parameter distribution scenarios, and simulate the dynamic game behavior of market participants under extreme weather parameter distribution scenarios; combine with the causal conduction path to conduct electricity market failure assessment, and obtain electricity market failure assessment indicators, such as quantifying the price fluctuation range, supply gap probability, and transmission congestion rate, and provide decision-making support for the electricity market based on the electricity market failure assessment indicators.

[0024] In a possible embodiment, based on the loss function of a conventional adversarial neural network model, add a physical constraint term. During the training process of the adversarial neural network model, after the generator generates simulation transaction data each time, calculate the physical constraint term loss through the energy conservation loss function, that is, the degree of violation of the physical constraint; add the physical constraint loss and the loss of the conventional adversarial neural network model as the overall loss function of the generator; In the embodiments of the present invention, it needs to be further explained that in the construction of the digital twin, embed the physical constraint term into the training process of the adversarial neural network model. The physical constraint term includes the DC power flow equation and the dynamic line capacity constraint, which specifically includes the following content: based on the DC power flow equation, balance the injection power of each node and the line power flow to ensure that the generated simulation transaction data conforms to the node energy conservation condition in the actual operation of the power grid (that is, the algebraic sum of power generation, load, and line loss is zero); at the same time, introduce a dynamic line capacity constraint function, compare the power value of each transmission line in the generated data with its maximum allowable capacity under the current working condition, and constrain the line power not to exceed the real-time capacity threshold through a penalty term; ensure that the generated simulation transaction data meets the transmission capacity limit and node energy balance, and the penalty term can be a loss function based on the degree of line power overlimit.

[0025] In the embodiments of the present invention, it needs to be further explained that use a graph neural network model to model the spatial dependence and causal conduction path among market participants, and through intervening in the key variables in the causal graph, simulate counterfactual scenarios under external shocks in the digital twin, and predict market price and supply-demand changes, including the following steps: Step 101: Extract market variables and perform standardization processing; search for causal graph structures through the Bayesian information criterion of the GES algorithm; embed prior knowledge of the power grid topology to constrain the direction of causal edges (e.g., power plant → transmission line); verify and determine key variables (such as regional power generation and line failure rate) through Granger causality test and N-1 security criterion; Step 102: Market participants (such as power generators, users, transmission companies, etc.) serve as nodes in the graph structure, and the trading relationships and geographical location relationships of market participants serve as edges in the graph structure; node features include historical prices, supply and demand situations, geographical locations, line impedances, and weather sensitivities; In a possible embodiment, during the process of converting the causal graph into a graph structure, through multi-layer attention calculation, attention coefficients are assigned to each pair of nodes based on node features and edge attributes, and neighbor node features are aggregated to generate node embeddings, representing the spatial dependence and causal conduction paths among market participants; Step 103: Learn the spatial dependence and causal conduction paths between nodes through the attention mechanism; learn the spatial dependence and causal conduction paths between nodes; Step 104: Intervene in key variables through the do-calculus rule (such as setting the power generation in a certain area to zero or adjusting the generation cost), simulate counterfactual scenarios under external shocks in the digital twin, and predict market price fluctuations, supply and demand balance, and transmission congestion risks.

[0026] For example, simulate the impact of an extreme weather event (such as a severe drought) on a region mainly relying on hydropower. By intervening in the generation cost node of this region, observe how this change propagates to other regions through the graph structure and affects the price and supply and demand balance of the entire market; this method can help market managers predict potential market risks in advance and formulate corresponding coping strategies.

[0027] Explanation: The do-calculus rule is a set of rules proposed by Judea Pearl for dealing with interventions in causal inference. It is based on causal graphs and helps to infer causal effects from observational data in the presence of hidden variables or unobserved confounding factors.

[0028] Furthermore, the key variables in the causal graph refer to: variables that have a significant causal impact on the risk of power market failure (such as price fluctuations and supply gaps), including: External shock variables: Regional power generation affected by extreme weather, transmission line failure rate; Market conduction variables: Node electricity price, cross-regional transmission power, reserve capacity demand; Physical constraint variables: Dynamic line capacity, node power balance deviation.

[0029] The acquisition method of the key variables is as follows: extract the causal dependencies between variables from the simulation transaction data through a causal discovery algorithm (such as weather → power generation → price); perform Granger causality test or Bayesian posterior probability analysis on the edge weights in the causal graph, and filter out the causal relationships with significance (p < 0.05); combine the operation rules of the power system and expert experience to determine the key variables (such as the transmission corridor capacity corresponding to the storm area).

[0030] Furthermore, use the key variables in the causal graph (such as weather conditions, power generation costs) to design the parameter ranges of multiple extreme weather scenarios (such as rainfall from 0 to 50 mm, power generation cost increase from 5% to 15%) to provide an input template for Monte Carlo simulation; for example, based on the causal conduction path of the drought scenario (weather conditions → power generation cost → market clearing price), set the intervention variable for the storm scenario (such as a 20% decrease in the transmission line capacity).

[0031] Furthermore, integrate the node embedding of the causal graph (the vector generated by GAT, representing the state of market participants) with the simulation transaction data, and convert it into the graph structure format required in step three (the node features include quotes and power generation, and the edges represent transaction relationships); configure the data interface to ensure that the multi-agent reinforcement learning model in step 3 can directly call the causal graph and prediction results.

[0032] In the embodiments of the present invention, it needs to be further explained that the multi-agent reinforcement learning model includes a graph attention network module and the policy networks of each agent. The graph attention network module is used to generate the node embedding of market participants to support dynamic game simulation. The node embedding is a way to represent the state and relationship of market participants with numbers, usually a fixed-length vector; the node embedding captures the dynamic interactions in the power market through a dynamic attention mechanism, improving the quantitative accuracy of market failure risk assessment. Configure a policy network for each agent. The policy network is a decision-making model based on a deep neural network. Analyze the simulation transaction data to obtain the agent state, including its own quotes, power generation or electricity demand. Using the node embedding and the agent state as inputs, as well as the fuel price, weather conditions, power generation cost, and market clearing price in the set of market variables, output the action decision of the market participant. The action decision includes at least one of the magnitude of adjusting the quote and the plan to change the power generation or electricity demand; the policy network is obtained through reinforcement learning training based on the simulation transaction data and the causal conduction path in the digital twin, and optimizes the decision-making behavior of each agent based on the reward function of the agent to adapt to the market dynamics under the extreme weather parameter distribution scenario.

[0033] Explanation: A strategy network is built using a deep neural network. Based on the causal conduction path, initial dependencies are preset (for example, a higher weight of the hydropower plant node on the adjacent load center node). The training parameters of the multi-agent reinforcement learning model are set, including the learning rate (0.001) and the discount factor (0.95), and the initialized multi-agent reinforcement learning model is output. The update of the agent strategy network is calibrated by pre-collected transaction data to ensure that the decision-making characteristics of each participant are consistent with the historical data, forming a closed-loop feedback mechanism and laying a data foundation for subsequent risk assessment. Furthermore, the present invention does not limit the specific quantization formula of the reward function. The reward function is set based on the attributes of the agents. For example, the reward function of the power generator is based on profit, calculated as the market clearing price multiplied by the power generation volume minus the power generation cost multiplied by the power generation volume; the reward function of the user is based on the negative value of the electricity consumption cost, calculated as the market clearing price multiplied by the opposite of the electricity demand; the reward function of the transmission company is based on the negative cost of transmission congestion.

[0034] In the embodiments of the present invention, it needs to be further explained that the acquisition method of the multi-agent reinforcement learning model is as follows: Step 201: Obtain the simulation transaction data of the digital twin, including bid prices, power generation volumes, electricity demands, etc.; supplement external data, such as market variables (fuel prices, weather conditions, power generation costs) and power network topology information (line impedance, capacity limits); construct a graph structure based on the causal graph. Step 202: Use the graph attention network module to aggregate node neighbor information and capture spatial dependencies; for example, the graph attention network module dynamically assigns edge weights through the attention mechanism to highlight key interactions; each agent uses an independent strategy network (such as a deep neural network) to optimize decisions based on the reinforcement learning algorithm; the graph neural network model outputs node embeddings (representing the market state sequence), which are input into the strategy network of each agent to guide action decisions; the action decisions of the agents (such as adjusting bid prices) are fed back to the graph neural network model to update the market state sequence. Step 203: Use the digital twin as a simulation environment to provide a dynamic market state sequence (such as price changes, supply-demand balance); design a reward function for each agent. Step 204: Initialize the parameters of the graph neural network model and the strategy network; run multiple rounds of simulations in the digital twin. The agents take actions according to the market state sequence output by the graph neural network model to obtain rewards; use the reinforcement learning algorithm to update the parameters of the strategy network, and at the same time optimize the graph neural network model to improve the state prediction accuracy; introduce a competition and cooperation mechanism (such as sharing part of the rewards) to balance the goals of multiple agents. Step 205: Verify the consistency between the power market failure evaluation indicators (such as price fluctuation threshold, transmission congestion rate) output by the multi-agent reinforcement learning model and historical data or actual trends; deploy the multi-agent reinforcement learning model to the digital twin to evaluate the power market failure risk in real time.

[0035] In the embodiments of the present invention, it needs to be further explained that the operation process of market failure risk assessment includes the following steps: Step 301: Construct a multi-agent environment, define each market participant (such as power generators, users, transmission companies) as an agent, and all agents form an agent set; output the state, action, and reward function of each agent. Step 302: Dynamic game simulation: Run the multi-agent reinforcement learning model in the digital twin, combine Monte Carlo simulation to generate extreme weather parameter distribution scenarios, simulate the dynamic game behavior of market participants in the extreme weather parameter distribution scenarios, calculate node embeddings through a graph attention network in each round of loop, the agent adjusts its action according to the node embedding, verify the physical feasibility of the power market state sequence, update the policy and record the market state sequence. Step 303: Quantify the power market failure evaluation indicators: Based on the set of market state sequences and the causal conduction path, calculate the power market failure evaluation indicators, including price fluctuation amplitude, supply gap probability, and transmission congestion rate; summarize the indicators of all scenarios to form a set of power market failure evaluation indicators.

[0036] In the embodiments of the present invention, it needs to be further explained that the way to obtain the price fluctuation amplitude is as follows: For each set of market state sequences, extract the time series of the power market clearing price, calculate the price change range, which is defined as the difference between the maximum price and the minimum price, and take the 95th percentile of the price change ranges of all scenarios as the price fluctuation amplitude to ensure capturing extreme situations. The way to obtain the supply gap probability is as follows: For each set of market state sequences, calculate the proportion of unmet electricity demand in each simulation (electricity demand minus power generation, divided by electricity demand); count the number of simulations with an unmet proportion greater than 0, divide by the total number of simulations to obtain the supply gap probability, and the unit is percentage. The way to obtain the transmission congestion rate is as follows: For each set of market state sequences, extract the time series of the transmission line load rate; calculate the proportion of the number of times the load rate exceeds the dynamic line capacity constraint as the transmission congestion rate, and the unit is percentage; combine the causal conduction path and give priority to analyzing key lines (for example, lines connecting high-load areas).

[0037] Explanation: The connection between Step 2 and Step 3 lies in that the causal graph generated in Step 2 is used to guide the variable intervention in extreme scenarios in Step 3 (for example, preferentially intervening in the root node variables in the causal graph), and the counterfactual reasoning results (such as "the increase in fuel price leads to the increase in power generation cost") are used as the input parameters for the Monte Carlo simulation in Step 3; Step 2 provides static causal relationships, and Step 3 simulates the cascading effect of the causal chain through dynamic game; the single shock analysis in Step 2 verifies the causal hypothesis, and the multi-variable coupling simulation in Step 3 evaluates the systemic risk.

[0038] Summary: The embodiments of the present invention focus on market modeling and risk assessment, forming a complete analysis process from data generation, causal analysis to risk quantification; Step 1 generates simulation transaction data (quotation, power generation, electricity demand), Step 2 constructs a causal graph and predicts the market changes in a single scenario, and Step 3 quantifies the power market failure evaluation index through a multi-agent reinforcement learning model. The three steps progress in sequence to form a closed loop of market analysis.

[0039] Embodiment 2 is different from Embodiment 1 in that the method further includes: Step 4: Power market transaction optimization Define the power trading rule parameters and the multi-objective optimization function, adopt a model-agnostic meta-learning framework, and generate multiple groups of extreme weather scenarios in the digital twin; solve the local optimal power trading rule parameters through an inner-loop double-layer game, and update the meta-parameters in the outer loop to improve the cross-scenario generalization ability of the optimal power trading rule parameters; design a progressive constraint tightening mechanism, combine with a graph neural network proxy model to accelerate the optimization process, output the Pareto optimal power trading rule parameter set, improve the cross-scenario generalization ability, and enhance the trading stability of the power market under extreme weather.

[0040] In the embodiments of the present invention, it needs to be further explained that referring to Figure 2 the power market transaction optimization flowchart, the operation process of the power market transaction optimization includes: Step 401: Define the power trading rule parameters and the multi-objective optimization function. The power trading rule parameters of the power market include the price ceiling, the reserve capacity threshold, and the penalty coefficient, and set the trading price correlation constraint and the dynamic reserve threshold constraint (the adjustable range of the parameters); the price ceiling is used to limit the market clearing price, the reserve capacity threshold is used to ensure power supply reliability, and the penalty coefficient is used to constrain the behavior of market participants and is the fine standard for agent violations, prompting market participants to act according to the rules; set the multi-objective optimization function to minimize the deviation of the power market failure evaluation index from the preset value. In one possible embodiment, the multi-objective optimization function is to minimize the weighted sum of the price fluctuation range, the supply gap probability, and the transmission congestion rate; the weights are set by the regulatory agency according to the market priority. For example, the weight of the price fluctuation range is 0.4, the supply gap probability is 0.3, and the transmission congestion rate is 0.3; Explanation: To ensure the rationality of power trading rule parameters, constraint conditions for power trading rule parameters are imposed. For example, the price ceiling needs to be higher than the market average clearing price in the simulation trading data, the reserve capacity threshold needs to meet the N - 1 security criterion (i.e., the system can still operate normally under a single component failure), and the penalty coefficient needs to be sufficient to motivate market participants to comply with the rules; Generate multiple sets of extreme weather scenario sets in the digital twin and divide them into a training set and a test set; solve the local optimal power trading rule parameters for each training set scenario through an inner - loop double - layer game. In the upper layer, the regulatory agency optimizes the power trading rule parameters, and in the lower layer, market participants conduct strategic games based on the parameters; in the outer loop, update the meta - parameters based on the power market failure evaluation indicators of the test set scenarios to improve the cross - scenario generalization ability of the power trading rule parameters; design a progressive constraint tightening mechanism to gradually narrow the range of power trading rule parameters as the number of optimization rounds increases to ensure the convergence stability of the optimization process; combine with a graph neural network proxy model to predict the power market failure evaluation indicators with the power trading rule parameters as the input to accelerate the optimization process; Step 402: Generate multiple sets of extreme weather scenario sets in the digital twin. Each extreme weather scenario is generated by intervening in the key variables (such as fuel price, weather conditions) in the causal graph; the inputs include the simulation trading data (bid price, power generation, electricity demand) in step 1, the causal graph (causal relationship between market variables) in step 2, the power trading rule parameters in step 401, and the multi - objective optimization function; divide the scenario set into a training set (accounting for 80%) and a test set (accounting for 20%). The training set is used for inner - loop optimization, and the test set is used for outer - loop verification. The output is the extreme weather scenario set, and each set of scenarios contains scenario trading data and market variable data; Explanation: The extreme weather scenario sets include drought, storm, and high temperature; for example, in the drought scenario, the key variables are adjusted through the causal graph (such as a 10% increase in the cost of hydropower generation), and the behavior of market participants is simulated in the digital twin to generate scenario trading data (updated bid price, power generation); the scenario generation refers to the counterfactual reasoning results in step 2, such as the causal chain of increased fuel price leading to increased generation cost, to ensure that the scenario is consistent with the market variables. The market variable data refers to the simulated corresponding market behavior in the digital twin; Step 403: Solve for the locally optimal electricity trading rule parameters in the inner loop: In the inner loop, solve for the locally optimal electricity trading rule parameters of the training set scenario set through a two-layer game method; the inputs are the training set scenario set (including scenario trading data), electricity trading rule parameters, and multi-objective optimization function; the processing process constructs a two-layer game model, where the upper layer optimizes the electricity trading rule parameters by the regulatory agency, and the lower layer adjusts strategies by market participants based on the parameters to maximize profits; after completing the iteration, the output is the set of locally optimal electricity trading rule parameters corresponding to each training set scenario. Explanation: In the lower-layer game, market participants (such as power generators) adjust their bids and power generation volumes according to the current electricity trading rule parameters (for example, the price cap is 120 yuan / MWh), and the profit is calculated as the market clearing price multiplied by the power generation volume minus the power generation cost. The solution result is the optimal strategy combination; in the upper-layer optimization, the regulatory agency adjusts the parameters according to the lower-layer results, with the goal of minimizing the electricity market failure evaluation index in Step 3; the optimization uses the gradient descent method with a step size of 0.01, and the convergence condition is that the parameter change amplitude is less than 0.001. Step 404: Update the meta-parameters and accelerate optimization in the outer loop: In the outer loop, update the meta-parameters based on the test set scenario set to improve the cross-scenario generalization ability of the electricity trading rule parameters, and accelerate the optimization by combining the graph neural network proxy model and the progressive constraint tightening mechanism; the inputs include the set of locally optimal electricity trading rule parameters, the test set scenario set, electricity trading rule parameters, and multi-objective optimization function. Update the meta-parameters through meta-learning methods, and iterate and optimize until convergence. The output is the Pareto optimal set of electricity trading rule parameters.

[0041] Explanation: The test set scenarios are used to evaluate the performance of the meta-parameters, calculate the electricity market failure evaluation index under the current electricity trading rule parameters, and compare it with the locally optimal parameters in Step 403; the update direction of the meta-parameters is to minimize the average value of the electricity market failure evaluation index of the test set, with a step size of 0.001; the progressive constraint tightening mechanism gradually narrows the parameter range; the graph neural network proxy model takes the electricity trading rule parameters as inputs and predicts the electricity market failure evaluation index (for example, input the price cap of 120 yuan / MWh and predict the supply gap probability of 5%), reducing the computational cost of digital twin simulation; the output Pareto optimal set of electricity trading rule parameters contains multiple combinations, such as a price cap of 120 yuan / MWh, a reserve capacity of 100 MW, and a penalty coefficient of 12 yuan / MWh; a price cap of 130 yuan / MWh, a reserve capacity of 90 MW, and a penalty coefficient of 13 yuan / MWh, which are applicable to different extreme weather scenarios.

[0042] In the embodiments of the present invention, it needs to be further explained that the graph neural network proxy model is a machine learning model based on the graph neural network model, which is used to quickly predict the power market failure evaluation index under different power trading rule parameters; taking the power trading rule parameters as input and outputting the power market failure evaluation index, approximating and simulating the dynamic behavior of the power market by learning the interaction relationships and historical data among market participants. The core features include: Input: Power trading rule parameters, such as price ceiling, reserve capacity threshold, penalty coefficient, etc.; Output: Power market failure evaluation index, such as price fluctuation range, supply gap probability, transmission congestion rate, etc.; Function: Accelerate the optimization of power trading rule parameters.

[0043] Furthermore, the specific implementation method of the progressive constraint tightening mechanism is: gradually narrow the adjustable range of power trading rule parameters during the optimization process, and the constraint range decays exponentially with the increase of the optimization rounds; the meta-parameters include the initial value, adjustment step size and direction of the power trading rule parameters, which are extracted from the optimization experiences of multiple extreme weather scenarios through the meta-learning method, so that the adjustment of power trading rule parameters in the new scenario can quickly converge to the Pareto optimal solution.

[0044] In the embodiments of the present invention, it needs to be further explained how to obtain the graph neural network proxy model: Based on the simulation trading data in Step 1, the causal graph in Step 2, and the power market failure evaluation index in Step 3, construct and train in the digital twin to predict the power market failure evaluation index corresponding to the trading rule parameters; First, obtain the simulation trading data (bid price, power generation, electricity demand) as the initial node features, extract the trading relationships and transmission lines among market participants (power generators, users, transmission companies) from the causal graph as the edges of the graph structure, and construct the graph structure, where the node features are supplemented with the power market failure evaluation index; Then, design a graph attention network as the proxy model, with the input being the trading rule parameters (price ceiling, reserve capacity threshold, penalty coefficient), and aggregate the node features through multi-layer attention calculation to generate the predicted power market failure evaluation index; Utilize the set of extreme weather scenarios generated in Step 4, train the model with the training set scenarios (80%), optimize the GAT parameters to make the predicted value close to the actual power market failure evaluation index, and verify the model accuracy with the test set scenarios (20%); Finally, output the graph neural network proxy model to predict the effect of trading rule parameters and reduce the computational cost of digital twin simulation.

[0045] Explanation: The model-agnostic meta-learning framework is used to help the electricity market quickly find the electricity trading rule parameters suitable for different extreme weather scenarios without having to start adjusting from scratch every time. It is not limited to a specific algorithm. In the inner loop, for each specific scenario (such as a drought scenario), it quickly adjusts the electricity trading rule parameters to find the parameter combination that performs best in this scenario (local optimal electricity trading rule parameters). In the outer loop, it summarizes the experiences of multiple scenarios (such as drought, storm, high temperature) to find a general adjustment method (meta-parameters), enabling the system to quickly find suitable parameters when facing new scenarios. In the model-agnostic meta-learning framework, the meta-parameters are a set of parameters used to initialize and guide the optimization of electricity trading rule parameters. The meta-parameters play their roles in the following ways: Set initial values for the electricity trading rule parameters (for example, a price ceiling of 120 yuan / MWh, a reserve capacity of 100 MW, a penalty coefficient of 12 yuan / MWh), ensuring that there is no need to start from scratch when optimizing at the beginning of each new extreme weather scenario (such as high temperature or storm). Determine the adjustment direction and step size of the electricity trading rule parameters, and evaluate the optimization effect according to the electricity market failure assessment index (for example, reducing the price fluctuation range to below 50 yuan / MWh), ensuring that the adjustment process is efficient and stable. By summarizing the optimization experiences of multiple groups of extreme weather scenarios, construct a general adjustment strategy, enabling the electricity trading rule parameters to quickly converge to the Pareto optimal electricity trading rule parameter set in unseen scenarios (such as future new storms), and maintaining the stability of the electricity market.

[0046] Summary: The embodiments of the present invention focus on trading rule optimization. Using the simulation trading data, causal diagrams, and electricity market failure assessment indicators output by Embodiment 1, a Pareto optimal trading rule parameter set (price ceiling, reserve capacity threshold, penalty coefficient) is generated through the model-agnostic meta-learning framework, improving the cross-scenario generalization ability and the trading stability of the electricity market under extreme weather.

[0047] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A cross-platform power trading data interaction optimization method, characterized in that: The following steps are involved: Generate simulated transaction data of market participants based on adversarial neural network model, including quotations, power generation and electricity demand; Use causal discovery algorithms to identify causal relationships between market variables from simulated transaction data of digital twins and construct causal graphs. Use graph neural network models to model spatial dependencies and causal transmission paths between market participants, and perform counterfactual reasoning in digital twins under a single external shock scenario by intervening in key variables in the causal graph. A multi-agent reinforcement learning model is deployed in the digital twin, and the Monte Carlo simulation is combined to generate extreme weather parameter distribution scenarios, simulating the dynamic game behavior of market participants in extreme weather parameter distribution scenarios; the causal transmission path is combined to conduct power market failure assessment, obtain power market failure assessment indicators, and issue early warnings based on the power market failure assessment indicators.

2. A cross-platform power trading data interactive optimization method according to claim 1, characterized in that: In the construction of digital twins, physical constraints are embedded in the training process of the adversarial neural network model. The physical constraints include DC power flow equations and dynamic line capacity constraints. Based on the DC power flow equations, the injected power of each node and the line power flow are balanced to ensure that the generated simulation transaction data meets the node energy conservation conditions in the actual operation of the power grid. A dynamic line capacity constraint function is introduced to compare the power value of each transmission line in the generated data with the maximum allowable capacity under its current working conditions, and the line power is constrained not to exceed the real-time capacity threshold through a penalty term; Ensure that the generated simulation transaction data meets the transmission capacity constraints and node energy balance.

3. A cross-platform power trading data interactive optimization method according to claim 1, characterized in that: The spatial dependency and causal transmission path between market participants are modeled using a graph neural network model. By intervening in the key variables in the causal graph, counterfactual scenarios under external shocks are simulated in the digital twin to predict market prices and supply and demand changes, including the following steps: Construct causal graph: extract market variables and standardize them; search for causal graph structure through Bayesian information criterion of GES algorithm; embed prior knowledge of power network topology and constrain the direction of causal edges; determine key variables through Granger causality test and N-1 safety criterion verification; Build a graph structure based on the causal graph: market participants serve as nodes of the graph structure, and the transaction relationships and geographical location relationships of market participants serve as edges of the graph structure; Node characteristics include historical prices, supply and demand, geographic location, line impedance, and weather sensitivity; Learn the spatial dependencies and causal paths between nodes through the attention mechanism; Learn the spatial dependencies and causal paths between nodes; By intervening in key variables through do-calculus rules, counterfactual scenarios under external shocks are simulated in the digital twin to predict market price fluctuations, supply and demand balance, and transmission congestion risks.

4. A cross-platform power trading data interactive optimization method according to claim 3, characterized in that: The multi-agent reinforcement learning model includes a graph attention network module and a strategy network of each agent; The graph attention network module is used to generate node embeddings of market participants to support dynamic game simulation. The node embeddings capture the dynamic interactions of the power market through a dynamic attention mechanism to improve the quantitative accuracy of market failure risk assessment; A policy network is configured for each intelligent agent. The policy network is a decision-making model based on a deep neural network. The simulated transaction data is analyzed to obtain the agent state, including its own quotation, power generation or electricity demand. The node embedding and agent state are used as input, as well as the fuel price, weather conditions, power generation cost and market clearing price in the market variable set. The action decision of the market participant is output. The action decision includes at least one of adjusting the quotation amplitude and changing the power generation or electricity demand plan. The policy network is obtained through reinforcement learning training based on simulated transaction data and causal conduction paths in the digital twin. The decision-making behavior of each intelligent agent is optimized based on the reward function of the intelligent agent to adapt to the market dynamics under the extreme weather parameter distribution scenario.

5. A cross-platform power trading data interactive optimization method according to claim 4, characterized in that: The multi-agent reinforcement learning model is obtained as follows: Obtain simulated transaction data of the digital twin and build a graph structure based on the cause-effect graph; A graph attention network module is used to aggregate node neighbor information and capture spatial dependencies. Input to each agent’s strategy network to guide action decisions; the agent’s action decisions are fed back to the graph neural network model to update the market state sequence; Use digital twins as simulation environments to provide dynamic market state sequences; design reward functions for each agent; Initialize the graph neural network model and policy network parameters; run multiple rounds of simulation in the digital twin, and the agent takes actions and obtains rewards based on the market state sequence output by the graph neural network model; use the reinforcement learning algorithm to update the policy network parameters, and optimize the graph neural network model to improve the accuracy of state prediction; introduce competition and cooperation mechanisms to balance the goals of multiple agents; Verify the consistency of the power market failure assessment indicators output by the multi-agent reinforcement learning model with historical data or actual trends; Deploy a multi-agent reinforcement learning model to a digital twin to assess power market failure risks in real time.

6. A cross-platform power trading data interactive optimization method according to claim 4, characterized in that: The operational process of market failure risk assessment includes the following steps: Build a multi-agent environment, define each market participant as an agent, and all agents form an agent set; output the state, action, and reward function of each agent; Dynamic game simulation: Run a multi-agent reinforcement learning model in the digital twin, combine Monte Carlo simulation to generate extreme weather parameter distribution scenarios, simulate the dynamic game behavior of market participants in extreme weather parameter distribution scenarios, calculate node embeddings through the graph attention network in each cycle, and the agent adjusts its actions according to the node embeddings to verify the physical feasibility of the power market state sequence, update the strategy and record the market state sequence; Quantify the power market failure assessment indicators: Based on the market state sequence set and causal transmission path, calculate the power market failure assessment indicators, including price fluctuation range, supply gap probability and transmission congestion rate; summarize the indicators of all scenarios to form a set of power market failure assessment indicators.

7. A cross-platform power trading data interactive optimization method according to any one of claims 1-6, characterized in that: It also includes the steps of optimizing electricity market transactions: The power trading rule parameters and multi-objective optimization functions are defined, and a model-independent meta-learning framework is adopted to generate multiple sets of extreme weather scenarios in the digital twin. The local optimal power trading rule parameters are solved through an inner-loop double-layer game, and the meta-parameters are updated in the outer loop to improve the cross-scenario generalization ability of the optimal power trading rule parameters. A progressive constraint tightening mechanism is designed, combined with a graph neural network agent model to accelerate the optimization process, and the Pareto optimal power trading rule parameter set is output to improve the trading stability of the power market under extreme weather conditions.

8. A cross-platform power trading data interactive optimization method according to claim 7, characterized in that: The operation process of power market transaction optimization includes: Define power trading rule parameters and multi-objective optimization functions. The power trading rule parameters of the power market include price cap, reserve capacity threshold and penalty coefficient. The multi-objective optimization function is set to minimize the deviation of the power market failure evaluation index from the preset value. Generate multiple sets of extreme weather scenario sets in the digital twin, each extreme weather scenario is generated by intervening key variables in the causal graph; the input includes simulated transaction data, the causal graph, and the power transaction rule parameters and multi-objective optimization function of step 401; and divide them into training sets and test sets, the training set is used for inner loop optimization, and the test set is used for outer loop verification, and the output is a set of extreme weather scenarios, each set of scenarios includes scenario transaction data and market variable data; In the inner loop, the local optimal power trading rule parameters of the training set scenario set are solved by a two-layer game method; the input is the training set scenario set, power trading rule parameters, and multi-objective optimization function; the processing process constructs a two-layer game model, the upper layer is for the regulator to optimize the power trading rule parameters, and the lower layer is for market participants to adjust strategies based on the parameters to maximize profits; after the iteration is completed, the output is the local optimal power trading rule parameter set corresponding to each set of training set scenarios; In the outer loop, meta-parameters are updated based on the test set scenario set to improve the cross-scenario generalization ability of power trading rule parameters, and the optimization is accelerated by combining the graph neural network proxy model and the progressive constraint tightening mechanism. The input includes the local optimal power trading rule parameter set, the test set scenario set, the power trading rule parameters, and the multi-objective optimization function. The meta-parameters are updated through the meta-learning method, and the optimization is iterated until convergence. The output is the Pareto optimal power trading rule parameter set.

9. A cross-platform power trading data interactive optimization method according to claim 8, characterized in that: The graph neural network agent model is a machine learning model based on the graph neural network model, which is used to quickly predict the power market failure evaluation indicators under different power trading rule parameters; it takes the power trading rule parameters as input and outputs the power market failure evaluation indicators, and approximately simulates the dynamic behavior of the power market by learning the interaction relationship and historical data between market participants.

10. A cross-platform power transaction data interactive optimization method according to claim 8, characterized in that: The specific implementation method of the progressive constraint tightening mechanism is: gradually narrowing the adjustable range of the power trading rule parameters during the optimization process, and the constraint range decays exponentially with the increase in optimization rounds; the meta-parameters include the initial value, adjustment step size and direction of the power trading rule parameters, which are extracted from the optimization experience of multiple groups of extreme weather scenarios through the meta-learning method, so that the power trading rule parameter adjustment under the new scenario can quickly converge to the Pareto optimal solution.

Images