A transaction decision method and system under a market joint clearing environment

Abstract

The application discloses a kind of market joint clearing under the transaction decision-making method and system of environment, it is related to electric power spot market transaction technical field, method includes: step S1, the predicted value of node marginal price is obtained, and the bidding optimization model of each power generation subject is constructed;Step S2, design agent, with bidding optimization model as reward function training agent;Step S3, based on the agent after training generates bidding data, and constructs the joint clearing model of all power generation subjects;Step S4, based on joint clearing model, the actual value of node marginal price is calculated, and the deviation between actual value and predicted value is calculated;Step S5, whether the deviation exceeds preset threshold value is judged, if exceeds, then actual value is used as its predicted value, returns step S1 and carries out closed loop learning;If not exceed, then the bidding data generated by each agent is used as final bidding strategy.The application can realize global optimization and intelligent collaboration of power generation subject transaction decision-making under complex market environment.

Description

Technical Field

[0001] This invention relates to the field of electricity spot market trading technology, and in particular to a trading decision-making method and system under a market joint clearing environment. Background Technology

[0002] With the deepening of power market reforms, the power market system has evolved from a single energy market into a complex structure involving multiple markets operating in tandem, including medium- and long-term markets, spot markets, and ancillary services. Against this backdrop, power generation companies, energy storage operators, and other market participants face the challenge of optimizing decisions across these multiple markets to maximize profits.

[0003] Currently, the technical methods applied to electricity trading decision-making have mainly undergone two generations of evolution. The first generation was based on traditional optimization algorithms, employing cost-plus models or linear programming methods based on a single electricity price prediction curve. These methods relied on accurate market environment predictions and were applicable to a certain extent in the early stages when the market structure was relatively stable. The second generation of technologies began to incorporate artificial intelligence, mainly focusing on strategy optimization in a single market environment. For example, Q-learning algorithms were used to solve multi-stage profit optimization problems under incomplete information, deep deterministic policy gradient algorithms were applied to handle continuous price spaces, and multi-agent reinforcement learning was used to simulate the game behavior among market participants.

[0004] When these existing technologies are applied to a joint clearing environment for the electricity and ancillary services markets, their inherent limitations are gradually exposed. Methodologically, existing technologies treat the electricity and ancillary services markets as independent decision-making environments, adopting a sequential decision-making model of "electricity first, ancillary services later." This fragmented decision-making approach fails to accurately quantify the opportunity costs between the electricity and ancillary service markets, ignoring the mutually exclusive relationship between the two markets in capacity allocation, leading power generators into a "local optimum trap" when allocating resources. At the algorithmic implementation level, existing technologies face severe technical bottlenecks. First, there is the problem of predictive vulnerability; existing methods rely excessively on a single electricity price prediction trajectory, and when actual prices deviate from the prediction due to factors such as fluctuations in renewable energy output and load changes, the decision-making effect deteriorates sharply. Second, there is the curse of dimensionality; in high-dimensional decision spaces, traditional algorithms struggle to obtain high-quality solutions within a limited application time. Furthermore, in multi-party game environments, algorithms often get trapped in local optima or exhibit oscillatory convergence, severely impacting decision-making effectiveness. From a system operation perspective, existing intelligent decision-making technologies have inherent flaws. In the early stages of market operation, intelligent agents, lacking prior knowledge, experience drastic policy fluctuations. These individual policy fluctuations, in turn, trigger market price volatility, creating a negative feedback loop and exacerbating policy instability. Since a lengthy "trial and error learning" phase is required to reach a stable state, this process not only affects the profitability of market participants but also threatens the smooth operation of the electricity market.

[0005] The aforementioned limitations not only restrict the competitiveness of power generation entities in complex market environments, but also hinder the improvement of the overall operational efficiency of the electricity market, and urgently need to be addressed through technological innovation. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a transaction decision-making method and system in a market joint clearing environment. This method and system can achieve global optimization and intelligent collaboration of transaction decisions by power generation entities in the complex environment of joint clearing of the electricity and ancillary services markets, improve the competitiveness of power generation entities in complex market environments, and effectively improve the overall operating efficiency of the electricity market.

[0007] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0008] In a first aspect, the present invention provides a transaction decision-making method under a market joint clearing environment, comprising:

[0009] Step S1: Obtain the predicted value of the marginal electricity price of the node, and construct a bidding optimization model for each power generation entity with the goal of maximizing the comprehensive benefits of the power generation entity.

[0010] Step S2: Design an agent for each power generation entity that can directly output continuous bids, and train the agent using the bid optimization model as the reward function;

[0011] Step S3: Based on the trained agents, generate the bidding data of each power generation entity, and construct a joint clearing model for all power generation entities with the goal of minimizing the total cost of the whole society.

[0012] Step S4: Calculate the actual value of the node marginal electricity price based on the joint clearing model, and calculate the deviation between the actual value and the predicted value of the node marginal electricity price;

[0013] Step S5: Determine whether the deviation exceeds a preset threshold. If it does, use the actual value of the node marginal electricity price as its predicted value and return to step S1 for closed-loop learning. If it does not exceed the threshold, use the bidding data generated by each agent as the final bidding strategy.

[0014] Optionally, the pricing optimization model is:

[0015] In the formula, These are the medium- and long-term electricity revenue and the base electricity revenue of the generating unit in time period t, respectively.

[0016] These represent the day-ahead market revenue and real-time market revenue of the generating unit in the spot electricity market corresponding to time period t; k is the prediction curve of the marginal electricity price at node k.

[0017] These represent the compensation costs and ancillary service market revenue for the generating unit during time period t, respectively.

[0018] These represent the power generation cost, start-up and shutdown cost, and shutdown penalty cost of the generating unit during time period t, respectively. The generating capacity of the unit during time period t;

[0019] Let T be the confidence probability of the predicted curve of the marginal electricity price of the k-th node, where T is the total number of time periods and K is the total number of nodes.

[0020] Optionally, the revenue from day-ahead market bids in the electricity spot market is:

[0021]

[0022] In the formula, For the electricity contracts won in the market before time period t, Let t be the predicted market node electricity price for time period t. The unit's plant power consumption rate;

[0023] The revenue from real-time market-winning electricity bids in the aforementioned electricity spot market is:

[0024]

[0025] In the formula, For the real-time market-winning electricity during time period t, This represents the predicted real-time market node electricity price for time period t.

[0026] Optionally, the constraints for the agent's decision-making process include:

[0027] The price quotation limits and physical operation constraints include upper and lower limits for unit output, unit output ramp-up rate constraints, minimum start-up and shutdown time constraints, and state of charge constraints for energy storage units.

[0028] Optionally, the joint clearing model is:

[0029]

[0030] In the formula, These represent the operating cost, startup cost, and power generation of the i-th unit during time period t. These represent the positive reserve bid and the awarded positive reserve capacity for the i-th generating unit during time period t, respectively; N is the total number of generating units, and T is the total number of time periods;

[0031] Let $\frac{e}{t}$ be the operating cost and charge / discharge power of the $e$-th energy storage unit during time period $t$. These are the positive reserve bid and the winning positive reserve capacity for the e-th energy storage unit during time period t; This represents the total number of energy storage units.

[0032] Let $\frac{p}{t}$ be the operating cost and power generation / pumping capacity of the $p$-th pumped-storage unit during time period $t$. These are the positive reserve bid and the winning positive reserve capacity of the p-th pumped-storage unit during time period t; This represents the total number of pumped storage units;

[0033] The network flow constraint relaxation penalty factor is used for market clearing optimization. are the forward and reverse power flow relaxation variables for the l-th line, respectively, and NL is the total number of lines; , where are the forward and reverse current relaxation variables for the s-th cross section, and NS is the total number of cross sections.

[0034] Optionally, the constraints for calculating the actual value of the nodal marginal electricity price based on the joint clearing model include:

[0035] Constraints include power balance, unit operating characteristics, unit contracted power volume, network power flow, and ancillary service demand.

[0036] Secondly, the present invention provides a transaction decision-making system under a market joint clearing environment, comprising:

[0037] The pricing optimization model module is configured to obtain the predicted value of the marginal electricity price of the node, and construct the pricing optimization model of each power generation entity with the goal of maximizing the comprehensive benefits of the power generation entity.

[0038] The agent design module is configured to design agents for each power generation entity that can directly output continuous bids, and train the agents using the bid optimization model as the reward function;

[0039] The joint clearing model module is configured to generate bid data for each power generation entity based on the trained agents, and construct a joint clearing model for all power generation entities with the goal of minimizing the total social cost.

[0040] The electricity price deviation calculation module is configured to calculate the actual value of the node marginal electricity price based on the joint clearing model, and to calculate the deviation between the actual value and the predicted value of the node marginal electricity price.

[0041] The closed-loop learning module is configured to determine whether the deviation exceeds a preset threshold. If it does, the actual value of the node marginal electricity price is used as its predicted value and returned to the pricing optimization model module for closed-loop learning. If it does not exceed the threshold, the pricing data generated by each agent is used as the final pricing strategy.

[0042] Thirdly, the present invention provides an electronic device, including a processor and a storage medium;

[0043] The storage medium is used to store instructions;

[0044] The processor is configured to operate according to the instructions to perform the steps according to the method described above.

[0045] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0046] Fifthly, the present invention provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0047] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0048] This invention provides a transaction decision-making method and system under a market joint clearing environment. By constructing a two-layer optimization model under a market joint clearing environment for electricity and ancillary services, namely a bid optimization model and a joint clearing model, in the bid optimization of power generation entities, a multi-entity bid strategy generation algorithm based on reinforcement learning is used to obtain a locally optimal decision. Then, through the iterative feedback and convergence mechanism between the two-layer optimization models, the selfish decisions of power generation entities are coordinated with the safe and economical operation of the power system, ultimately guiding the market to reach or approach an equilibrium state. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the transaction decision-making method under a market joint clearing environment provided in an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram comparing the actual value and the predicted value of the nodal marginal electricity price and their deviation rate provided in an embodiment of the present invention;

[0051] Figure 3 This is a block diagram of a transaction decision-making system under a market joint clearing environment provided in an embodiment of the present invention. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0053] Example 1

[0054] In a first aspect, the present invention provides a transaction decision-making method under a market joint clearing environment, comprising:

[0055] Step S1: Obtain the predicted value of the marginal electricity price of the node, and construct a bidding optimization model for each power generation entity with the goal of maximizing the comprehensive income of the power generation entity.

[0056] In this embodiment, an operating day is taken as a cycle and divided into 96 time periods, each time period being 15 minutes. The predicted curve of the node marginal electricity price for the 96 time periods of the operating day, obtained based on historical data and market analysis, is used as the initial predicted value of the node marginal electricity price.

[0057] The pricing optimization model comprehensively considers revenue from medium- and long-term contract electricity, base electricity revenue, spot market revenue, compensation fees, and ancillary service market revenue. The specific expression is as follows:

[0058]

[0059] In the formula, These are the medium- and long-term electricity revenue and the base electricity revenue of the generating unit in time period t, respectively.

[0060] These represent the day-ahead market revenue and real-time market revenue of the generating unit in the spot electricity market corresponding to time period t; k is the prediction curve of the marginal electricity price at node k.

[0061] These represent the compensation costs and ancillary service market revenue for the generating unit during time period t, respectively.

[0062] These represent the power generation cost, start-up and shutdown cost, and shutdown penalty cost of the generating unit during time period t, respectively. The generating capacity of the unit during time period t;

[0063] Let T be the confidence probability of the predicted curve of the marginal electricity price of the k-th node, where T is the total number of time periods and K is the total number of nodes.

[0064] Specifically, in this embodiment, the revenue from day-ahead electricity bids in the electricity spot market is as follows:

[0065]

[0066] In the formula, For the electricity contracts won in the market before time period t, Let t be the predicted market node electricity price for time period t. The unit's plant power consumption rate;

[0067] The revenue from real-time electricity sales in the spot electricity market is as follows:

[0068]

[0069] In the formula, For the real-time market-winning electricity during time period t, This represents the predicted real-time market node electricity price for time period t.

[0070] The compensation amount is:

[0071]

[0072]

[0073] In the formula, To compensate for the cost of market electricity, To cover operating costs; Compensation for unit startup; The price is subsidized to cover the market electricity cost.

[0074] The revenue of the ancillary services market is:

[0075]

[0076] In the formula, The expected revenue from the unit's participation in the frequency regulation market, For auxiliary service coefficients;

[0077] Unit start-up and shutdown costs and downtime penalty costs:

[0078]

[0079]

[0080] In the formula, The start-up cost of the unit is included in the cost during the start-up operation period; The downtime cost of the unit is included in the cost for the period of downtime; if the user expects to start up, a downtime penalty cost is incurred for the downtime period. Downtime incurs cost penalties Set to a larger value.

[0081] Step S2: Design an agent for each power generation entity that can directly output continuous bids, and train the agent using the bid optimization model as the reward function.

[0082] Specifically, in this embodiment, a deep reinforcement learning algorithm, especially a deep deterministic policy gradient algorithm, is used to construct an environment model that includes market state and unit state. An agent that can directly output continuous bidding actions is designed, and the agent is trained through repeated interaction and learning between itself and the environment model.

[0083] The configuration of the environmental model includes: market rule model configuration, power system model construction, market boundary condition preparation, main cost and prediction model configuration, etc., which are configured based on existing technologies and will not be described in detail in this application.

[0084] During strategy generation, strict consideration must be given to upper and lower price limits and various physical operational constraints, such as:

[0085] Unit output upper and lower limit constraints , The upper and lower limits of the unit's output. This is a start / stop status variable; a value of 1 indicates that the unit is started, and a value of 0 indicates that the unit is stopped.

[0086] Unit output ramp-up rate constraint , These are the upper and lower limits of the climbing rate;

[0087] Minimum start-stop time constraints, state of charge constraints of energy storage units, etc.

[0088] Step S3: Based on the trained agents, generate the bidding data of each power generation entity, and construct a joint clearing model for all power generation entities with the goal of minimizing the total cost of the whole society.

[0089] Specifically, in this embodiment, the joint clearing model is as follows:

[0090]

[0091] In the formula, These represent the operating cost, startup cost, and power generation of the i-th unit during time period t. These represent the positive reserve bid and the awarded positive reserve capacity for the i-th generating unit during time period t, respectively; N is the total number of generating units, and T is the total number of time periods;

[0092] Let $\frac{e}{t}$ be the operating cost and charge / discharge power of the $e$-th energy storage unit during time period $t$. These are the positive reserve bid and the winning positive reserve capacity for the e-th energy storage unit during time period t; This represents the total number of energy storage units.

[0093] Let $\frac{p}{t}$ be the operating cost and power generation / pumping capacity of the $p$-th pumped-storage unit during time period $t$. These are the positive reserve bid and the winning positive reserve capacity of the p-th pumped-storage unit during time period t; This represents the total number of pumped storage units;

[0094] The network flow constraint relaxation penalty factor is used for market clearing optimization. are the forward and reverse power flow relaxation variables for the l-th line, respectively, and NL is the total number of lines; , where are the forward and reverse current relaxation variables for the s-th cross section, and NS is the total number of cross sections.

[0095] During the clearing process, relevant constraints must be strictly met, such as power balance constraints and unit operating characteristic constraints. , For maximum power generation, For coefficient parameters, unit winning bid power constraints , This is due to factors such as backup capacity requirements, network traffic constraints, and ancillary service requirements.

[0096] Step S4: Calculate the actual value of the node marginal electricity price based on the joint clearing model, and calculate the deviation between the actual value and the predicted value of the node marginal electricity price.

[0097] Specifically, in this embodiment, the average deviation rate is used as the deviation index, such as... Figure 2 As shown. Other deviation indices can also be selected in other alternative implementations.

[0098] Step S5: Determine whether the deviation exceeds the preset threshold. If it does, use the actual value of the node's marginal electricity price as its predicted value and return to step S1 for closed-loop learning. If it does not exceed the threshold, use the bidding data generated by each agent as the final bidding strategy.

[0099] In the closed-loop learning process following the return to step S1, the bidding strategy is readjusted and optimized based on the actual value of the node marginal electricity price feedback, forming a new round of decision-making schemes. Steps S1 to S5 are repeated to establish a closed-loop learning mechanism of "strategy generation - market clearing - result feedback - strategy optimization". After multiple iterations, when the deviation meets the requirements, the system is determined to have reached a stable state, and the final optimized bidding strategy is output for actual market bidding.

[0100] Example 2

[0101] This invention provides a transaction decision-making system under a market joint clearing environment, comprising:

[0102] The pricing optimization model module is configured to obtain the predicted value of the marginal electricity price of the node, and construct the pricing optimization model of each power generation entity with the goal of maximizing the comprehensive benefits of the power generation entity.

[0103] The agent design module is configured to design agents for each power generation entity that can directly output continuous bids, and train the agents using the bid optimization model as the reward function;

[0104] The joint clearing model module is configured to generate bid data for each power generation entity based on the trained agents, and construct a joint clearing model for all power generation entities with the goal of minimizing the total cost to the whole society.

[0105] The electricity price deviation calculation module is configured to calculate the actual value of the node marginal electricity price based on the joint clearing model, and to calculate the deviation between the actual value and the predicted value of the node marginal electricity price.

[0106] The closed-loop learning module is configured to determine whether the deviation exceeds a preset threshold. If it does, the actual value of the node's marginal electricity price is used as its predicted value and returned to the pricing optimization model module for closed-loop learning. If it does not exceed the threshold, the pricing data generated by each agent is used as the final pricing strategy.

[0107] like Figure 3 As shown, the trading decision-making system can also be divided into a power generation entity bidding decision layer and a market clearing layer based on the power generation entity and market clearing. Through artificial intelligence algorithms and iterative feedback mechanisms, it aims to maximize the profits of power generation entities in complex multi-market environments. This method coordinates the self-interested decisions of power generation entities with the safe and economical operation of the power system, ultimately guiding the market to reach or approach an equilibrium state.

[0108] Example 3

[0109] Based on the transaction decision-making method provided in Embodiment 1, this embodiment of the invention provides an electronic device, including a processor and a storage medium;

[0110] Storage media are used to store instructions;

[0111] The processor is used to perform operations according to instructions to execute the steps according to the method described above.

[0112] Example 4

[0113] Based on the transaction decision-making method provided in Embodiment 1, this embodiment of the invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0114] Example 5

[0115] Based on the transaction decision-making method provided in Embodiment 1, this embodiment of the invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-described method.

[0116] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0118] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0119] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0120] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A transaction decision-making method under a market joint clearing environment, characterized in that, include: Step S1: Obtain the predicted value of the marginal electricity price of the node, and construct a bidding optimization model for each power generation entity with the goal of maximizing the comprehensive benefits of the power generation entity. Step S2: Design an agent for each power generation entity that can directly output continuous bids, and train the agent using the bid optimization model as the reward function; Step S3: Based on the trained agents, generate the bidding data of each power generation entity, and construct a joint clearing model for all power generation entities with the goal of minimizing the total cost of the whole society. Step S4: Calculate the actual value of the node marginal electricity price based on the joint clearing model, and calculate the deviation between the actual value and the predicted value of the node marginal electricity price; Step S5: Determine whether the deviation exceeds a preset threshold. If it does, use the actual value of the node marginal electricity price as its predicted value and return to step S1 for closed-loop learning. If the bid is not exceeded, the bid data generated by each of the aforementioned agents will be used as the final bid strategy. The pricing optimization model is as follows: In the formula, These are the medium- and long-term electricity revenue and the base electricity revenue of the generating unit in time period t, respectively. These represent the day-ahead market revenue and real-time market revenue of the generating unit in the spot electricity market corresponding to time period t; k is the prediction curve of the marginal electricity price at node k. These represent the compensation costs and ancillary service market revenue for the generating unit during time period t, respectively. These represent the power generation cost, start-up and shutdown cost, and shutdown penalty cost of the generating unit during time period t, respectively. The generating capacity of the unit during time period t; Let T be the confidence probability of the predicted curve of the marginal electricity price of the k-th node, where T is the total number of time periods and K is the total number of nodes. The joint clearing model is as follows: In the formula, These represent the operating cost, startup cost, and power generation of the i-th unit during time period t. These represent the positive reserve bid and the awarded positive reserve capacity for the i-th generating unit during time period t, respectively; N is the total number of generating units, and T is the total number of time periods; Let $\frac{e}{t}$ be the operating cost and charge / discharge power of the $e$-th energy storage unit during time period $t$. These are the positive reserve bid and the winning positive reserve capacity for the e-th energy storage unit during time period t; This represents the total number of energy storage units. Let $\frac{p}{t}$ be the operating cost and power generation / pumping capacity of the $p$-th pumped-storage unit during time period $t$. These are the positive reserve bid and the winning positive reserve capacity of the p-th pumped-storage unit during time period t; This represents the total number of pumped storage units; The network flow constraint relaxation penalty factor is used for market clearing optimization. are the forward and reverse power flow relaxation variables for the l-th line, respectively, and NL is the total number of lines; , where are the forward and reverse current relaxation variables for the s-th cross section, and NS is the total number of cross sections.

2. The transaction decision-making method under a market joint clearing environment according to claim 1, characterized in that, The revenue from day-ahead electricity bids in the aforementioned electricity spot market is: In the formula, For the electricity contracts won in the market before time period t, Let t be the predicted market node electricity price for time period t. The unit's plant power consumption rate; The revenue from real-time market-winning electricity bids in the aforementioned electricity spot market is: In the formula, For the real-time market-winning electricity during time period t, This represents the predicted real-time market node electricity price for time period t.

3. The transaction decision-making method under a market joint clearing environment according to claim 1, characterized in that, The constraints on the agent's generation decision include: The price quotation limits and physical operation constraints include upper and lower limits for unit output, unit output ramp-up rate constraints, minimum start-up and shutdown time constraints, and state of charge constraints for energy storage units.

4. The transaction decision-making method under a market joint clearing environment according to claim 1, characterized in that, The constraints for calculating the actual value of the nodal marginal electricity price based on the joint clearing model include: Constraints include power balance, unit operating characteristics, unit contracted power volume, network power flow, and ancillary service demand.

5. A transaction decision-making system under a market joint clearing environment, characterized in that, The system is configured to perform the steps of the method as described in any one of claims 1-4, the system comprising: The pricing optimization model module is configured to obtain the predicted value of the marginal electricity price of the node, and construct the pricing optimization model of each power generation entity with the goal of maximizing the comprehensive benefits of the power generation entity. The agent design module is configured to design agents for each power generation entity that can directly output continuous bids, and train the agents using the bid optimization model as the reward function; The joint clearing model module is configured to generate bid data for each power generation entity based on the trained agents, and construct a joint clearing model for all power generation entities with the goal of minimizing the total social cost. The electricity price deviation calculation module is configured to calculate the actual value of the node marginal electricity price based on the joint clearing model, and to calculate the deviation between the actual value and the predicted value of the node marginal electricity price. The closed-loop learning module is configured to determine whether the deviation exceeds a preset threshold. If it does, the actual value of the node marginal electricity price is used as its predicted value and returned to the pricing optimization model module for closed-loop learning. If it does not exceed the threshold, the pricing data generated by each agent is used as the final pricing strategy.

6. An electronic device, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-4.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-4.

Images