Generator bidding strategy method and device considering bounded rationality, equipment and medium

Abstract

The invention relates to the technical field of power regulation and control. The invention discloses a power generator bidding strategy method, device and equipment considering bounded rationality and a medium, and the method comprises the steps: constructing a bidding strategy modeling and solving framework integrating behavioral economics and an advanced deep reinforcement learning technology; a power market decision is converted into a mathematical problem containing behavioral psychological and physical constraints by constructing a power market transaction settlement model, an influence model of medium and long term transactions on a bidding strategy of a power supplier in a spot market, a bidding model of the power supplier participating in the power spot market and a power spot market clearing model; a deep reinforcement learning algorithm is trained based on the model, and a power generator bidding strategy is determined based on the trained deep reinforcement learning algorithm, so that an optimal strategy in a continuous quotation space can be efficiently solved in a medium and long term market environment and a spot market environment; therefore, a clearing analysis and mechanism design basis closer to market reality is provided for a power transaction mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power regulation, in particular to a method and device for bidding strategy of power generator considering bounded rationality, equipment and medium. BACKGROUND

[0002] With the deepening of the reform of the electricity market system, the centralized planning mode is gradually transitioning to a competitive market with parallel long-term contracts and spot markets. Power generators lock in part of the electricity and revenue through long-term contracts such as monthly and annual contracts, and then compete for the remaining electricity in the spot market, thereby balancing between maximizing revenue and hedging risks. Traditional research is generally based on the expected utility theory, assuming that power generators are completely rational and have complete information, and using game theory equilibrium, stochastic optimization or robust optimization methods to derive the optimal offer. However, in the real market, the decision-making of power generators is influenced by multiple factors such as cognitive ability, historical profit and loss, emotions and market noise, showing obvious characteristics of bounded rationality, which leads to deviations in their behavior from classical theory, such as excessive loss avoidance, anchoring on existing gains, and unbalanced trade-offs when facing uncertainty.

[0003] Therefore, there is an urgent need for a method that can accurately depict the bounded rationality of power generators and efficiently solve the optimal strategy in the continuous offer space under the two-stage market environment of long-term and spot markets, thereby providing power trading agencies with more market-actual clearing analysis and mechanism design basis. SUMMARY

[0004] In view of the above, the embodiments of the present application provide a method and device for bidding strategy of power generator considering bounded rationality, equipment and medium, aiming to solve the above problems or at least partially solve the above problems.

[0005] In a first aspect, the present application provides a method for bidding strategy of power generator considering bounded rationality, comprising: According to the transaction electricity and price of the power generator in the long-term market and the spot market, a power market transaction settlement model is constructed to obtain the electricity selling revenue of the power generator in the long-term market and the spot market; According to the electricity selling revenue of the power generator in the long-term market and the spot market, an influence model of long-term transaction on the bidding strategy of the power generator in the spot market is constructed to obtain the subjective value of the power generator for profit obtained by clearing the spot market; According to the subjective value of the power generator for profit obtained by clearing the spot market, a bidding model of the power generator participating in the power spot market is constructed to estimate the bidding strategy of the power generator; Based on the pre-set system boundary conditions and the declaration information of the power generator, a power spot market clearing model is constructed to verify the estimated bidding strategy of the power generator; The deep reinforcement learning algorithm is trained according to the power market transaction settlement model, the influence model of the mid-to-long-term transaction on the bidding strategy of the power generator in the spot market, the bidding model of the power generator participating in the power spot market, and the power spot market clearing model, and the bidding strategy of the power generator is determined based on the trained deep reinforcement learning algorithm.

[0006] The power market transaction settlement model includes a full energy settlement algorithm and a deviation settlement algorithm. The power market transaction settlement model is constructed according to the transaction energy and price of the power generator in the mid-to-long-term market and the spot market, and the power generator's electricity selling revenue in the mid-to-long-term market and the spot market is obtained, including: The power generator's electricity selling revenue in the mid-to-long-term market is obtained based on the full energy settlement algorithm according to the transaction energy and price of the power generator in the mid-to-long-term market. The power generator's electricity selling revenue in the spot market is obtained based on the deviation settlement algorithm according to the transaction energy and price of the power generator in the spot market.

[0007] The influence model of the mid-to-long-term transaction on the bidding strategy of the power generator in the spot market is constructed according to the power generator's electricity selling revenue in the mid-to-long-term market and the spot market, and the subjective value of the power generator's profit obtained from the spot market clearing is obtained, including: The transaction cost of the power generator in the mid-to-long-term market and the spot market is obtained based on the pre-set power generation cost coefficient of the power generator and the transaction energy of the power generator in the mid-to-long-term market and the spot market. The transaction profit of the power generator in the mid-to-long-term market and the spot market is obtained according to the power generator's electricity selling revenue in the mid-to-long-term market and the spot market and the transaction cost of the power generator in the mid-to-long-term market and the spot market. The psychological reference point of the power generator for the profit of the spot market is obtained according to the transaction profit of the power generator in the mid-to-long-term market based on the pre-set psychological reference point of the power generator for the total profit. The subjective value of the power generator's profit obtained from the spot market clearing is obtained according to the psychological reference point of the power generator for the profit of the spot market and the transaction profit of the power generator in the spot market.

[0008] The subjective value of the power generator's profit obtained from the spot market clearing is obtained according to the psychological reference point of the power generator for the profit of the spot market and the transaction profit of the power generator in the spot market, including: When the generator's trading profit in the spot market is greater than or equal to the generator's psychological reference point for spot market profit, the subjective value of the generator's profit from clearing the spot market is determined based on the generator's risk preference coefficient in the profit area, the generator's trading profit in the spot market, and the generator's psychological reference point for spot market profit. When the generator's trading profit in the spot market is less than the generator's psychological reference point for spot market profit, the subjective value of the generator's profit from clearing the spot market is determined based on the generator's risk preference coefficient and loss aversion coefficient in the loss zone, the generator's trading profit in the spot market, and the generator's psychological reference point for spot market profit.

[0009] For example, the objective function in the bidding model for the power generator participating in the electricity spot market is to maximize the subjective value of the power generator's profit from clearing the spot market based on the bid price of the power generator in the spot market. The constraint condition in the bidding model for the power generator participating in the electricity spot market is the constraint price condition of the power generator's bid price in the spot market.

[0010] For example, the objective function in the electricity spot market clearing model is to minimize the total transaction costs of all power generators in the spot market; The constraints in the electricity spot market clearing model include node power balance constraints, unit output upper and lower limit constraints, unit ramping constraints, line power flow constraints, and node voltage phase angle constraints.

[0011] For example, the deep reinforcement learning algorithm includes a state space, an action space, an immediate reward, and an experience pool; The state space includes the target time period, the electricity volume and price of power generators in the spot market during the target time period; The action space includes the estimated bidding strategy of power generators determined based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategy of power generators in the spot market, and the bidding model of power generators participating in the electricity spot market. The immediate reward includes the subjective value of the generator's profit from clearing the spot market, obtained based on the estimated generator bidding strategy and the impact model of the medium- and long-term transactions on the generator bidding strategy in the spot market. The training of the deep reinforcement learning algorithm includes: The input of the deep reinforcement learning algorithm is determined based on the state space, and the output of the deep reinforcement learning algorithm is determined based on the action space. The output of the deep reinforcement learning algorithm is input into the electricity spot market clearing model to obtain four-tuple data. The four-tuple data is placed into an experience pool, and a set of four-tuple data is randomly selected from the experience pool to train the deep reinforcement learning algorithm. The four-tuple data includes the state corresponding to the target time period, the estimated generator bidding strategy obtained based on the state corresponding to the target time period, the immediate reward corresponding to the estimated generator bidding strategy, and the state corresponding to the next time period.

[0012] Secondly, this application provides a power generator bidding strategy apparatus that considers bounded rationality, comprising: The settlement module is used to construct a power market transaction settlement model based on the electricity volume and price traded by power generators in the medium- and long-term market and the spot market, and to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market. The bounded rationality module is used to construct a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, and to obtain the subjective value of the profit obtained by power generators from clearing the spot market. The bidding module is used to construct a bidding model for power generators to participate in the electricity spot market based on the subjective value of power generators to obtain profits by clearing the spot market, and to predict the bidding strategies of power generators. The clearing module is used to construct a power spot market clearing model based on pre-set system boundary conditions and power generator declaration information, and to verify the estimated bidding strategies of the power generators. The deep reinforcement learning module is used to train the deep reinforcement learning algorithm based on the electricity market transaction settlement model, the impact model of the medium- and long-term transactions on the bidding strategies of power generators in the spot market, the bidding model of power generators participating in the electricity spot market, and the electricity spot market clearing model, and to determine the bidding strategies of power generators based on the trained deep reinforcement learning algorithm.

[0013] Thirdly, this application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the generator bidding strategy method considering bounded rationality as described in the first aspect.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method for considering bounded rationality in the bidding strategy of power generators as described in the first aspect.

[0015] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: This application constructs a bidding strategy modeling and solution framework that integrates behavioral economics and advanced deep reinforcement learning technology. By building a power market transaction settlement model, a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market, a bidding model for power generators participating in the power spot market, and a power spot market clearing model, power market decision-making is transformed into a mathematical problem containing behavioral, psychological, and physical constraints. This framework can accurately characterize the bounded rationality of power generators. Furthermore, based on the above models, a deep reinforcement learning algorithm is trained, and the bidding strategy of power generators is determined based on the trained deep reinforcement learning algorithm. This framework can efficiently solve for the optimal strategy in the continuous price space under both medium- and long-term and spot market environments, thereby providing power trading institutions with a clearing analysis and mechanism design basis that is closer to market reality. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of an application environment for a power generator bidding strategy method that considers bounded rationality in one embodiment of the present invention. Figure 2 This is a flowchart illustrating a power generator bidding strategy method considering bounded rationality in one embodiment of the present invention. Figure 3 yes Figure 2 A schematic diagram of a specific implementation method for step S10; Figure 4 yes Figure 2 A schematic diagram of a specific implementation method for step S20; Figure 5 This is a schematic diagram of a power generation bidding strategy device considering bounded rationality in one embodiment of the present invention. Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 7 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the term "comprising" and its variations should be interpreted as open-ended terms meaning "including but not limited to."

[0019] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0020] As mentioned earlier, current medical guidance services are provided through manual processes or web pages or apps with fixed symptom and disease options, all of which have limitations. To address this technical issue, this application provides a method for power generator bidding strategies that considers bounded rationality.

[0021] The power generator bidding strategy method considering bounded rationality provided in this invention can be applied to, for example... Figure 1In this application environment, the device communicates with the server via a network. The server can use the device to construct a model of the impact of medium- and long-term transactions on the bidding strategies of generators in the spot market, based on the generators' revenue from electricity sales in the medium- and long-term market and the spot market. This model yields the subjective value of generators' profits from clearing the spot market. Based on this subjective value, the server constructs a bidding model for generators participating in the spot market, predicting their bidding strategies. Using pre-set system boundary conditions and generator application information, a spot market clearing model is constructed to verify the predicted generator bidding strategies. Finally, a deep reinforcement learning algorithm is trained using the electricity market transaction settlement model, the model of the impact of medium- and long-term transactions on generator bidding strategies in the spot market, the generator bidding model, and the spot market clearing model. The trained deep reinforcement learning algorithm is then used to determine the generator bidding strategies. This application constructs a bidding strategy modeling and solution framework that integrates behavioral economics and advanced deep reinforcement learning technology. By building a power market transaction settlement model, a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market, a bidding model for power generators participating in the power spot market, and a power spot market clearing model, power market decision-making is transformed into a mathematical problem containing behavioral, psychological, and physical constraints. This framework can accurately characterize the bounded rationality of power generators. Furthermore, based on the above models, a deep reinforcement learning algorithm is trained, and the bidding strategy of power generators is determined based on the trained deep reinforcement learning algorithm. This framework can efficiently solve for the optimal strategy in the continuous price space under both medium- and long-term and spot market environments, thereby providing power trading institutions with a clearing analysis and mechanism design basis that is closer to market reality.

[0022] The device side can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server side can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will now be described in detail through specific embodiments.

[0023] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a power generator bidding strategy method considering bounded rationality provided in an embodiment of the present invention includes the following steps: S10: Based on the electricity volume and prices traded by power generators in the medium- and long-term market and the spot market, construct an electricity market transaction settlement model to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market. The electricity volume and prices traded by power generators in the medium- and long-term market and the spot market are obtained through observation. Step S10 provides the mathematical basis for subsequent calculations of the revenue of power generators in different markets and is the starting point for profit calculation.

[0024] In one embodiment, such as Figure 3The electricity market transaction settlement model described in step S10 includes a total electricity settlement algorithm and a deviation settlement algorithm; Step S10 constructs an electricity market transaction settlement model based on the electricity volume and price traded by power generators in the medium- and long-term market and the spot market, to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market, including: S11: Based on the electricity volume and price traded by the power generator in the medium- and long-term market, and using the full electricity volume settlement algorithm, obtain the electricity sales revenue of the power generator in the medium- and long-term market;

[0025] in, For power generation Electricity sales revenue in the medium to long term market; for Time-based delivery e-commerce Electricity volume traded in the medium- and long-term markets; They are respectively Time-based delivery e-commerce Trading prices in the medium to long term market.

[0026] S12: Based on the electricity volume and price traded by the power generator in the spot market, and using the aforementioned deviation settlement algorithm, obtain the power generator's revenue from electricity sales in the spot market.

[0027]

[0028] in, For power generation Revenue from selling electricity in the spot market; for Time-based delivery e-commerce Trading volume in the spot market; for Time-based delivery e-commerce Node The spot market transaction price.

[0029] S20: Based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, a model is constructed to assess the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market, thus obtaining the subjective value of profit generation from spot market clearing for power generators. Step S20 introduces prospect theory, simulating the bounded rationality behavior of power generators based on prospect theory, and establishing a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market. This incorporates behavioral economics into decision-making, replacing the traditional rational man assumption, and explaining psychological biases such as the excessive sensitivity of power generators to losses.

[0030] In one embodiment, such as Figure 4Step S20, as shown, constructs a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market. This model yields the subjective value of the profit gained by power generators from clearing the spot market, including: S21: Based on the pre-set power generation cost coefficient of the power generator and the transaction volume of the power generator in the medium- and long-term market and the spot market, obtain the transaction cost of the power generator in the medium- and long-term market and the spot market; Specifically, based on the pre-set power generation cost coefficient of the power generator and the electricity volume traded by the power generator in the medium- and long-term market, the transaction cost of the power generator in the medium- and long-term market is obtained through the following formula:

[0031] in, , , Generators The coefficients of each term in the secondary power generation cost function are preset; for Time-based delivery e-commerce Electricity volume traded in the medium- and long-term markets; For power generators, transaction costs in the medium to long term market.

[0032] Specifically, based on the pre-set power generation cost coefficient of the power generator and the electricity volume traded by the power generator in the spot market, the transaction cost of the power generator in the spot market is obtained through the following formula:

[0033] in, , , Generators The cost coefficients of each item in the secondary power generation cost function are preset; for Time-based delivery e-commerce Electricity volume traded in the medium- and long-term markets; for Time-based delivery e-commerce Trading volume in the spot market; For transaction costs of generators in the spot market.

[0034] S22: Based on the electricity sales revenue of the power generator in the medium- and long-term market and the spot market and the transaction costs of the power generator in the medium- and long-term market and the spot market, obtain the transaction profit of the power generator in the medium- and long-term market and the spot market; Specifically, based on the electricity sales revenue and transaction costs of the power generator in the medium- and long-term market, the transaction profit of the power generator in the medium- and long-term market is obtained through the following formula:

[0035] in, For power generators' transaction profits in the medium to long term market; For power generation Electricity sales revenue in the medium to long term market; For power generators, transaction costs in the medium to long term market.

[0036] Specifically, based on the electricity sales revenue and transaction costs of the power generator in the spot market, the transaction profit of the power generator in the spot market is obtained through the following formula:

[0037] in, For the profit of power generators in the spot market; For power generation Revenue from selling electricity in the spot market; For transaction costs of generators in the spot market.

[0038] S23: Based on the pre-set psychological reference point for the generator's total profit, and according to the generator's trading profits in the medium- and long-term market, the generator's psychological reference point for the spot market profit is obtained through the following formula:

[0039] in, This serves as a psychological reference point for generators regarding profits in the spot market. For power generation A psychological reference point for total profit is pre-set; For the transaction profits of power generators in the medium and long term market.

[0040] S24: Based on the generator's psychological reference point for spot market profits and the generator's trading profits in the spot market, the subjective value of the generator's profit from clearing the spot market is obtained.

[0041] Specifically, based on power generation operators Psychological reference point for spot market profits We construct a prospect theory value function to simulate the subjective perception of decision outcomes.

[0042] Specifically, step S24, based on the generator's psychological reference point for spot market profits and the generator's trading profits in the spot market, derives the subjective value of the generator's profit from clearing the spot market, including: S241: When the generator's trading profit in the spot market Greater than or equal to the generator's psychological reference point for spot market profits At that time, based on the pre-set risk preference coefficient of the generator in the revenue region. The transaction profits of the power generator in the spot market And the psychological reference point for spot market profits for the aforementioned power generators. The subjective value of the generator's profit from clearing the spot market is determined. S242: When the generator's trading profit in the spot market Less than the psychological reference point for spot market profits of the generators. At that time, based on the pre-set risk preference coefficient of the power generator in the loss area Loss aversion coefficient The transaction profits of the power generator in the spot market And the psychological reference point for spot market profits for the aforementioned power generators. The subjective value of the generator's profit from clearing the spot market is determined.

[0043] Specifically, step S24 is described by the following formula:

[0044] in, The subjective value of power generators in clearing the spot market to obtain profits; For the profit of power generators in the spot market; This serves as a psychological reference point for generators regarding profits in the spot market. Risk appetite coefficient of power generators in revenue regions; The risk appetite coefficient of power generators in loss-making areas; This represents the loss aversion coefficient.

[0045] The model of the impact of medium- and long-term transactions on the bidding strategies of generators in the spot market in step S20 makes the reward function non-smooth and non-convex, which solves the technical problem that traditional optimization algorithms are difficult to solve.

[0046] S30: Based on the subjective value of power generators in obtaining profits from clearing the spot market, construct a bidding model for power generators to participate in the electricity spot market and predict their bidding strategies.

[0047] The model of the impact of medium- and long-term transactions on the bidding strategies of generators in the spot market is a bounded rationality model. Step S30 transforms the bounded rationality model of step S20 into an optimizable mathematical objective.

[0048] Specifically, the objective function in the bidding model for power generators participating in the electricity spot market is to maximize the subjective value of the power generator's profit from clearing the spot market, based on the power generator's bid price in the spot market, as shown in the following formula:

[0049] Among them, decision variables For power generation Bidding prices in the spot market.

[0050] The constraint condition in the bidding model for power generators participating in the electricity spot market is the constraint price condition of the power generator's bid price in the spot market, as shown in the following formula:

[0051] in, For power generation Bidding prices in the spot market; For power generation The upper limit of the bid price is set in advance; For power generation The cost coefficient of the secondary power generation cost function is also used to characterize the lower limit of the bid price, and is preset.

[0052] S40: Based on pre-set system boundary conditions and generator bid information, construct a power spot market clearing model to verify the estimated generator bidding strategies. The system boundary conditions, i.e., the constraints in the power spot market clearing model, include node power balance constraints, unit output upper and lower limit constraints, unit ramping constraints, line power flow constraints, and node voltage phase angle constraints. Generator bid information refers to the generator's bid prices in the spot market.

[0053] Step S40 is used to define the environmental boundaries of power generator decisions to ensure that the strategy complies with grid security requirements.

[0054] Specifically, the objective function in the electricity spot market clearing model described in step S40 is to minimize the total transaction costs for all power generators in the spot market, which can be expressed by the following formula:

[0055] in, Total number of generators; for Time-based delivery e-commerce Trading volume in the spot market; For power generation Bidding prices in the spot market; , Generators The cost coefficients of the secondary power generation cost function are preset.

[0056] Specifically, the constraints in the electricity spot market clearing model include node power balance constraints, unit output upper and lower limit constraints, unit ramping constraints, line power flow constraints, and node voltage phase angle constraints.

[0057] Specifically, the node power balance constraint can be expressed by the following formula:

[0058] in, For connecting nodes and nodes Branch admittance; For nodes During the period The voltage phase angle; For nodes During the period The voltage phase angle; for Time-based delivery e-commerce Trading volume in the spot market; for Time-of-use load Electricity demand in the spot market.

[0059] The upper and lower limits of the unit's output constraints can be expressed by the following formula:

[0060] in, For power generation The maximum output capacity of the generator set is preset; For power generation The lower limit of the generator set's output is preset; for Time-based delivery e-commerce Electricity traded in the spot market.

[0061] The unit's ramp-up constraint can be expressed by the following formula:

[0062] in, For power generation The upper limit of the ramp rate of the generator set is preset; For power generation The maximum downhill gradient rate of the generator set is preset; for Time-based delivery e-commerce Trading volume in the spot market; for -1 period for shipping e-commerce Electricity traded in the spot market.

[0063] The power flow constraints of a line can be expressed by the following formula:

[0064] in, For connecting nodes and nodes The upper limit of the tributary transmission capacity; For connecting nodes and nodes Branch admittance; For nodes During the period The voltage phase angle; For nodes During the period The voltage phase angle; For nodes A set of connected nodes.

[0065] The node voltage phase angle constraint can be expressed by the following formula:

[0066] in, For nodes During the period The upper limit of the voltage phase angle is preset; For nodes During the period The lower limit of the voltage phase angle is preset; For nodes During the period The voltage phase angle.

[0067] S50: The deep reinforcement learning algorithm is trained based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategy of power generators in the spot market, the bidding model of power generators participating in the electricity spot market, and the electricity spot market clearing model. The bidding strategy of power generators is determined based on the trained deep reinforcement learning algorithm.

[0068] In one specific embodiment, the deep reinforcement learning algorithm is the DDPG (Deep Deterministic Policy Gradient) deep reinforcement learning algorithm. DDPG is a deep reinforcement learning algorithm for solving continuous action space problems, combining deterministic policy gradient and DQN techniques, and is suitable for continuous control tasks.

[0069] Steps S10 to S40 together construct a high-dimensional, nonlinear, and constrained decision-making environment, while the DDPG deep reinforcement learning algorithm in step S50 is the solver for the decision-making environment.

[0070] In one specific embodiment, step S50 includes: S51: Construct the state space, action space, and immediate reward for the DDPG reinforcement learning algorithm.

[0071] The deep reinforcement learning algorithm includes a state space, an action space, an immediate reward, and an experience pool; The state space This includes the target time period, the electricity volume traded by power generators in the spot market during the target time period, and the corresponding prices. Specifically, the target time period is the period observed by the agent in the spot market environment, and the state space includes the period observed by the agent in the spot market environment, the electricity volume traded by power generators in the spot market, and the nodal price. This can be expressed by the following formula:

[0072] in, t For the target time period; for Time-based delivery e-commerce Trading volume in the spot market; for Time-based delivery e-commerce Node The spot market transaction price.

[0073] The action space includes, based on the electricity market transaction settlement model, the medium- and long-term transactions, and the power generation and investment in the spot market. The bidding strategy of the power generator is estimated based on the impact model of the bidding strategy and the bidding model for the power generator's participation in the electricity spot market. Specifically, the estimated bidding strategy of the power generator is the bid price in the spot market when the power generator maximizes its subjective value of clearing the spot market to obtain profits. The action space is the bidding strategy of the power generator in the spot market, which can be expressed by the following formula:

[0074] The instant reward Including the model of the impact of the estimated generator bidding strategy and the medium- and long-term transaction on the generator bidding strategy in the spot market, the subjective value of the generator's profit from spot market clearing, obtained by the generator, can be expressed by the following formula:

[0075] S52: Establishing the Policy-Value Network Architecture for the DDPG Reinforcement Learning Algorithm. The Policy-Value Network architecture includes a policy network and a value network. The policy network is used to control the agent's decisions, particularly in the state... Down Output Action Value networks are used to analyze state-action pairs. Scoring is used to evaluate policy networks and measure state-action pairs. The advantages and disadvantages of the policy network. The goal of the policy network is to improve the score of the value network on state-action pairs in order to optimize the agent's policy. Its objective function is as follows:

[0076] in, For value networks; For value network parameters; For policy networks; For the output action of the policy network ; For policy network parameters; This is an experience replay pool.

[0077] The policy network updates its parameters using gradient descent, as shown in the following formula:

[0078] Among them, the policy network is Updated at a frequency of [frequency] Pre-set.

[0079] S53: Establishing the Target Policy-Target Value Network Architecture for the DDPG Reinforcement Learning Algorithm. The Target Policy-Target Value Network architecture includes a Target Policy Network and a Target Value Network. The Target Policy Network is used to control the agent's decisions, in the state... Down Output Action ,state The market clearing is achieved by inputting the action in step S52 into the electricity spot market clearing model. The target value network is used to analyze state-action pairs. Scoring. The target value network is used to evaluate the target policy network, measuring the state-action pairs. The advantages and disadvantages.

[0080] The goal of the target policy-target value network architecture is to improve the accuracy of policy network evaluation and help the policy network update the policy in the correct direction. The target policy-target value network architecture updates based on the temporal difference algorithm, and its loss function is shown in equation (22), representing the value network's response to state-action pairs. The smaller the difference between the score and the target value of the target value network, the better.

[0081]

[0082] in, For value networks, state-action pairs Rate; The target value of the target value network.

[0083] The target value of the target value network is calculated as follows:

[0084] in, For the target value network; Target value network parameters; For the target policy network; For the target policy network parameters; Discount factors are pre-set; the target value network is based on... Updated at a frequency of [frequency] Pre-set.

[0085] S54: Training the DDPG deep reinforcement learning algorithm. This includes: determining the input of the deep reinforcement learning algorithm based on the state space, determining the output of the deep reinforcement learning algorithm based on the action space, inputting the output of the deep reinforcement learning algorithm into the electricity spot market clearing model to obtain four-tuple data, placing the four-tuple data into an experience pool, and randomly selecting a set of four-tuple data from the experience pool to train the deep reinforcement learning algorithm; wherein, the four-tuple data includes the state corresponding to the target time period. Based on the state corresponding to the target time period, the estimated bidding strategy for power generators is obtained. , Estimating the immediate rewards corresponding to the bidding strategy of the power generation company The state corresponding to the next time period , can be represented as .

[0086] In a specific embodiment, the process of training the DDPG reinforcement learning algorithm to solve the generator bidding strategy is as follows: 1) Initialize the agent's policy network parameters and value network parameters ; 2) Initialize the target policy network parameters of the agent. and target value network parameters ; 3) for to do, which means the number of rounds of interaction between the agent and the environment is 1 to 10. M ; 4) for to do, which is the agent's observation period. t From 1 to T ; 5) Intelligent agents observe the current state of the spot market environment. ; 6) Actions output by the agent policy network This step is based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategies of power generators in the spot market, and the bidding model of power generators participating in the electricity spot market.

[0087] 7) Perform market clearing through the aforementioned electricity spot market clearing model and output rewards. and the status of the next period Return to the agent; 8) Obtain the quadruple And store it in the experience replay pool D; steps 5) to 8) are executed repeatedly, so there are multiple quadruple data in the experience pool.

[0088] 9) if Then, that is, if the number of interactions between the agent and the environment is divisible by the update frequency. Update the value network parameters; 10) Randomly select quadruplets from the experience replay pool ; 11) Calculate the target policy network output and target value network output

[0089] 12) Update value network parameters ;in The value network update coefficients are preset. 13) if Then, that is, if the number of interactions between the agent and the environment is divisible by the update frequency. Update the policy network parameters; 14) Update policy network parameters ;in The policy network update coefficients are preset; 15) end if 16) end if 17) Soft update of target value network parameters ;in The soft update factor is preset. 18) Soft update target strategy network parameters ; 19) end for 20) end for The final training yields an optimal policy network in the DDPG deep reinforcement learning algorithm, based on the state of the input observation period. You can then obtain the corresponding bidding strategy. .

[0090] The DDPG deep reinforcement learning algorithm overcomes the limitations of traditional optimization methods by handling continuous actions (bidding) and high-dimensional constraints (hill climbing, current flow). Steps S10 to S40 define the environment, state, action, and reward for the DDPG deep reinforcement learning algorithm, which then outputs the optimal strategy that conforms to market rules and the psychology of generators through interactive learning.

[0091] As can be seen, in the above scheme, this application constructs a bidding strategy modeling and solution framework that integrates behavioral economics and advanced deep reinforcement learning technology. By constructing a power market transaction settlement model, a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market, a bidding model for power generators participating in the power spot market, and a power spot market clearing model, the power market decision-making is transformed into a mathematical problem containing behavioral, psychological, and physical constraints. This can accurately characterize the bounded rationality of power generators. Furthermore, based on the above models, a deep reinforcement learning algorithm is trained, and the bidding strategy of power generators is determined based on the trained deep reinforcement learning algorithm. This can efficiently solve the optimal strategy in the continuous bidding space under both medium- and long-term and spot market environments, thereby providing power trading institutions with a clearing analysis and mechanism design basis that is closer to the actual market.

[0092] In step S20, this application introduces prospect theory from behavioral economics into electricity market analysis, using value functions, loss aversion coefficients, and reference points to characterize the risk preferences and decision weights of market participants. Simultaneously, by optimizing the reference points based on the spot market and the medium- to long-term market, it can systematically characterize the transmission mechanism of prospect effects under the coupling of the medium- to long-term market and the spot market, avoiding the assumption that generator bidding space is discrete or linear, and accurately reflecting the true strategy space under continuous bidding, multi-period interaction, and complex constraints.

[0093] On the other hand, the DDPG deep reinforcement learning algorithm in this application, with its closed-loop iterative capability of "trial and error-feedback-update," can mine complex state-action mappings in scenarios such as unit start-up and shutdown, energy storage scheduling, and flexible load incentives without the need for explicit physical models. It can be applied to fields such as power dispatching, energy storage control, and demand response. In particular, it is well-suited for electricity spot bidding tasks, addressing the following challenges: First, decision variables are naturally continuous and high-dimensional, and discretization of actions both drastically increases the search space and reduces bidding accuracy; second, the payoff function exhibits sharp multi-peak characteristics with triggering conditions such as capacity boundaries and price ceilings, leading to value estimation bias and perturbation of strategy updates; third, multi-agent parallel learning makes the environment non-stationary, and gradient propagation and exploration are hindered in high-dimensional coupled states, resulting in decreased convergence speed and repeatability; fourth, electricity market decisions focus on tail risk and profit-loss asymmetry, and simply maximizing expected returns cannot reflect loss aversion or risk premium. Research has attempted to incorporate behavioral economics descriptions of preferences, but prospect theory's value function suffers from asymmetric, heavy-tailed, and locally sensitive reward distributions. Furthermore, the reference point dynamically shifts with returns, leading to a non-convex and non-smooth objective function, a sliding policy space over time, and a tendency for the learning process to oscillate and fall into local maxima. This application's deep reinforcement learning framework mitigates problems such as limited action expression, value assessment bias, insufficient risk measurement, and low convergence efficiency when dealing with bounded rationality bidding in the electricity market.

[0094] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0095] In one embodiment, a power generator bidding strategy apparatus considering bounded rationality is provided, which corresponds one-to-one with the power generator bidding strategy method considering bounded rationality in the above embodiments. For example... Figure 5 As shown, the generator bidding strategy device considering bounded rationality includes a settlement module 101, a bounded rationality module 102, a bidding module 103, a clearing module 104, and a deep reinforcement learning module 105. Detailed descriptions of each functional module are as follows: Settlement module 101 is used to construct an electricity market transaction settlement model based on the electricity volume and price traded by power generators in the medium- and long-term market and the spot market, and to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market. Bounded rationality module 102 is used to construct a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, and to obtain the subjective value of the profit obtained by power generators from clearing the spot market. Bidding module 103 is used to construct a bidding model for power generators to participate in the electricity spot market based on the subjective value of power generators to obtain profits from clearing the spot market, and to predict the bidding strategies of power generators. The clearing module 104 is used to construct a power spot market clearing model based on pre-set system boundary conditions and power generator declaration information, and to verify the estimated bidding strategy of the power generators. The deep reinforcement learning module 105 is used to train the deep reinforcement learning algorithm based on the electricity market transaction settlement model, the impact model of the medium and long-term transactions on the bidding strategy of power generators in the spot market, the bidding model of power generators participating in the electricity spot market, and the electricity spot market clearing model, and to determine the bidding strategy of power generators based on the trained deep reinforcement learning algorithm.

[0096] Specifically, the electricity market transaction settlement model in the settlement module 101 includes a full-volume settlement algorithm and a deviation settlement algorithm; the settlement module 101 is also used to obtain the electricity sales revenue of the power generator in the medium and long-term market based on the full-volume settlement algorithm, according to the power generator's transaction volume and price in the medium and long-term market. Based on the electricity volume and price traded by the power generator in the spot market, and using the aforementioned deviation settlement algorithm, the power generator's revenue from electricity sales in the spot market is obtained.

[0097] Specifically, the bounded rationality module 102 is also used to obtain the transaction costs of the power generator in the medium- and long-term market and the spot market based on the pre-set power generation cost coefficient of the power generator and the transaction volume of the power generator in the medium- and long-term market and the spot market. The transaction profit of the power generator in the medium- and long-term market and the spot market is obtained based on the electricity sales revenue and transaction costs of the power generator in the medium- and long-term market and the spot market. Based on the pre-set psychological reference point of the generator for total profit, and according to the generator's trading profit in the medium and long term market, the generator's psychological reference point for spot market profit is obtained. Based on the generator's psychological reference point for spot market profits and the generator's trading profits in the spot market, the subjective value of the generator's profit from clearing the spot market is obtained.

[0098] Specifically, the bounded rationality module 102 is also used to determine the subjective value of the generator's profit from clearing the spot market when the generator's trading profit in the spot market is greater than or equal to the generator's psychological reference point for the spot market profit, based on the generator's risk preference coefficient in the profit area, the generator's trading profit in the spot market, and the generator's psychological reference point for the spot market profit. When the generator's trading profit in the spot market is less than the generator's psychological reference point for spot market profit, the subjective value of the generator's profit from clearing the spot market is determined based on the generator's risk preference coefficient and loss aversion coefficient in the loss zone, the generator's trading profit in the spot market, and the generator's psychological reference point for spot market profit.

[0099] Specifically, the bidding module 103 is also used in the bidding model for the power generator to participate in the electricity spot market, where the objective function is to maximize the subjective value of the power generator's profit from clearing the spot market based on the power generator's bid price in the spot market. The constraint condition in the bidding model for the power generator participating in the electricity spot market is the constraint price condition of the power generator's bid price in the spot market.

[0100] Specifically, the clearing module 104 is also used in the electricity spot market clearing model to minimize the total transaction cost of all power generators in the spot market; The constraints in the electricity spot market clearing model include node power balance constraints, unit output upper and lower limit constraints, unit ramping constraints, line power flow constraints, and node voltage phase angle constraints.

[0101] Specifically, the deep reinforcement learning module 105 is also used in the deep reinforcement learning algorithm, which includes a state space, an action space, an immediate reward, and an experience pool. The state space includes the target time period, the electricity volume and price of power generators in the spot market during the target time period; The action space includes the estimated bidding strategy of power generators determined based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategy of power generators in the spot market, and the bidding model of power generators participating in the electricity spot market. The immediate reward includes the subjective value of the generator's profit from clearing the spot market, obtained based on the estimated generator bidding strategy and the impact model of the medium- and long-term transactions on the generator bidding strategy in the spot market. The training of the deep reinforcement learning algorithm includes: The input of the deep reinforcement learning algorithm is determined based on the state space, and the output of the deep reinforcement learning algorithm is determined based on the action space. The output of the deep reinforcement learning algorithm is input into the electricity spot market clearing model to obtain four-tuple data. The four-tuple data is placed into an experience pool, and a set of four-tuple data is randomly selected from the experience pool to train the deep reinforcement learning algorithm. The four-tuple data includes the state corresponding to the target time period, the estimated generator bidding strategy obtained based on the state corresponding to the target time period, the immediate reward corresponding to the estimated generator bidding strategy, and the state corresponding to the next time period.

[0102] This invention provides a device for power generator bidding strategies that considers bounded rationality. It constructs a bidding strategy modeling and solution framework that integrates behavioral economics and advanced deep reinforcement learning technology. By building a power market transaction settlement model, a model of the impact of medium- and long-term transactions on power generator bidding strategies in the spot market, a bidding model for power generators participating in the power spot market, and a power spot market clearing model, power market decisions are transformed into mathematical problems involving behavioral, psychological, and physical constraints. This accurately characterizes the bounded rationality of power generators. Furthermore, based on the above models, a deep reinforcement learning algorithm is trained, and the trained deep reinforcement learning algorithm determines the power generator bidding strategy. This can efficiently solve for the optimal strategy in the continuous price space under both medium- and long-term and spot market conditions, thus providing power trading institutions with a clearing analysis and mechanism design basis that is closer to market reality.

[0103] Specific limitations regarding the device for considering bounded rationality in power generator bidding strategies can be found in the limitations of the method for considering bounded rationality in the above text, and will not be repeated here. Each module in the aforementioned device for considering bounded rationality in power generator bidding strategies can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0104] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used for communication with external devices via a network connection. When the computer program is executed by the processor, it implements server-side functions or steps of a power generator bidding strategy method considering bounded rationality.

[0105] In one embodiment, a computer device is provided, which may be a device terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements functions or steps on the device side, considering a bounded rationality-based bidding strategy method for generators.

[0106] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Based on the electricity volume and price traded by power generators in the medium- and long-term market and the spot market, a power market transaction settlement model is constructed to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market. Based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, a model is constructed to assess the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market, thereby obtaining the subjective value of power generators' profits from clearing the spot market. Based on the subjective value of power generators in obtaining profits from clearing the spot market, a bidding model for power generators to participate in the electricity spot market is constructed to predict their bidding strategies. Based on pre-set system boundary conditions and power generator application information, a power spot market clearing model is constructed to verify the predicted bidding strategies of the power generators. The deep reinforcement learning algorithm is trained based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategies of power generators in the spot market, the bidding model of power generators participating in the electricity spot market, and the electricity spot market clearing model. The bidding strategy of power generators is then determined based on the trained deep reinforcement learning algorithm.

[0107] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Based on the electricity volume and price traded by power generators in the medium- and long-term market and the spot market, a power market transaction settlement model is constructed to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market. Based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, a model is constructed to assess the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market, thereby obtaining the subjective value of power generators' profits from clearing the spot market. Based on the subjective value of power generators in obtaining profits from clearing the spot market, a bidding model for power generators to participate in the electricity spot market is constructed to predict their bidding strategies. Based on pre-set system boundary conditions and power generator application information, a power spot market clearing model is constructed to verify the predicted bidding strategies of the power generators. The deep reinforcement learning algorithm is trained based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategies of power generators in the spot market, the bidding model of power generators participating in the electricity spot market, and the electricity spot market clearing model. The bidding strategy of power generators is then determined based on the trained deep reinforcement learning algorithm.

[0108] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and device side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0109] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0110] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0111] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A bidding strategy method for power generators that considers bounded rationality, characterized in that, include: Based on the electricity volume and price traded by power generators in the medium- and long-term market and the spot market, a power market transaction settlement model is constructed to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market. Based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, a model is constructed to assess the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market, thereby obtaining the subjective value of power generators' profits from clearing the spot market. Based on the subjective value of power generators in obtaining profits from clearing the spot market, a bidding model for power generators to participate in the electricity spot market is constructed to predict their bidding strategies. Based on pre-set system boundary conditions and power generator application information, a power spot market clearing model is constructed to verify the predicted bidding strategies of the power generators. The deep reinforcement learning algorithm is trained based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategies of power generators in the spot market, the bidding model of power generators participating in the electricity spot market, and the electricity spot market clearing model. The bidding strategy of power generators is then determined based on the trained deep reinforcement learning algorithm.

2. The method according to claim 1, characterized in that, The electricity market transaction settlement model includes a total electricity settlement algorithm and a deviation settlement algorithm; Based on the electricity volume and prices traded by power generators in the medium- and long-term market and the spot market, a power market transaction settlement model is constructed to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market, including: Based on the electricity volume and price traded by the power generator in the medium- and long-term market, and using the aforementioned total electricity volume settlement algorithm, the electricity sales revenue of the power generator in the medium- and long-term market is obtained. Based on the electricity volume and price traded by the power generator in the spot market, and using the aforementioned deviation settlement algorithm, the power generator's revenue from electricity sales in the spot market is obtained.

3. The method according to claim 1, characterized in that, Based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, a model is constructed to assess the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market. This model yields the subjective value of power generators' profits from clearing the spot market, including: Based on the pre-set power generation cost coefficient of the power generator and the transaction volume of the power generator in the medium- and long-term market and the spot market, the transaction cost of the power generator in the medium- and long-term market and the spot market is obtained. The transaction profit of the power generator in the medium- and long-term market and the spot market is obtained based on the electricity sales revenue and transaction costs of the power generator in the medium- and long-term market and the spot market. Based on the pre-set psychological reference point of the generator for total profit, and according to the generator's trading profit in the medium and long term market, the generator's psychological reference point for spot market profit is obtained. Based on the generator's psychological reference point for spot market profits and the generator's trading profits in the spot market, the subjective value of the generator's profit from clearing the spot market is obtained.

4. The method according to claim 3, characterized in that, Based on the generator's psychological reference point for spot market profits and the generator's trading profits in the spot market, the subjective value of the generator's profit from clearing the spot market is derived, including: When the generator's trading profit in the spot market is greater than or equal to the generator's psychological reference point for spot market profit, the subjective value of the generator's profit from clearing the spot market is determined based on the generator's risk preference coefficient in the profit area, the generator's trading profit in the spot market, and the generator's psychological reference point for spot market profit. When the generator's trading profit in the spot market is less than the generator's psychological reference point for spot market profit, the subjective value of the generator's profit from clearing the spot market is determined based on the generator's risk preference coefficient and loss aversion coefficient in the loss zone, the generator's trading profit in the spot market, and the generator's psychological reference point for spot market profit.

5. The method according to claim 1, characterized in that, The objective function in the bidding model for power generators participating in the electricity spot market is to maximize the subjective value of the power generator's profit from clearing the spot market, based on the power generator's bid price in the spot market. The constraint condition in the bidding model for the power generator participating in the electricity spot market is the constraint price condition of the power generator's bid price in the spot market.

6. The method according to claim 1, characterized in that, The objective function in the electricity spot market clearing model is to minimize the total transaction cost for all power generators in the spot market; The constraints in the electricity spot market clearing model include node power balance constraints, unit output upper and lower limit constraints, unit ramping constraints, line power flow constraints, and node voltage phase angle constraints.

7. The method according to claim 1, characterized in that, The deep reinforcement learning algorithm includes a state space, an action space, an immediate reward, and an experience pool; The state space includes the target time period, the electricity volume and price of power generators in the spot market during the target time period; The action space includes the estimated bidding strategy of power generators determined based on the electricity market transaction settlement model, the impact model of medium- and long-term transactions on the bidding strategy of power generators in the spot market, and the bidding model of power generators participating in the electricity spot market. The immediate reward includes the subjective value of the generator's profit from clearing the spot market, obtained based on the estimated generator bidding strategy and the impact model of the medium- and long-term transactions on the generator bidding strategy in the spot market. The training of the deep reinforcement learning algorithm includes: The input of the deep reinforcement learning algorithm is determined based on the state space, and the output of the deep reinforcement learning algorithm is determined based on the action space. The output of the deep reinforcement learning algorithm is input into the electricity spot market clearing model to obtain four-tuple data. The four-tuple data is placed into an experience pool, and a set of four-tuple data is randomly selected from the experience pool to train the deep reinforcement learning algorithm. The four-tuple data includes the state corresponding to the target time period, the estimated generator bidding strategy obtained based on the state corresponding to the target time period, the immediate reward corresponding to the estimated generator bidding strategy, and the state corresponding to the next time period.

8. A device for a power generator bidding strategy that considers bounded rationality, characterized in that, include: The settlement module is used to construct a power market transaction settlement model based on the electricity volume and price traded by power generators in the medium- and long-term market and the spot market, and to obtain the electricity sales revenue of power generators in the medium- and long-term market and the spot market. The bounded rationality module is used to construct a model of the impact of medium- and long-term transactions on the bidding strategies of power generators in the spot market based on the electricity sales revenue of power generators in the medium- and long-term market and the spot market, and to obtain the subjective value of the profit obtained by power generators from clearing the spot market. The bidding module is used to construct a bidding model for power generators to participate in the electricity spot market based on the subjective value of power generators to obtain profits by clearing the spot market, and to predict the bidding strategies of power generators. The clearing module is used to construct a power spot market clearing model based on pre-set system boundary conditions and power generator declaration information, and to verify the estimated bidding strategies of the power generators. The deep reinforcement learning module is used to train the deep reinforcement learning algorithm based on the electricity market transaction settlement model, the impact model of the medium- and long-term transactions on the bidding strategies of power generators in the spot market, the bidding model of power generators participating in the electricity spot market, and the electricity spot market clearing model, and to determine the bidding strategies of power generators based on the trained deep reinforcement learning algorithm.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the generator bidding strategy method that takes into account bounded rationality as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the generator bidding strategy method that takes into account bounded rationality as described in any one of claims 1 to 7.

Images