CN-122022875-A - Power market price limiting optimization method based on multidimensional data and adaptive evaluation

Abstract

The invention discloses an electric power market price limiting optimization method based on multi-dimensional data and self-adaptive evaluation, and relates to the technical field of electric power markets, wherein the method comprises the steps of constructing a multi-dimensional market state index system, training an intelligent body by adopting a dual delay depth deterministic strategy gradient TD3 algorithm to automatically generate an initial price limiting interval, designing a reward function to define a training objective function of the intelligent body, outputting an optimal price limiting interval, constructing a multi-element main body decision optimization model based on the market state and the optimal price limiting interval, re-optimizing the optimal price limiting interval, outputting a price limiting interval taking main body response into consideration, constructing a multi-dimensional evaluation matrix quantization price limiting effect based on the price limiting interval taking main body response into consideration, calculating market comprehensive scoring indexes corresponding to different price limiting intervals according to the market multi-dimensional evaluation matrix, and adjusting core parameters by utilizing Bayesian optimization iteration to realize self-evolution of a price limiting mechanism. The invention reduces price fluctuation rate of the electric power market.

Inventors

• HUANG YIZHOU
• CHEN WEI
• LIAO QIWEN
• TAN SHAOSI
• SONG LONG
• ZHANG YI
• WEI XINGCHEN
• WANG JUNTAO
• ZHANG ZHE
• ZHAO TAO
• LIU MING
• LU WEI
• ZHANG RUIRUI

Assignees

• 华润电力投资有限公司中西分公司
• 河南华润电力首阳山有限公司
• 华润电力焦作有限公司
• 华润电力登封有限公司
• 河南华润电力古城有限公司
• 华润电力(河南)销售有限公司

Dates

Publication Date

20260512

Application Date

20251202

Claims (10)

1. 1. The utility model provides an electric power market limit optimizing method based on multidimensional data and self-adaptive evaluation, which is characterized by comprising the following steps: constructing a multidimensional market state index system, training an agent by adopting a dual delay depth deterministic strategy gradient TD3 algorithm, automatically generating an initial price limiting interval, and meanwhile, defining a training objective function of the agent by designing a reward function to output an optimal price limiting interval; constructing a multi-element main body decision optimization model based on the perceived market state and the optimal price limiting interval, re-optimizing the optimal price limiting interval, and outputting the price limiting interval after the main body response is considered; Based on the price limiting interval after the main body response is considered, constructing a multidimensional evaluation matrix quantization price limiting effect, calculating market comprehensive scoring indexes corresponding to different price limiting intervals according to the market multidimensional evaluation matrix, and adjusting core parameters by utilizing Bayesian optimization iteration to realize self-evolution of a price limiting mechanism.
2. 2. The method of claim 1, wherein the constructing a multi-dimensional market state index system and training the agent using a dual delay depth deterministic strategy gradient TD3 algorithm to automatically generate an initial limit interval further comprises: The following market key factors are selected to construct a multi-dimensional market state index system: Wherein, the A market state vector representing a period T, T representing a vector transpose, Representing the total load demand of the system for the period t, Representing the power generation capacity available to the system for the period t, Representing the system supply and demand tension of the period t, The new energy output prediction deviation rate is represented, Representing the positive rotation reserve margin of the system for the period t, Representing the severity of the system critical section blockage for period t; the fuel price index change rate for the period t is expressed, Representing the market price volatility at time t, Market centralization index representing period t; based on the market state index system, a TD3 training agent is adopted to automatically generate a limit interval, and the state interval and the action space are as follows: Wherein the state space Representing a multidimensional data space in the face of an agent, i.e Historical samples of seven-dimensional data in the model are used as training data, dynamic space Representing the specific actions to be determined by the agent during training, Represents the lower limit of the price for the period t, Representing the upper bound of the t-period limit.
3. 3. The method according to claim 2, wherein the defining the training objective function of the agent by designing the bonus function, outputting the optimal price-limiting interval, further comprises: after the state interval and the action interval are determined, the following reward function is designed for the intelligent agent: Wherein, the — Representing the five weight coefficients of the model, Represents the sum of social benefits for the period t, Represents the market clearing price for the period t, Representing the system marginal cost for the period t, Representing a virtual indication function that is to be used, Representing a specific limit value for the period t, Representing the market risk value estimate for period t, Representing the limit price change amplitude of two adjacent time periods; Based on the reward function, the training objective function of the agent is: Wherein, the Representing the discount factor of step t.
4. 4. The method of claim 3, wherein constructing a multi-element body decision optimization model based on the perceived market state and the optimal price limit interval, re-optimizing the optimal price limit interval, further comprises: constructing a multi-element main body decision optimization model comprising four types of market main bodies, wherein the four types of market main bodies comprise traditional power generators, new energy stations, flexible resource aggregators and large users/power selling companies; Based on a multi-element main body decision optimization model, substituting four types of market main bodies into a game process to obtain Nash equilibrium solutions, and obtaining the game equilibrium solutions based on a price limit interval given by a TD3 algorithm.
5. 5. The method of claim 4, wherein constructing a multidimensional evaluation matrix quantifies a price limit effect based on the price limit interval after considering the subject response, further comprising: under the condition of a given price limiting interval, the following multidimensional evaluation matrix is designed: where N represents a period of scrolling forward from the t period, Representing a collection of power market principals, Represents the clearing price for the k period of forward scrolling, The average value of the electricity price is represented, For the purpose of characterizing the price stability, Represents social benefits assuming that the electric market is clear without limit, The social efficiency is characterized in that, Representing the revenue function of the market subject i over the period t, The coefficient of the basis is represented by the coefficient of the basis, Indicating the fairness of the competition in the market, Representing the profit margin of the market subject i for period t, Representing the conditional risk value at 95% confidence, For measuring the system risk of the market.
6. 6. The method of claim 1, wherein calculating market synthesis score indicators corresponding to different price-limiting intervals according to the market multidimensional assessment matrix further comprises: based on the multi-dimensional evaluation matrix of the market, calculating comprehensive scoring indexes of the market in different price limiting intervals: Wherein, the Representation of The j index of (a); Representation of The weight of the j index in (a); The generation mode is as follows: Wherein, the Representing an adjustment sensitivity coefficient; Representation of Historical means of the j-th index in (a).
7. 7. The method of claim 6, wherein the iteratively adjusting the core parameters and utilizing bayesian optimization achieves self-evolution of the price limiting mechanism, further comprising: the Bayesian optimization method is adopted to adjust the parameters of the price limiting model, and the corresponding optimization process is as follows: Wherein, the Representing a set of core parameters to be adjusted; The method comprises the following steps: Wherein, the For each weight coefficient of the bonus function, A change threshold value representing a new energy output prediction deviation; Representing the coefficient of severity of the obstruction A change threshold of (2); meaning to find a given set of parameters Lower part(s) Is set to a maximum value that is desirable for the condition of (c), the constraint conditions are as follows: Wherein, the Representing a minimum value of market synthesis score set according to regulatory requirements, A set of parameters representing the kth iteration, Representing the first order of the norm, Representing the deviation threshold of the parameter update.
8. 8. An electric power market limit optimizing device based on multidimensional data and self-adaptive evaluation, which is characterized by comprising: The initial price limiting interval generation module is used for constructing a multidimensional market state index system, training the intelligent agent by adopting a TD3 algorithm to automatically generate an initial price limiting interval, and meanwhile, defining a training objective function of the intelligent agent by designing a reward function to output an optimal price limiting interval; The main body response type price limiting interval optimizing module is used for constructing a multi-main body decision optimizing model based on the perceived market state and the optimal price limiting interval, simulating game decisions of four types of market main bodies, re-optimizing the optimal price limiting interval and outputting the price limiting interval after main body response is considered; And the price limiting mechanism self-evolution module is used for constructing a multidimensional evaluation matrix quantization price limiting effect based on a price limiting interval after the response of the main body is considered, calculating market comprehensive scoring indexes corresponding to different price limiting intervals according to the market multidimensional evaluation matrix, and utilizing Bayesian optimization iteration to adjust core parameters so as to realize the self-evolution of the price limiting mechanism.
9. 9. An electronic device comprising a processor and a memory; wherein the processor runs a program corresponding to executable program code stored in the memory by reading the executable program code for implementing the method according to any one of claims 1-7.
10. 10. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the program, when executed by a processor, implements the method according to any one of claims 1-7.

Description

Power market price limiting optimization method based on multidimensional data and adaptive evaluation Technical Field The invention relates to the technical field of electric power markets, in particular to an electric power market price limiting optimization method and device based on multidimensional data and self-adaptive evaluation. Background The pricing mechanism is one of core elements of the electric power market, in order to realize the coupling of social benefit maximization and safe operation of an electric power system, the electric power market is generally designed with a corresponding price limiting mechanism, but the price limiting mechanism of the electric power market in China has three core problems in actual operation, namely, the price limiting rule is stiff and dynamic response is absent, the existing mechanism mostly adopts a fixed price upper and lower limit or month adjustment mode, dynamic scenes such as new energy fluctuation, fuel cost jump and partial blocking are difficult to adapt in real time, and administrative scheduling instructions squeeze market space, negative electricity price frequent caused by excessive starting of non-marketized coal and electricity, the main behavior prediction and market effect evaluation are absent, a main behavior mapping model of 'price limiting signal- > main decision' is not established, the price limiting policy is disjointed with main behaviors such as force up commodity prices, VPP and load side response are absent, the supervision depends on a single-dimension index, social benefit, fairness and system risk quantitative analysis is absent, the adjustment period is long and hysteresis is difficult, and the scheduling instructions, the market price prediction instructions, the new energy prediction and other multi-source data belong to independent systems, the existing policy is not dependent on the price limiting policy is difficult to apply, the actual price feedback parameters are difficult to realize, and the self-adaptive to the actual market configuration, and the price limiting effect is difficult to develop. Disclosure of Invention The invention mainly aims to provide an electric power market price limiting optimization method based on multidimensional data and self-adaptive evaluation. Another object of the present invention is to provide an electric power market price limiting optimizing device based on multidimensional data and adaptive evaluation. A third object of the present invention is to propose an electronic device. A fourth object of the present invention is to propose a non-transitory computer readable storage medium. To achieve the above objective, an embodiment of a first aspect of the present invention provides a power market price limiting optimization method based on multidimensional data and adaptive evaluation, including: constructing a multidimensional market state index system, training an agent by adopting a dual delay depth deterministic strategy gradient TD3 algorithm, automatically generating an initial price limiting interval, and meanwhile, defining a training objective function of the agent by designing a reward function to output an optimal price limiting interval; constructing a multi-element main body decision optimization model based on the perceived market state and the optimal price limit interval, simulating game decisions of four types of market main bodies, re-optimizing the optimal price limit interval, and outputting the price limit interval after main body response is considered; Based on the price limiting interval after the main body response is considered, constructing a multidimensional evaluation matrix quantization price limiting effect, calculating market comprehensive scoring indexes corresponding to different price limiting intervals according to the market multidimensional evaluation matrix, and adjusting core parameters by utilizing Bayesian optimization iteration to realize self-evolution of a price limiting mechanism. Optionally, the constructing a multidimensional market state index system, training the agent by adopting a dual delay depth deterministic strategy gradient TD3 algorithm to automatically generate an initial price limiting interval, and further includes: The following market key factors are selected to construct a multi-dimensional market state index system: Wherein, the A market state vector representing a period T, T representing a vector transpose,Representing the total load demand of the system for the period t,Representing the power generation capacity available to the system for the period t,Representing the system supply and demand tension of the period t,The new energy output prediction deviation rate is represented,Representing the positive rotation reserve margin of the system for the period t,Representing the severity of the system critical section blockage for period t; the fuel price index change rate for the period t is expressed, Representing the mar