CN-122022885-A - Day-ahead electricity price prediction method based on market subject behavior mode and finite rational theory simulation

CN122022885ACN 122022885 ACN122022885 ACN 122022885ACN-122022885-A

Abstract

The invention discloses a day-ahead electricity price prediction method based on a market subject behavior mode and a finite rationality theory simulation, which belongs to the technical field of electricity price prediction and comprises the following steps of S1, analyzing day-ahead market transaction flow of a typical market at home and abroad, S2, introducing a simulation declaration information strategy coefficient, constructing a day-ahead electricity price prediction framework based on a market participant behavior mode and a finite rationality theory simulation, representing and simulating market subject declaration information, S3, predicting the market subject declaration information based on a typical time sequence prediction model, and S4, combining the predicted subject declaration information, superposing declaration quantity price curve information forming the whole market and obtaining predicted day-ahead electricity price. The invention solves the problems that the dynamic characteristics of the influencing factors of the current day-ahead electricity price change are difficult to extract and the formation mechanism of the day-ahead electricity price is difficult to embody. According to the invention, through reasonably simulating the coupling relation between the market declaration behavior information and the day-ahead electricity price, more accurate day-ahead electricity price prediction is obtained.

Inventors

ZHANG JINGYU
WANG YUQING
WANG FEI
ZHEN ZHAO

Assignees

华北电力大学

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (10)

1. The day-ahead electricity price prediction method based on the market subject behavior mode and the finite rational theory simulation is characterized by comprising the following steps of: s1, analyzing a daily market transaction flow of typical markets at home and abroad; S2, introducing a simulated declaration information strategy coefficient, constructing a day-ahead power price prediction framework based on market participant behavior mode and limited rationality theory simulation, and representing and simulating market subject declaration information; S3, predicting the main body declaration information of the market based on a typical time sequence prediction model by combining the main body declaration information of the simulated market; And S4, combining the predicted declaration information of each main body, superposing the declaration price curve information of the whole market, and acquiring the predicted day-ahead electricity price.
2. The method for predicting the day-ahead electricity price based on the market subject behavior mode and the finite rationality theory simulation according to claim 1, wherein in the step S2, a simulation declaration information policy coefficient is introduced and used for representing the market subject declaration information, and the simulation declaration information policy coefficient comprises a simulation declaration policy coefficient, a simulation quotation policy coefficient and a simulation quotation segment bit coefficient.
3. The method for predicting the day-ahead electricity price based on the market subject behavior pattern and the finite rational theory simulation according to claim 2, wherein the formula of the simulation declaration strategy coefficient is as follows: ; Wherein k i,t is an analog reporting policy coefficient of the ith main body at the t moment, when k i,t =1, the model represents quotation according to marginal cost, C i,t,n is n-th segment quotation information of the ith main body at the t moment, and MC i,t,n is corresponding cost; The formula of the simulated quotation strategy coefficient is as follows: ; Wherein, l i,t is the simulation quotation strategy coefficient of the ith main body at the t moment, wherein l i,t <1 indicates lower than the predicted electricity price quotation for improving the winning probability, l i,t >1 indicates higher than the predicted electricity price quotation for improving the winning income, and y i,t is the day-ahead predicted electricity price of the ith main body for the t moment; the formula of the simulated quotation segment coefficient is as follows: ; Where z i,t is a simulated quotation segment coefficient at the ith time of the ith main body, and z i,t =1, which indicates that the market main body is willing to take the declared electricity price based on the predicted electricity price as the declared electricity price of the first segment.
4. The method for predicting the day-ahead electricity price based on the market subject behavior pattern and the finite rational theory simulation according to claim 3, wherein in S2, a day-ahead electricity price prediction framework based on the market participant behavior pattern and the finite rational theory simulation is constructed, and the following operations are performed: Inputting known basic data, market public information and predicted electricity prices of various market subjects based on the public information, wherein the known basic data comprise constant coefficients a i 、b i 、c i of a secondary cost model of various market subjects, the unit capacity of a thermal power unit and the minimum technical output proportion, and the market public information comprises historical electricity price information, forecast market information and weather information; Simulating a main body declaration at a T moment, wherein the main body declaration comprises the steps of predicting the selection preference of a declaration section corresponding to the current price of the current day and the current day, formulating a quotation strategy and formulating a declaration strategy by each main body, and simulating the market main body to predict the behavior information of the current price of the current day based on the public information of the current price information, the market information and the weather information contained in total T 1 time periods of the previous N 1 days by utilizing an LSTM model; simulating the main body declaration price information at the t moment, and simulating corresponding declaration price information based on declaration behaviors of each main body; Simulating a total declaration price curve of the market at the t moment, and superposing the total declaration price curve information of the market based on declaration price information of each main body; simulating the market price at the t moment, and obtaining simulated market price based on the whole declaration price curve information of the market and in combination with the market load demand; And simulating the solution of the main body declaration information of the market at the t moment, and solving to obtain the reasonable simulated declaration information strategy coefficients of different moments of each main body by utilizing a PSO algorithm and basic constraint conditions and combining strategy coefficient fluctuation range constraint.
5. The method for predicting the day-ahead electricity price based on the market subject behavior pattern and the finite rationality theory simulation according to claim 4, wherein in the step S2, the basic constraint conditions include a bid amount constraint, a bid price constraint, a thermal power bidding space constraint, a market clearing error constraint and a simulated declaration information policy coefficient upper and lower limit constraint.
6. The day-ahead electricity price prediction method based on market subject behavior mode and finite rationality theory simulation according to claim 5, wherein in the step S2, in the process of solving the reasonable simulation reporting information policy coefficients of different subjects, based on the finite rationality theory, the inherent association mechanism between the simulation reporting information policy coefficients of different market subjects under different time periods is revealed, the coupling relation and uncertainty between the fuzzy mathematical theory policy coefficients are introduced for modeling, the membership functions are adopted for representing the fluctuation degree and amplitude characteristics, 2 fluctuation amplitude grades are set, the membership functions respectively represent small fluctuation and large fluctuation, wherein the fuzzy set formula based on the zade representation method is as follows: ; Where A (x 1 )/x 1 has only a symbolic meaning, meaning that the membership of point x 1 to fuzzy set A is A (x 1 ).
7. The method for predicting the day-ahead electricity price based on the market subject behavior mode and the finite rationality theory simulation according to claim 6, wherein in the step S2, under the constraint condition based on the fluctuation amplitude, all subjects are required to meet the fluctuation amplitude constraint in all time periods, and on the basis of reasonably setting the fluctuation amplitude constraint, penalty coefficients are introduced by combining the simulation quotation strategy coefficients and the simulation quotation segment coefficient deviation average values of the same time period of the last time period and the last day to ensure the smoothness of the market subject strategy coefficients; Wherein, the formula of penalty coefficient is as follows: ; ; Wherein, θ t is the penalty coefficient of the t period, I is the number of subjects, and θ max is the penalty coefficient upper limit.
8. The method for predicting the day-ahead electricity price based on the market subject behavior pattern and the finite rationality theory simulation according to claim 7, wherein in the step S3, based on solving to obtain the simulated reporting information policy coefficients of each market subject for the total of T 2 time periods in the previous N 2 days, the simulated reporting information policy coefficients of the next T 2 +1 time period of the ith subject are respectively predicted by using a typical time sequence prediction model, and the simulated reporting information policy coefficients of the 24 time periods of the next day are further predicted in the form of a sliding window.
9. The method for predicting the day-ahead electricity price based on the market subject behavior mode and the finite rational theory simulation according to claim 8, wherein in the step S4, the simulated declaration information policy coefficient of each market subject in the next 24 time period is obtained based on solving, the corresponding declaration quantity price information is simulated, the declaration quantity price curve information of the whole market is formed by superposition, the price of the simulated market is obtained by combining the thermal power bidding space of the next market, and the day-ahead electricity price is predicted.
10. The method for predicting the day-ahead electricity price based on the market subject behavior pattern and the finite rational theory simulation according to claim 9, wherein in S4, the day-ahead electricity price prediction error is estimated by adopting a normalized root mean square error, a normalized mean absolute error and a decision coefficient, and the accuracy of the day-ahead electricity price prediction is estimated by combining the actual day-ahead electricity price upper limit of the current month; The root mean square error RMSE is used for measuring the variance level of the error and is sensitive to abnormal values, and the normalized root mean square error NRMSE is normalized according to the maximum value, and the formula is as follows: ; ; The mean absolute error MAE is used for measuring the mean absolute difference level of the error, reflecting the overall mean deviation degree of the model, and the normalized mean absolute error NMAE is normalized according to the maximum value, and the formula is as follows: ; ; Determining coefficients For measuring the fitting degree of the model predicted value and the true value, wherein, =1 Represents perfect prediction, and the formula is as follows: ; Wherein N is the number of predicted-actual values, Y k is the actual value, and Y k is the predicted value; is the average of the true values.

Description

Day-ahead electricity price prediction method based on market subject behavior mode and finite rational theory simulation Technical Field The invention relates to the technical field of electricity price prediction, in particular to a day-ahead electricity price prediction method based on market subject behavior mode and limited rational theory simulation. Background Current electricity prices are affected by many factors and highly accurate predictions are quite difficult and challenging. In terms of influencing factors including weather, fuel cost, load demand, bidding strategies and the like, the influence degree of different influencing factors on the spot electricity price in different time periods is different, and the high sensitivity and volatility of the spot electricity price in time and space dimension before the day are determined based on a complex spot electricity price forming mechanism and the characteristic that electric power cannot be stored. Meanwhile, with gradual access of renewable energy sources, the fluctuation and uncertainty of the power system are further increased, so that more complex space-time dynamic characteristics are caused, and even more frequent negative electricity prices occur. Therefore, the day-ahead electricity price is predicted from the external complex characteristic data only, is greatly influenced by the selected influence factor data, and meanwhile, the dynamic characteristics of the influence factors of the change of the day-ahead electricity price are difficult to extract, and the formation mechanism of the day-ahead electricity price is also difficult to embody. Disclosure of Invention The invention aims to provide a day-ahead electricity price prediction method based on a market subject behavior mode and limited rationality theory simulation, which introduces a simulated reporting information strategy coefficient, combines the limited rationality theory simulation and characterizes complex reporting strategy behaviors and reporting price information of different market subjects in each period of a training set, thereby accurately reflecting the adaptability of an electricity price prediction model to the market behavior complexity and the dynamic change thereof, and solving the problems in the background technology. In order to achieve the above purpose, the present invention provides the following technical solutions: a day-ahead electricity price prediction method based on market subject behavior mode and finite rational theory simulation comprises the following steps: s1, analyzing a daily market transaction flow of typical markets at home and abroad; S2, introducing a simulated declaration information strategy coefficient, constructing a day-ahead power price prediction framework based on market participant behavior mode and limited rationality theory simulation, and representing and simulating market subject declaration information; S3, predicting the main body declaration information of the market based on a typical time sequence prediction model by combining the main body declaration information of the simulated market; And S4, combining the predicted declaration information of each main body, superposing the declaration price curve information of the whole market, and acquiring the predicted day-ahead electricity price. And combining with market rules, reasonably simulating declaration strategy behaviors and price information of each main body under clear electricity price at different moments as much as possible, wherein the declaration strategy behaviors at different moments can also be used for representing dynamic change characteristics of key factors affecting the electricity price, so that construction of a day-ahead electricity price prediction model comprising an electricity price formation mechanism and a data driving method is realized, and the purpose of predicting the day-ahead electricity price is realized. Preferably, in the step S2, a simulated declaration information policy coefficient is introduced, and is used for characterizing the declaration information of the market main body, where the simulated declaration information policy coefficient includes a simulated declaration policy coefficient, a simulated quotation policy coefficient and a simulated quotation segment coefficient. Preferably, the formula of the simulated reporting policy coefficient is as follows: Wherein k i,t is an analog reporting policy coefficient of the ith main body at the t moment, when k i,t =1, the model represents quotation according to marginal cost, C i,t,n is n-th segment quotation information of the ith main body at the t moment, and MC i,t,n is corresponding cost; The formula of the simulated quotation strategy coefficient is as follows: Wherein, l i,t is the simulation quotation strategy coefficient of the ith main body at the t moment, wherein l i,t <1 indicates lower than the predicted electricity price quotation for improving the winning proba