CN-122022884-A - Electricity price prediction method and system based on statistics

CN122022884ACN 122022884 ACN122022884 ACN 122022884ACN-122022884-A

Abstract

The invention belongs to the technical field of electricity price prediction, and discloses a statistical-based electricity price prediction method and a statistical-based electricity price prediction system, wherein the method comprises the steps of obtaining an original data set, wherein the original data set comprises a historical electricity price sequence, load data, weather indexes and fuel prices; the method comprises the steps of preprocessing an original data set to obtain a feature set after dimension reduction, decomposing the feature set after dimension reduction through a time sequence to obtain a component set, constructing a trend prediction model, a seasonal prediction model and a nonlinear residual prediction model according to the component set, fusing prediction results to obtain a power price prediction sequence, comparing the power price prediction sequence with a historical actual power price prediction sequence to obtain an error evaluation result, adjusting super parameters of the nonlinear residual prediction model based on the error evaluation result and the power price prediction sequence to obtain an optimized prediction model, predicting the power price of a target period by using the optimized prediction model to obtain a power price prediction curve, providing a basis for a power market, and reducing operation cost.

Inventors

WEN LIMING
WU MIN
LI DALIN
ZHANG DU
HUANG MIN

Assignees

华能江西清洁能源有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260119

Claims (10)

1. The electricity price prediction method based on statistics is characterized by comprising the following steps of: acquiring an original data set, wherein the original data set comprises a historical electricity price sequence, load data, weather indexes and fuel prices; preprocessing the original data set to obtain a feature set after dimension reduction, and decomposing the feature set after dimension reduction through a time sequence to obtain a component set; Constructing a trend prediction model, a seasonal prediction model and a nonlinear residual prediction model according to the component set, and fusing prediction results of the trend prediction model, the seasonal prediction model and the nonlinear residual prediction model to obtain an electricity price prediction sequence; comparing the electricity price prediction sequence with a historical actual electricity price prediction sequence to obtain an error evaluation result; And adjusting the super parameters of the nonlinear residual error prediction model based on the error evaluation result and the electricity price prediction sequence to obtain an optimized prediction model, and predicting the electricity price of the target period by using the optimized prediction model to obtain an electricity price prediction curve.
2. The statistical-based electricity price prediction method of claim 1, wherein: The electricity price prediction sequence is ; Wherein, the 、、 The fusion weights of the trend prediction result, the seasonal prediction result and the residual prediction result are respectively, 、、 Respectively is Predicted values of temporal trend component, seasonal component, and residual component.
3. The statistical-based electricity price prediction method of claim 2, wherein: the said Predicted value of time trend component The method comprises the following steps: Wherein, the As a coefficient of the slope of the trend, As a function of the time variable, Is the trend intercept coefficient.
4. The statistical-based electricity price prediction method of claim 2, wherein: The predicted values of the seasonal components are: Wherein, the As the reference coefficient of the seasons, As the harmonic order of the fourier series, Is the first The cosine coefficient of the subharmonic wave, In order to be able to take time, For the length of the seasonal period, Is the first Sine coefficients of subharmonics.
5. The statistical-based electricity price prediction method of claim 4, wherein: preprocessing the original data set to obtain a feature set after dimension reduction, wherein the feature set comprises; Filling the missing values in the original data set based on linear interpolation of adjacent time point data; And identifying and eliminating the abnormal values in the original data set based on a standard deviation or quartile range statistical method.
6. The statistical-based electricity price prediction method of claim 1, wherein: the error evaluation result comprises a root mean square error and an average absolute percentage error; the root mean square error is : Wherein, the The total number of points for the historical data of the error assessment, Is the first The historical actual electricity prices of the moments in time, Is the first Predicted value of electricity price at time.
7. The statistical-based electricity price prediction method of claim 6, wherein: the average absolute percentage error is : Wherein, the The total number of points for the historical data of the error assessment, Is the first The historical actual electricity prices of the moments in time, Is the first Predicted value of electricity price at time.
8. The statistical-based electricity price prediction method of claim 1, wherein: The nonlinear residual prediction model construction comprises the following steps: And modeling and predicting the residual component by using a random forest regression algorithm or a long-short-term memory neural network, wherein the nonlinear residual prediction model is input into the preprocessed load data, weather indexes and historical residual sequences.
9. The statistical-based electricity price prediction method of claim 8, wherein: adjusting the super parameters of the nonlinear residual prediction model based on the error evaluation result and the electricity price prediction sequence, comprising: and (3) using grid search or Bayesian optimization, and taking the weighted sum of root mean square error and average absolute percentage error as an objective function to automatically optimize the super-parameters of the nonlinear residual error prediction model.
10. A statistics-based electricity price prediction system, comprising: The acquisition module is used for acquiring an original data set, wherein the original data set comprises a historical electricity price sequence, load data, weather indexes and fuel prices; The decomposition module is used for preprocessing the original data set to obtain a feature set after dimension reduction, and decomposing the feature set after dimension reduction through a time sequence to obtain a component set; the fusion module is used for constructing a trend prediction model, a seasonal prediction model and a nonlinear residual prediction model according to the component set, and fusing the prediction results of the trend prediction model, the seasonal prediction model and the nonlinear residual prediction model to obtain an electricity price prediction sequence; the comparison module is used for comparing the electricity price prediction sequence with the historical actual electricity price prediction sequence to obtain an error evaluation result; and the prediction module is used for adjusting the super parameters of the nonlinear residual error prediction model based on the error evaluation result and the electricity price prediction sequence to obtain an optimized prediction model, and predicting the electricity price of the target period by using the optimized prediction model to obtain an electricity price prediction curve.

Description

Electricity price prediction method and system based on statistics Technical Field The invention belongs to the technical field of electricity price prediction, and relates to an electricity price prediction method and system based on statistics. Background The electricity price is a core regulation index of the operation of the electric power market, is influenced by coupling of multiple factors such as power load fluctuation, meteorological condition change, fuel cost fluctuation and the like, and shows remarkable trend, seasonal and nonlinear fluctuation characteristics. The existing electricity price prediction method only adopts a single prediction model, and is difficult to simultaneously process different components such as trend, seasonality and nonlinear residual errors in electricity price data. The nonlinear model can not effectively capture nonlinear variation characteristics in electricity price data, and although the nonlinear model can process nonlinear relations, modeling on trending and seasonal components is inaccurate, so that the prediction effect is poor. In addition, after the electricity price prediction sequence is obtained, error information between the prediction result and the historical actual electricity price cannot be fully utilized for optimization. Especially, the nonlinear residual prediction model lacks scientific basis for adjusting the super parameters, the model parameters are difficult to dynamically optimize according to the actual prediction errors, the prediction precision under different scenes is limited, and the requirements of the electric power market on high accuracy and high stability of electricity price prediction cannot be met. Disclosure of Invention Aiming at the defects existing in the prior art, the invention aims to provide a statistical-based electricity price prediction method and system. In order to achieve the above purpose, the present invention adopts the following technical scheme: The invention provides a statistical-based electricity price prediction method, which comprises the steps of obtaining an original data set, preprocessing the original data set to obtain a feature set after dimension reduction, decomposing the feature set after dimension reduction through a time sequence to obtain a component set, constructing a trend prediction model, a seasonal prediction model and a nonlinear residual prediction model according to the component set, fusing prediction results of the trend prediction model, the seasonal prediction model and the nonlinear residual prediction model to obtain an electricity price prediction sequence, comparing the electricity price prediction sequence with a historical actual electricity price prediction sequence to obtain an error evaluation result, adjusting super parameters of the nonlinear residual prediction model based on the error evaluation result and the electricity price prediction sequence to obtain an optimized prediction model, and predicting the electricity price of a target period by using the optimized prediction model to obtain an electricity price prediction curve. Further, the electricity price prediction sequence is that; Wherein, the 、、The fusion weights of the trend prediction result, the seasonal prediction result and the residual prediction result are respectively,、、Respectively isPredicted values of temporal trend component, seasonal component, and residual component. Further, the saidPredicted value of time trend componentThe method comprises the following steps: Wherein, the As a coefficient of the slope of the trend,As a function of the time variable,Is the trend intercept coefficient. Further, the predicted value of the seasonal component is: Wherein, the As the reference coefficient of the seasons,As the harmonic order of the fourier series,Is the firstThe cosine coefficient of the subharmonic wave,In order to be able to take time,For the length of the seasonal period,Is the firstSine coefficients of subharmonics. Further, preprocessing the original data set to obtain a feature set after dimension reduction, wherein the feature set comprises filling missing values in the original data set based on linear interpolation of adjacent time point data, and identifying and eliminating abnormal values in the original data set based on a standard deviation or a statistics method of four bit distances. Further, the error assessment result comprises a root mean square error and an average absolute percentage error; the root mean square error is : Wherein, the The total number of points for the historical data of the error assessment,Is the firstThe historical actual electricity prices of the moments in time,Is the firstPredicted value of electricity price at time. Further, the average absolute percentage error is: Wherein, the The total number of points for the historical data of the error assessment,Is the firstThe historical actual electricity prices of the moments in time,Is the firstPredicted v