CN-121983948-A - Power grid load prediction method and system based on daily load management

CN121983948ACN 121983948 ACN121983948 ACN 121983948ACN-121983948-A

Abstract

The application relates to a power grid load prediction method and system based on daily load management, which comprises the steps of obtaining daily load data, daily weather data and holiday data in a preset time period in a prediction day, inputting the load data, the weather data and the holiday data in the preset time into a trained load prediction model to obtain a daily maximum load predicted value of the prediction day, and inputting the daily maximum load predicted value into a load prediction linear regression equation to obtain a morning maximum load predicted value, a afternoon maximum load predicted value and a evening maximum load predicted value of the prediction day. According to the scheme provided by the application, the refined load prediction of different time periods of the prediction day can be realized by integrating the daily load data, the weather data and the holiday data, and the effectiveness of power grid dispatching and the stability of power grid operation are improved.

Inventors

XU LIPING
ZHANG XINXIN
WANG FENG
ZHU ZIZHAO

Assignees

广东电网有限责任公司江门供电局

Dates

Publication Date: 20260505
Application Date: 20251208

Claims (8)

1. The utility model provides a power grid load prediction method based on daily load management, which is characterized by comprising the following steps: acquiring load data in the day, weather data in the day and holiday data in the day within a preset time period in a prediction day; Inputting the load data, the weather data and the holiday data in the preset time into a trained load prediction model to obtain a daily highest load predicted value of the predicted day; And inputting the daily maximum load predicted value into a load prediction linear regression equation to obtain the morning maximum load predicted value, the afternoon maximum load predicted value and the evening maximum load predicted value of the predicted day.
2. The method according to claim 1, characterized in that it comprises: The weather data comprise day temperature data, day illumination data, day rainfall data, day wind power data and day humidity data; the forecast day is a working day or a holiday.
3. The method of claim 1, wherein the training process of the load prediction model comprises: Acquiring historical load data, historical weather data and historical holiday data in the preset time period; Comparing the daily load data with the historical load data, and screening out at least one historical target day corresponding to a historical load curve meeting preset conditions, wherein the historical load curve comprises current daily load data of the historical target day in a preset time period; The load data, weather data and holiday data of the historical target day are used as training samples to be input into a load prediction model, and the highest load prediction value of the historical target day is obtained; And optimizing the load prediction model through a preset back propagation algorithm based on the highest load prediction value of the historical target day until the highest load prediction value output by the load prediction model reaches a preset expected value.
4. The method of claim 3, wherein the obtaining historical load data, historical weather data, and historical holiday data for the preset time period comprises: acquiring historical load data, historical weather data and historical holiday data in a first preset range from a database at intervals of preset first time thresholds; and obtaining the historical load data, the historical weather data and the historical holiday data in a second preset range from the database.
5. A method according to claim 3, further comprising: Performing data cleaning and normalization processing on the historical load data and the historical weather data to obtain processed historical load data and processed historical weather data; Calculating a standard deviation of the historical load data based on the processed historical load data and the processed historical weather data; and constructing a historical load curve based on the standard deviation of the historical load data.
6. The method of claim 3, wherein optimizing the load prediction model by a preset back propagation algorithm based on the highest load prediction value of the historical target day until the highest load prediction value output by the load prediction model reaches a preset expected value further comprises: And if the mean square error of the load prediction model gradually becomes larger in iteration of continuous preset times, stopping training the load prediction model.
7. The method according to claim 1, characterized in that it comprises: and generating a visual chart or a visual curve based on the predicted morning maximum load value, the predicted afternoon maximum load value and the predicted evening maximum load value of the predicted day.
8. A daily load management-based grid load prediction system, comprising: the acquisition module is used for acquiring daily load data, daily weather data and holiday data of the current day in a preset time period in a prediction day; The input module is used for inputting the load data, the weather data and the holiday data in the preset time into the trained load prediction model to obtain a daily highest load predicted value of the predicted day; And the prediction module is used for inputting the daily maximum load predicted value into a load prediction linear regression equation to obtain the morning maximum load predicted value, the afternoon maximum load predicted value and the evening maximum load predicted value of the predicted day.

Description

Power grid load prediction method and system based on daily load management Technical Field The application relates to the technical field of power, in particular to a method and a system for predicting power grid load based on daily load management. Background With the increasing level of electrical energy replacement and the frequent occurrence of extreme climatic events, grid load prediction faces unprecedented challenges. In the related technology, the traditional method for predicting the daily load of the power grid has the problems of inaccurate prediction results and overlarge deviation, so that demand response offers are cancelled for many times, power load gaps cannot be filled up for many times, and the like. Disclosure of Invention In order to solve or partially solve the problems in the related art, the application provides a power grid load prediction method and a power grid load prediction system based on daily load management, which can realize the fine load prediction of different time periods of a prediction day and improve the effectiveness of power grid dispatching and the stability of power grid operation by integrating daily load data, weather data and holiday data. The application provides a power grid load prediction method based on daily load management, which comprises the steps of obtaining daily load data, daily weather data and holiday data in a preset time period in a prediction day, inputting the load data, the weather data and the holiday data in the preset time into a trained load prediction model to obtain a daily maximum load predicted value of the prediction day, and inputting the daily maximum load predicted value into a load prediction linear regression equation to obtain a morning maximum load predicted value, a afternoon maximum load predicted value and a evening maximum load predicted value of the prediction day. With reference to the first aspect, in a possible implementation manner of the first aspect, the weather data includes day air temperature data, day illumination data, day rainfall data, day wind power data and day humidity data, and the predicted day is a working day or a holiday. With reference to the first aspect, in one possible implementation manner of the first aspect, the training process of the load prediction model includes obtaining historical load data, historical weather data and historical holiday data in the preset time period, comparing the daily load data with the historical load data, screening out a historical target day corresponding to at least one historical load curve meeting preset conditions, wherein the historical load curve includes current day load data of the historical target day in the preset time period, inputting the load data, the weather data and the holiday data of the historical target day into the load prediction model as training samples to obtain a highest load prediction value of the historical target day, and optimizing the load prediction model through a preset back propagation algorithm based on the highest load prediction value of the historical target day until the highest load prediction value output by the load prediction model reaches a preset expected value. With reference to the first aspect, in a possible implementation manner of the first aspect, the acquiring the historical load data, the historical weather data and the historical holiday data in the preset time period includes acquiring the historical load data, the historical weather data and the historical holiday data in a first preset range from a database at intervals of preset first time threshold values, and acquiring the historical load data, the historical weather data and the historical holiday data in a second preset range from the database. With reference to the first aspect, in one possible implementation manner of the first aspect, the method further includes performing data cleaning and normalization processing on the historical load data and the historical weather data to obtain processed historical load data and processed historical weather data, calculating a standard deviation of the historical load data based on the processed historical load data and the processed historical weather data, and constructing a historical load curve based on the standard deviation of the historical load data. With reference to the first aspect, in one possible implementation manner of the first aspect, the optimizing the load prediction model by a preset back propagation algorithm based on the highest load prediction value of the historical target day until the highest load prediction value output by the load prediction model reaches a preset expected value further includes stopping training the load prediction model if a mean square error of the load prediction model becomes gradually larger in a continuous preset number of iterations. With reference to the first aspect, in a possible implementation manner of the first aspect, the method includes generati