Method and system for short-term load forecasting of regional power grid

Abstract

The invention relates to a short-term load forecasting method and system for a regional power grid. The method includes: the network terminal collects historical multivariate feature information data related to load forecasting, the network crawls the future weather forecast broadcast information, and performs data preprocessing on the collected feature information; the weather information is fine-grained, and the weather on the day to be measured is virtually similar Day selection; virtual similar day selection for daily load to be predicted; based on historical characteristic information data, and based on deep learning theory to determine the trained short-term load forecasting model, and then realize the prediction of future load. The invention can improve the accuracy of short-term load forecasting.

Description

technical field [0001] The invention relates to the field of short-term load forecasting of power systems, in particular to a method and system for short-term load forecasting of regional power grids. Background technique [0002] With the development of economy and the integration of large-scale distributed energy sources, the base of weather-sensitive loads in regional power grids continues to rise, the daily load peak-to-valley difference continues to widen, and the local load changes become more random and complex. How to scientifically and effectively combine the fine-grained meteorological feature data to deeply mine the relationship between relevant factors and local load changes in short-term load forecasting is an inevitable direction to further improve the accuracy of load forecasting and achieve refined management of load forecasting. [0003] Traditional research on short-term load forecasting of regional power grids can generally be divided into two steps: one i...

AI Technical Summary

Problems solved by technology

[0004] For the existing load-similar days, most of the current load-similar day selection methods still use coarse-grained daily characteristic weather as the feature quantity for similar day selection, which is difficult to map to local changes in load

And most of the methods select load similar days from historical days, and there is often an excessive partial similarity between historical days and similar days, which has a great impact on the effect of load forecasting

Images