Medium-term regional power load prediction method based on data clustering theory

Abstract

Provided is a medium-term regional power load prediction method based on a data clustering theory. The method includes region partition, influence factor correlation analysis and load space distribution prediction. By utilizing a clustering theory method and adopting a region block as a basic unit, each region block within a district is repartitioned and combined according to the similarity to serve as a basic unit for later load prediction. By utilizing a neural network algorithm to predict the load density and the synchronous rate, the time, space and other multi-attribute multilayer factorsare taken into account, the influence factors of different layers on the load density and the synchronous rate are mined, the precision and efficiency of the load prediction method can be improved, and refined planning of a power grid is facilitated. Meanwhile, a BP neural network algorithm is suitable for load prediction and has the advantages that calculation is stable, the classification capability in an arbitrary complicated pattern is excellent, and the multi-dimensional function mapping capability is great.

Description

technical field [0001] The invention relates to a mid-term regional power load forecasting method based on data clustering theory, which belongs to the technical field of power load forecasting. Background technique [0002] Medium-term load forecasting is the basic work of grid planning and operation, and accurate forecasting can bring huge economic benefits to grid planning. However, the time span of medium-term load forecasting is large, and it is affected by various types of nonlinear and uncertain factors, and has significant regional characteristics, making forecasting difficult. [0003] At present, in my country's actual grid planning work, medium-term power load forecasting usually adopts traditional combined forecasting methods, such as combining regression analysis, trend extrapolation, elastic coefficient method and large user + natural growth method. This combined forecasting method is simple to operate and requires less historical data, but this method only co...

AI Technical Summary

Problems solved by technology

These research methods all have the problem of subjectively weighting the load influencing factors, which reduces the accuracy of load forecasting, and the clustering algorithm only performs simple classification and forecasting on the composition of the load, without taking into account the differences in the characteristics of power loads in different regions. The factors affecting the load considered are also rough

Images