Load integration prediction method based on CNN-SVR model

Abstract

The invention relates to a load integration prediction method based on a CNN-SVR model. The method comprises the steps of firstly initializing influence factors influencing loads in a certain region and daily load data of all users in the certain region, calculating a Pearson correlation coefficient, then performing extremum normalization and clustering on the data to obtain different types of user data tags, and combining the user data according to step grouping tags to serve as training input data; secondly, constructing a CNN-SVR load prediction model, then carrying out model training, preprocessing input data, inputting the data into the trained CNN-SVR model for prediction, obtaining a prediction result, and carrying out reverse normalization on the result to obtain multiple groups offinal prediction loads; and finally, summing the multiple groups of loads to obtain a daily final predicted load of a certain region, and summing the loads according to months if the monthly load ofa certain region is predicted. Compared with the prior art, the method can perform integrated classification on a large number of users in a certain region, and has the advantages of autonomous data feature extraction, high prediction precision and the like.

Description

technical field [0001] The invention relates to the field of power system load forecasting, in particular to an integrated load forecasting method based on a CNN-SVR model. Background technique [0002] Traditional power load forecasting methods include regression analysis, wavelet analysis, decision tree, random forest, support vector machine, neural network and other algorithms, among which the support vector regression machine is often used as a variant of the support vector machine algorithm for time series data The prediction of is a commonly used algorithm with good quasi-prediction accuracy, but because its prediction accuracy is greatly affected by its own parameters and input feature data, the input data often needs to go through the step of feature extraction. The neural network algorithm is a popular algorithm in recent years. It has a good ability to independently extract nonlinear features and can perform nonlinear fitting very well. However, the neural network...

AI Technical Summary

Problems solved by technology

Behind the continuous growth of electricity consumption is a huge number of electricity users. When performing regional load forecasting, in the face of a large number of electricity users, most methods only classify user clusters, but do not analyze how to classify users after classification. Carry out load forecasting; or propose a load forecasting method only for a certain user or a certain whole, without analyzing the load forecasting situation of a large number of users with different power consumption behaviors

Images