A short-term power load forecasting method based on emd-gru based on feature selection

Abstract

The example of the present invention provides a kind of EMD-GRU short-term power load forecasting method based on feature selection, comprising: using the empirical mode decomposition method (EMD) to decompose the original load sequence into a plurality of time series components, and all time series components are used as initial The feature set, which constitutes the potential input variables of the forecasting model; the correlation analysis of the initial features is carried out through the Pearson correlation coefficient method, and the time series component with a greater correlation with the original load sequence is selected as the input feature of the forecasting model; it will be The selected time series components combined with the original load sequence are input into the gated recurrent unit network (GRU) forecasting model to perform the final load forecasting. According to the technical solutions provided by the embodiments of the present invention, the accuracy of short-term power load forecasting can be improved.

Description

【Technical field】 [0001] The invention relates to an EMD-GRU short-term power load forecasting method based on feature selection, which belongs to the short-term load forecasting method in the field of machine learning. 【Background technique】 [0002] In the field of short-term power load forecasting, forecasting methods are mainly divided into traditional methods based on statistics, artificial intelligence methods based on machine learning, and combined forecasting methods. Traditional methods include multiple linear regression method, time series method, exponential smoothing method, etc. These methods generally require a clear mathematical model to give the relationship between load and input factors. The stationarity requirements of the series are relatively high, which cannot accurately reflect the nonlinear characteristics of the load data. Machine learning methods include artificial neural network (ANN), support vector machine (SVM), random forest (RF), etc. Althoug...

AI Technical Summary

Problems solved by technology

The combined forecasting method includes model combination on the forecasting mechanism and weighted combination on the forecasting results. By concentrating the advantages of various methods, a more adaptable method is formed. The signal decomposition method is used to decompose different load components from the original load sequence. Analyzing and modeling can effectively improve the prediction accuracy, but directly modeling and predicting the multiple time series components obtained by decomposing will introduce multiple random errors and generate a large amount of forecasting workload. Consider introducing the Pearson correlation coefficient method to decompose the time series components Time-series components are selected for feature selection, and the time-series components with greater correlation with the original load sequence are selected and input into the GRU prediction model together with the original load sequence to perform the final load prediction

Images