Research On Short-Term Power Load Forecasting Based On VMD And Improved Random Forest

With the rapid development of the power industry and social economy,the country has formulated the thirteenth five-year plan according to the development requirements of national conditions.The construction of new industrialization,the integration of urban and rural areas,and the modernization of rural agriculture will be continuously promoted and gradually improved.While promoting the rapid development of various economic systems,electric energy has become an indispensable and important element in people’s production and life.In the process of promoting the construction of a strong smart grid in the State Grid,load forecasting is related to the optimal distribution of energy and reasonable dispatch of electric energy.Therefore,it is essential to improve the accuracy of load forecasting for the operation of power system.For the purpose of realizing the short-term power load forecasting by using an intelligent forecasting model,the random forest(RF)regression algorithm is used to forecast the short-term power load in this paper.The gray wolf optimization(GWO)algorithm is applied to improve the RF regression model to find out the best decision trees and split characteristic numbers,which enhance the accuracy of power load forecasting.On this basis,it is attempted to improve the GWO algorithm by using chaotic sequence,which improves the convergence speed of conventional functions and the ability to find optimal adaptability values,however,the prediction accuracy is almost equal to the prediction accuracy before the GWO algorithm is not improved.In view of the problem of further improving the prediction accuracy of RF regression model,this paper proposes a method of short-term power load accurate prediction based on variational mode decomposition(VMD)and GWO optimized random forest regression model.Firstly,VMD is used to decompose the actual short-term load data to obtain multiple groups of modal function components with different characteristics.Then,GWO algorithm is used to optimize the best decision trees and split characteristic numbers in the RF regression model,and each group of modal function components is predicted intelligently and optimally.Finally,the predicted modal function is reconstructed to obtain the final prediction results.The feasibility of the proposed VMD-GWO-RF is illustrated by the example results to predict the short-term power load,and the prediction accuracy are significantly better than the traditional RF algorithm.