Short-Term Load Forecasting Based On EEMD-LSSVM

As the power can not be stored in a large number,the power need to follow the dynamic changes with the load and make the timely adjustment.If not,it will affect the quality of power supply and the stable operation of the electric power system.In recent years,the research on new methods of load forecasting is developing.This paper adopts the prediction method of power load forecasting based on the ensmble empirical mode decomposition and least squares support vector machines.Considering the influence of date type,weather and other factors on load forecasting,this paper use fuzzy clustering method selected similar days ensmble which is integrated as an initial load signal to empirical mode decompose,obtained series of IMF component and a residual component.Then the least squares support vector machine predict of each component.Both components forecasting value are added as the final load forecasting value of the next moment.In order to solve the conventional empirical mode decomposition of aliasing problem,ensmble empirical mode decomposition is used,the original signal is added to the white noise;and the use of IMF selection stop condition that Huang proposed,to decompose the added noise signal.Taking into account the prediction accuracy,the bi-group artificial bee colony algorithm is be use to choose the parameter of the least squares support vector machin.Simulation results show that the optimized least squares support vector machin have better predictive ability.Using control variable method,simulation results show that: the higher prediction accuracy is given when we use the ensmble empirical mode decomposition combined least squares support vector machines;the smaller the predictable size is,the higher forecast accuracy is;the smaller historical data length is,the higher forecast accuracy is.