Research On Short-term Power Load Forecasting Based On HHT

The short-term power load forecasting has important significance to economicdispatch, real-time control, operating plan and development plans, and it is helpful toimprove the utilization rate of power generation equipment and the effectiveness ofeconomic operation. With the increase of electric power industry’s importance,thedeepening of electric power market’s reform and the introduction of competitionmechanism,the accuracy of short-term load forecasting has gotten more and moreattention.Hilbert-Huang Transform (HHT) is suitable for processing the nonlinear andnon-stationary signal, and it can get the signal’s time-frequency distributioncharacteristics.So HHT is with a fully adaptability. In the field of power system, HHThas been widely applied in the detection of power quality, harmonic analysis, etc., andachieved a good effect.This paper firstly introduces the characteristics of the electricity power system’smain users, the main influencing factors of the short-term power load and thestatistical indicators of forecasting error analysis.Then the artificial neuralnetwork,least squares support vector machine and other methods have been carried onthe thorough analysis. A continuation method based on BP neural network wasproposesd to suppress the end effect in HHT. Firstly pretreat the original loaddata,and quantify the temperature, weather type and date type. Then by empiricalmode decomposition(EMD),the load series can be decomposed to several intrinsicmode functions(IMF) with different frequencies. Then by hilbert transform(HT), theaverage instantaneous frequency can be gotten.According to the characteristics ofevery IMF, the selection of forecasting model is different. Finally the ultimateforecasting value can be obtained by superposited the forecasting results of every IMF.In this paper,the actual load data of Hefei in Anhui Province in2012is used as thesample to set prediction model. The evaluation indexes are choosen as mean absolutepercentage error and relational degree. In order to verify the accuracy of HHT based on BP neural network continuated method, this paper uses the RBF neural network,LS-SVM, basic HHT algorithm to predict at the same time.The results show that thealgorithm of HHT based on BP neural network continuated method has higherprediction accuracy.