The Short-term Power Load Forcasting Based on Genetic Algorithm Optimized LS-SVM
WU Yan-zeng · Journal of Lanzhou Jiaotong University · 2012
In order to resolve the disadvantages of traditional methods on selecting LS-SVM regression model parameters,the GA+LS-SVM(Genetic Algorithm Least Squares Support Vector Machine)model is established based on the analysis of GA which could get the best parameters in the progress.The data of power load in this paper is obtained from ISO New England during 2005～2006,and the train data which is choosen to establish the prediction model is from 2005-01-01 to 2005-12-31 daily 24-point power load data.By this method,GA+LS-SVM load forcast model is established by taking model input of historical loading,temperature and humidity,and the data of the same hour in the last day and in last week to forcast the 168 point power load in the first week in January 2006.At the same time,BP,SVM,LS-SVM prediction model is also established.Analysis of the experimental result proves that LS-SVM optimization by using GA can achieve greater accuracy than BP,SVM,LS-SVM.