The Study Of Power Load Forecast Based On Support Vector Machine

Power load forecast is the process that using its past value and some relatedinformation to forecast its future value, and this process is considered to be afundamental work. For a long time, people have studied on this process much andcreated many effective method. Power load forecast method can be divided into twotypes, one is certain method and the other is uncertain method. Certain method is baseon traditional mathematic tools, through which the load value and the variable relatedone by one. Uncertain method considers lots of factors that will affect the load value,and it is the research focus now and future.Power load forecast method based on support vector machine (SVM) is one ofthe uncertain methods, which belong to the human intelligence area. SVM is the mostimportant achievement in the area of machine learning these years. It has manydominant characteristic such as global optimization, preventing over training, goodcalculating speed and accuracy and so on.This paper discussed how to use SVM into load forecast and how to considerrelevant environmental factors particularly, and created the detailed process of thismethod. We also give some real examples on short-term forecast and long-termforecast with this method, and analyzed the impacts of parameters, environmentalfactors and raw date. The accuracy of predict results were calculated and theapplication effect of this method was evaluated.