Short-term load forecasting (STLF) is the precondition of economic and secure operation of power systems. With power systems getting more and more marketable, STLF with high quality is getting more and more important and exigent. This paper is concerned with the application of Support Vector Machine (SVM) to STLF in order to improve the accuracy of load forecasting.Firstly, the paper does much research on the characteristic of electrical load, which offers reasons to the features selection of SVM and the modeling of load. Then the load forecasting model of SVM which parameters are determined by across-validation is studied, and its such defects as hard sledding, time-consuming, blindness, unable to select the parameters automatically and so on are found. The influence of features selection of samples to forecast accuracy is also studied, and the result indicates that when such information of climate, calendar, and past load included, the accuracy is better, but the training time does not get longer. As to important holidays, for which the model of SVM is not fit their load forecasting, the paper presents a new method of adding yearly load increases, which is simple but accurate.To aim at the defects of parameters selection of SVM with across validation, finally, the paper researches the problem of automatically selection of SVM parameters. A short-term load forecasting model based on SVM is presented in which the parameters in SVM are optimized by Particle Swarm Optimizer (PSO). The experimental results prove its more excellent accuracy than ordinary SVM load forecasting method.
|