Forecasting of the Daily Maximum Load Based on a Combined Model of Kohonen Clustering and SVM

Power system load is affected by a variety of factors including meteorological conditions. To improve the power system short-term prediction accuracy,this paper pre-sents a prediction algorithm combined with Kohonen clustering and SVM. The methodology takes into consideration of the characteristics of cyclical changes in power system, and chooses the similar days with the same characteristics in history by clustering the leaning sample and constitutes the training samples for the similar days,and prediction is conducted by the SVM model. The sample feature vector contains a variety of factors such as meteorological factors and holidays. The simulation has verified the effectiveness of the algorithm.