Load forecasting based on self-organizing map and support vector machines

Forecasting of future load demand is very important for decision making in system operation and planning. This paper presents a load forecasting model based on SOM (self-organizing map) and SVM (support vector machine). SOM is used as a clustering tool to divide the training data into subsets with different centers. SVM is used to fit the testing data based on the clustering subsets for predicting. Besides, the input vectors of the multi-step forecasting are constructed with virtual forecasted values that substitute for real values, and the genetic algorithm is used for SVM parameter optimization. The proposed model was tested on EUNITE competition data to predict the month-ahead electricity load, and the result illustrates the effectiveness and efficiency of clustering and prediction model.