Support vector machines with similar day’s training sample application in short-term load forecasting

A novel method is proposed based on support vector machines using similar day´s load data as the training sample data for power load forecasting in this paper, where the power load has the specificity of periodicity and randomness. Analyzing the natural characteristic of power load and the forecasting day´s weather features and getting the calculating data for load forecasting. Quantifying the forecasting day´s weather features and calculating the similar degree between the similar days and forecasting day, choosing the most properly data as the training data for SVM according to the similar degree between them. The support vector machines is a novel learning method based on dimension and local minima and the new load forecasting model is based on SVM, whose parameters and kernel function is properly chosen for different conditions. And similar day´s training sample data for SVM can improve the degree of belief of original data and reduce the training time which is most important for load forecasting. The simulation results show that the novel method based on SVM using similar day´s training sample has faster speed, higher precision than other method and which proves that it is an effective method.