A new short-term load forecasting model based on relevance vector machine

Considering the limitation of traditional feature extracting only the algebraic features of samples to the neglect of the practical significance of the original problem, a short-term load forecasting model based on the relevance vector machine (RVM) is proposed. By using the nonnegative matrix factorization (NMF) algorithm, the dimension of input variables is reduced, a short-term load forecasting model based on the RVM is proposed. The input data is decomposed by using the NMF algorithm, where the nonnegative lower-dimension mapping matrix derived is taken as the input of RVM for training and predicting. Due to the nonnegative property of the lower-dimension matrix, it retains the practical significance of the original problem while eliminating the redundant data and reducing dimensions. Simulation results show that the dimensions of the input variables can be effectively reduced and the predicting accuracy can be greatly improved.