Numerical taxonomy method for STLF

Abstract

This paper proposes a numerical taxonomy-based method for short-term load forecasting. The method is formulated as an optimization problem using MAPE as an objective function. The data matrix and the cluster size are optimization variables. The accuracy of the method is demonstrated on a practical system data. Two models are developed using the proposed method. Model 1 uses two temperature variables and Model 2 uses two load and two temperature variables. Both models were compared with 35 input supervised backpropagation model and 6 input unsupervised/supervised radial basis function reported in literature for the data of the same system. It is observed that both the models gave better prediction than feedforward neural network and radial basis function neural networks.