A LASSO-GRBFN-based Method for Electricity Price Forecasting with BSO

Abstract

This paper proposes a new method for one-step ahead electricity price forecasting. It is based on GRBFN (Generalized Radial Basis Function Network) that is an extension of RBFN of Artificial Neural Network (ANN). GRBFN has advantage over RBFN that the Gaussian function parameters are evaluated by the learning process. The conventional ANN methods consider overfitting with the weight decay method that corresponds to the L2 norm of weights between neurons, but there is still room for improvement. According to the idea of the sparse modeling, this paper proposes the use of Least Absolute Shrinkage and Selection Operator (LASSO) to improve the model performance. Also, this paper presents BSO of evolutionary computation to evaluate the cost function with the term of the L1 norm. That is because the conventional methods with the gradient do not work for the L1 norm. The proposed method is successfully applied to real data of ISO New England in USA.