Recurrent Wavelet Network with New Initialization and its Application on Short-Term Load Forecasting

A key issue in intelligent demand-side management is the accurate prediction of electricity consumption. This paper presents a dynamic model for short-term special days load forecasting which uses a recurrent wavelet network (RWN). However, initialization of this network encounters a major problem. Thus, a new initialization method is suggested based on orthogonal least square (OLS) technique. Finally, a RWN with the proposed initialization method is applied to experimental special days load data. Simulation results show that the proposed network is capable of handling the inherent complexity of load forecasting problem.