Electricity load forecasting using hybrid wavelet neural network based on parallel prediction method

This paper presents a new hybrid load forecast model to improve the accuracy and robustness of load profile forecasting (1-24 hours ahead). It comprises of Wavelet transform and Neural network based on parallel prediction method, which is called 'PWNN'. Wavelet transform is used to decompose the original load series into multiple load sub-series with different frequencies. Then, neural network is used to predict each load sub-series using parallel prediction method. The load forecast can be obtained by inverse wavelet transform. The results indicate that PWNN has a significant improvement of accuracy and robustness in load forecasting over other models used for comparison in this study. © 2016 IEEE.