Short-Term Load Forecasting Using Wavelet De-noising Signal Processing Techniques

Abstract

Electrical load data are non-stationary and very uproarious on the grounds that an assortment of components influences power markets. The immediate prediction of electrical load with noisy information is typically subject to vast mistakes. This venture proposes a novel methodology for load forecasting by applying wavelet de-noising in a neural network models. The procedure of the proposed methodology initially deteriorates the chronicled information into an approximate part connected with low frequency and a detailed part connected with high frequencies through a wavelet transform (WT). A backpropagation neural network (BPNN) is built up by the low-frequency signal to estimate the future value. At last, the load is predicted by BPNN with and without utilizing WT. To assess the execution of the proposed methodology, the load data in New South Wales, Australia, are utilized as an illustrative model.