Solar irradiance forecasting using wavelet neural network

Abstract

Short-term solar power forecasting is vital for reliable and secure operation of power systems with high PV penetration. This paper implements wavelet neural network (WNN) with Levenberg-Marquardt (LM) training for solar irradiance forecasting for finding the solar power output. It employs wavelets basis as activation functions whose shapes are adaptive in nature. The proposed model has better generalization capability and more accuracy than the conventional sigmoidal neural network (SNN). The outcomes demonstrate that the model can be implemented easily and can enhance the forecasting accuracy.