An improved ANN model for prediction of solar radiation using machine learning approach

Abstract

An accurate forecast of weather is essential for obtaining energy from Renewable sources. The objective of this paper is to present an analysis of weather parameters and comparison among different models of the weather prediction from accessible parameters and finally deriving a new technique for solar prediction in support of photovoltaic output power. Artificial Neural Network model with 5 weather parameters from NASA POWER dataset have been utilized to predict the day-ahead solar radiation and evaluated against real data measured for 4 years at Agartala, India (Latitude 23.83° N and Longitude 91.282° E). Results detailed in this work confirm the best predicting potential of the proposed method. The proposed model has been shown to predict solar radiation with accuracy of 83% shows the robustness of the system.