Select the most relevant input parameters using WEKA for models forecast Solar radiation based on Artificial Neural Networks

Abstract

Forecasting solar radiation is important for many applications in research related to renewable energy. Solar radiation is forecasted by solar radiation forecast models including the traditional models and artificial neural network (ANN) based model. There are geographical and meteorological variables that affect the solar radiation, thus identifying the appropriate variables to forecast solar radiation correctly is an important issue in the research area. Accordingly Waikato Environment for Knowledge Analysis (WEKA) Software was used in 11 points in Guilan based on different weather conditions to find the most effective input parameters to forecast solar radiation in different ANN models. Input parameters include latitude, longitude, maximum wind speed, average temperatures in each month, the average maximum air temperature, average minimum air temperature, sunshine, monthly rainfall, maximum rainfall in a day  for different cities of Gilan. In order to check the reliability of the forecasts by known parameters, three ANN models have developed (ANN-1, ANN-2 and ANN-3). The maximum MAPE for ANN-1, ANN-2 and ANN-3 equals 22.15%, 20.29% and 22.14%, respectively indicating 1.86% improvement in the accuracy in the prediction of ANN-2.