Single Step Ahead Assessment of Solar Irradiation Using Ann Model Based on Various Combination Of Meterological Parameters

Abstract

Solar energy is a valuable resource on earth but the availability of solar resources relies on meteorological variables. In this paper, forecasting models using the artificial neural network are developed by changing the input meteorological variables from five to seven. The two years data are used to train the model whereas the testing is performed using one year data on different seasons following single step ahead. The input parameters are relative humidity, pressure, temperature, solar zenith angle, wind speed, wind direction and perceptible water. Three artificial neural network models (ANN-I5, ANN-I6, ANN-I7) are developed to estimate the global horizontal irradiation and performance of all models are measured on the basis of Mean Absolute Percentage Error (MAPE), Relative Root Mean Square Error (RRMSE) and Correlation Coefficient (R2). Result indicates that ANN-I7 shown better performance as comparison to other developed models. The average MAPE and RRMSE of ANN models such as ANN-I7, ANN-I6,  ANN-I5 are 14.52%, 16.53%, 18.97% and 20.74%, 22.28%, 24.43% respectively. The ANN-I7 haiving an input meteorological parameters relative humidity, pressure, temperature, solar zenith angle, wind speed, wind direction and perceptible water showed good accuracy as comparison to other two developed models. This study indicates that accuracy of solar irradiation forecasting depends on meteorological parameters.