Development of ANN Based Model for Solar Potential Assessment Using Various Meteorological Parameters

Abstract

Abstract Solar potential assessment is very useful for various applications like solar heating, agriculture, solar lighting system and solar power plant erection etc. The objective of the current study is to identify theoretical potential of solar radiation for solar energy applications in hilly state Himachal Pradesh. Artificial Neural Network (ANN) is used to predict solar radiation using site specific measured data of Hamirpur for training and testing. The input variables used are temperature, rainfall, sunshine hours, humidity &barometric pressure to predict solar radiations. To identify the effect of various input parameters on solar radiations three ANN based models have been developed represented by ANN-I5, ANN-I4& ANN-I3.To obtain best prediction result, the number of input parameters of the input layer have been varied between 3 to 5 and hidden layer neuron have also been varied between 10 to 20. The best mean absolute percentage error (MAPE) calculated for these models (ANN-I5, ANN-I4 & ANN-I3) are 16.45%, 18.77% and 19.39% respectively. The ANN-I5 (temperature, humidity, barometric pressure, rainfall and sun shine hours), model has shown good prediction accuracy as compared to other two models. This study shows that various numbers of meteorological parameters mostly affect the forecasting of solar radiation. The method in this paper can also be used to identify the solar energy potential of any location worldwide where it is not possible to install direct measuring instrument.