Predicting The Monthly Average Incident Shortwave Solar Energy for Hubli, India by Using Training Functions in ANN

Abstract

Solar radiation is one of the vital resources found on Earth which can be renewed and hence tested and tried to be beneficiary for humankind. Solar energy is harnessed to fulfill the basic requirements of humans i.e.; supply power to operate any kind of machine or device. The way to utilize the energy for our maximum benefit is by approximating the radiation values of Sun and this can be achieved by installing measuring equipments. The main issue arises here as the equipment's maintenance and installation cost is too high to be affordable by the general people. To overcome this inconvenience, an affordable solution was developing models and methods to calculate the radiation and find the approximate values. We focus on city, Hubli, India and estimate the monthly mean radiation received on this particular city by creating a neural network in ANN (Artificial Neural network) using MATLAB. The models are validated for 3 training functions: resilient back-propagation (RP), Scaled Conjugate Gradient (SCG) and Levenberg-Marquardt (LM). The predicted values accuracy is also tested through statistical indicators like MSE, RMSE, MBE and MAPE.