Prediction of hourly solar radiation in Tamil Nadu using ANN model with different learning algorithms

The solar radiation is a critical metric for design and optimal operation of the solar energy system. Most of the locations available solar radiation records are not available due to high cost for buying and maintaining the measuring instruments. Aim of this paper is building an efficient neural network model that can reliably estimate solar radiation. In this work, different ANN models with three popular algorithms admired from the literature are investigated. The models were trained using meteorological data collected over a year from six different places from the hot area locations of Tamil Nadu, India. Based on minimal Mean Absolute Deviation (MAD), Mean Square Error (MSE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE), the optimum ANN algorithm and model are determined (R). Furthermore, the research illustrates that the ANN model's prediction performance is dependent on the entire set of data utilised to train the network. Also, this article aims at finding out the accurate number of hidden layer neurons for the developed model. The proposed ANN model offers improved accuracy and applicability for estimating hourly average global radiation for the purpose of designing or evaluating photovoltaic (PV) installations in areas without meteorological data collection facilities.