Modeling of Monthly Mean Solar Energy Potential using Artificial Neural Network

The aim of this study is to develop an artificial neural network (ANN) model for accurately predicting monthly mean solar radiation and irradiance for Mersin (36.8o N, 34.6o E, Türkiye). The prediction of monthly mean solar radiation and irradiance was made by using two different ANN (NN-1 and NN-2) models with different input parameters and thus, a dual solution strategy for the monthly mean solar radiation and irradiance forecasts was presented. The ANN models were trained for the target parameters (monthly mean solar radiation and irradiance) at each month of the year. The training, testing and validating for both models were conducted using the data obtained for the period from 2004 to 2024. The performance results of these alternative models compared with each other. The accuracy of the models to predict the monthly mean solar radiation and irradiance are identified based on root mean square errors (RMSE) and cross-correlation coefficients (R). The NN-2 model has smaller RMSE values and has bigger R values. That is, the NN-2 model has higher prediction success with lower prediction error for both monthly mean solar radiation and irradiance intensity. The presence of two models may be advantageous for more precise forecasting situations and the NN-2 model can be chosen for such cases. In addition, the application of the NN-2 model proposed in this study can be extended to other locations.