Global Solar Radiation Modeling Using The Artificial Neural Network Technique

Global solar radiation is need knowledge for solar energy system design. In this work, the artificial neural networks (ANN) were applied to estimate the daily global solar radiation in China. Eight-year meteorological data from ten weather stations, which located at very different locations and climate zone, was randomly split into training, validation and test set with the proportion of 2:1:1. Daily Meteorological data (sunshine duration, air temperature, rainfall, relative humidity, and atmospheric pressure), geographical parameters (latitude, longitude, and altitude), and day of year (DOY) were used in the input layer of the ANN models. Twelve combinations of input variables were considered and the performance of the models was evaluated. The ANN model with all input variables achieve the best results (R² = 0.932; RMSE = 1.915 MJ · m^-2 · d^-1). Compared to the most widely used regression model, Angstrom formula, ANN models are more accuracy. The ANN model was applied to forecast the daily solar radiation at 12 independent stations and the performance was fairly good (R² > 0.85; RMSE < 3.4 MJ · m^-2 · d^-1). Results indicated that the ANN models show promising in daily global solar radiation estimation at the places where the radiation data is missing or not available.