Solar radiation estimation via neural network

A study was conducted to determine if a neural network could adequately estimate the solar radiation for a location that does not have suitable colection equipment. The traditional method for estimating solar radiation employs deterministic equations and computes equation coefficients to fit the local conditions. Typically, regression analysis of temperature and precipitation is used. In contrast, this study used actual weather data collected in Gainesvilie, FL, to train a back propagation neural network to estimating solar radiation. Results indicated that the neural network performed better than the analytical approach. The correlation coefficient (r) for the nerual network was 0.86 versus 0.77 for the analytical approach. Thus, the neural network allowed users to estimate solar radiation using simple weather data.