Estimation of hourly average solar radiation on tilted surface via ANNs

An innovative approach is proposed for the estimation of hourly average solar radiation on tilted surface. The proposed approach, which is based on artificial neural networks and problem decomposition, demonstrates the following characteristics: Accuracy. Superior estimation is attained over both conventional approaches and theoretical models. Simplicity and efficiency. A small training set is employed and the training/test patterns involve few easily obtainable parameters. Generalization capability and robustness to noise. Apart from being of interest in meteorology, the accurate and efficient estimation of hourly average solar radiation on tilted surface is especially important in solar energy applications.