A novel three-segment solar radiation forecasting model

The alteration of solar radiation data is dependent to stochastic and deterministic internal features. Deterministic feature is caused by solar geometry and movements in the nature whereas, the stochastic features are composed from random atmospheric events. It is considered that, focusing on these features separately will increase the accuracy of forecasting. Therefore, in this study, a new solar radiation estimation model was developed by considering these intrinsic features separately. While dealing the deterministic part, the hours without sunlight were determined using corresponding extraterrestrial solar information of region. On the other hand, the sudden changes in sun set and rise times are considered separately. Finally, a "similar pattern-based approach" was developed to capture stochastic changes. In this way a three-segmented prediction model is proposed. The accuracy of the proposed model has been tested and the accuracy results have been compared with other models in the literature in terms of different benchmarks. The results indicate that, proposed model can predict data quite accurately.