A unified forecasting model for solar irradiation and wind speed using ground-based data

Abstract

The objective of this study is to realize short-term forecasting of time-varying solar irradiation and wind speed through a unified neural-network-based algorithm. In previous work, a neural-network forecasting model for solar irradiation—benchmarked against the Smart Persistence Model based on the Clear Sky Index (CSI)—demonstrated stable and accurate performance. In the present study, this framework was extended to wind speed forecasting by introducing a new indicator, the Average Index (AI), which plays an analogous role to CSI. Solar forecasting was applied to the Kyushu region, while wind forecasting was conducted for Okinawa and Iwate Prefectures, where high wind-power potential exists. Using meteorological data from the Japan Meteorological Agency, the proposed method achieved high accuracy and demonstrated strong agreement with the Smart Persistence Model. The integration of AI into the neural-network system enabled reliable short-term forecasting of wind speed, confirming that solar irradiation and wind speed can both be predicted within a single, unified algorithm. This framework provides a robust foundation for multi-source renewable-energy forecasting and intelligent grid management.