Wind energy, as a renewable and green energy source with substantial value that is vital for sustainable human development, is gaining more and more attention around the world. The variability of wind implies that wind power is random, intermittent, and volatile. In order to overcome the unfavourable factors brought by wind power and enhance the reliable, stable, and secure operation of electrical grids that incorporate wind power systems, a multi-hour ahead wind power forecasting system consisting of an optimal combination of statistical, physical, and artificial intelligence (AI) models for real wind farm applications was proposed in this research.

Except for a direct persistence model that was able to produce wind power forecasts directly, an indirect persistence, an autoregressive integrated moving average (ARIMA), and a Weather Research and Forecasting (WRF) model were used to provide wind speed forecasts which, in turn, could be converted to wind power forecasts by using a power curve model. A technique for order of preference by similarity to ideal solution (TOPSIS) scheme was applied to construct a novel 5-in-1 (ensemble) WRF model for wind speed and wind power forecasting. An adaptive neuro-fuzzy inference system (ANFIS) model was employed to determine the power curve model, and another ANFIS model was utilised to build a wind speed correction model exclusively for correcting the wind speed forecasts provided by the 5-in-1 (ensemble) WRF model.

By using a set of 24-day historical wind speed and wind power measurements acquired from an operational wind turbine in a real wind farm located in North China, the multi-hour ahead wind power forecasting system was proposed comprising the following components over various forecast time horizons:​ the direct and indirect persistence models for 30-minute ahead forecasting, the ARIMA model for 1-hour ahead forecasting, and the WRF-TOPSIS model (with corrections obtained from the ANFIS-based wind speed correction model) for 1.5-hour to 24-hour (with a 30-minute temporal resolution) ahead forecasting. The primary contribution of this research is the novel WRF-TOPSIS model strategy used to select and combine the best-performing WRF models from a vast ensemble of possible models. The results demonstrated that the proposed multi-hour ahead wind power forecasting system has excellent predictive performance and is of practical relevance.