Development of a hybrid model for medium-term wind farm power output forecasting

Nowadays there are two main approaches to operational physical forecasting of wind farm power generation. The first relies on numerical weather predictions and technical characteristics of wind farms, while the second employs statistical techniques implemented using neural networks that analyze empirical data and establish correlations between weather conditions and wind farm output. However, each approach has limitations in accuracy when used separately. This paper presents a hybrid model of wind farm short-term power forecasting that combines the advantages of physical and statistical approaches. The developed model utilizes publicly available meteorological forecasts, adjusted specifically for the wind farm site. The model calculates wind speeds at each turbine, considering aerodynamic interactions among turbines. Actual operating data from wind turbines are used to estimate their power output. Thus, the developed hybrid model integrates operational data and numerical methods to improve the accuracy of power output forecasts under specific climatic and geographical conditions. The effectiveness of the proposed model was verified using the case study of the Adygeya wind farm in Russia. Results demonstrated high accuracy, confirming the model's capability to reliably account for turbine aerodynamic interactions, climatic and other factors. The proposed approach ensures more reliable wind farm operation planning, which is crucial for integrating renewable energy sources into power systems.