Impact of the accuracy of NWP wind speed forecasts on wind power forecasting

Accurate wind power forecasting (WPF) is an effective way to alleviate the negative effects brought by the integration of large-scale wind power into electricity grid. Most short-term WPF approaches take the mesoscale Numerical Weather Prediction (NWP) results as input to achieve high accuracy. However, the deviation of input NWP wind speed (WS) becomes a major source of WPF error, and the deviation of NWP WS are amplified at different levels in different WPF processes. Therefore, knowing about the impacts of NWP WS accuracy on WPF results will contribute to the improvement of WPF system. Two WPF models based on statistical algorithm and physical Computational Fluid Dynamics method were established respectively. The low-fidelity mesoscale NWP WS, high-fidelity measured WS and medium-fidelity corrected NWP WS with the same time-stamps were separately adopted as the input of WPF models. Via the error analysis of various combinations of input data and forecasting models, results show that the accuracy of forecasting power is proportional to that of input WS. Compared with taking the low-fidelity WS as input, the medium-fidelity inputs could reduce the annual forecasting error by 11% and 10.3% in average for statistical and physical models separately, while more than 15% for the high-fidelity inputs.