Analog Bias Correction of Numerical Model on Wind Power Prediction

A new post-processing method is proposed to reduce numerical weather prediction's systematic and random errors. The method overcomes a difficulty of a post-processing algorithm inspired by Kalman filtering and a 7-day running-mean correction in dealing with sudden changes of the forecast error that could be caused by rapid weather transitions. The analog forecast for a given location and time is defined as a past prediction that matches selected features of the current forecast. The method is the weighted average of observations that verifies when the best analogs are valid. The method is tested for 70-m wind speed prediction from Weather Research and Forecasting (WRF) model, with observations from one wind farm sited at Yanchang, Shaanxi Province for 3 months.The analog bias correction method is able to produce skillful corrections of the raw forecasts, even with large day-to-day changes in forecast error, and thus the method can predict drastic changes in forecast error. Moreover, being a prediction based solely on observations, it results in an efficient downscaling procedure that eliminates representativeness discrepancies between observations and predictions.Also, it is able to reduce random errors, therefore improving the predictive skill of raw forecast. The correction method is much better, with average improvement of 9.3% and 9.8% measured by root mean square error (RMSE) and centered root mean square error (CRMSE), respectively. Meanwhile, the method shows a better pattern of correspondence between predictions and observations.Moreover, the correction method for middle wind speed (5—12 m·s^-1), which plays the most important role on wind power prediction, is much better, with average improvement of 12.3% and 21.7% measured by RMSE and CRMSE, respectively. Thus the analog bias correction method is very suitable for wind power prediction.The analog bias correction method is based purely on verifying observations of past predictions that are similar to the forecast (i.e., the analogs), which provide physically based insight about the atmospheric state, thus improving the predictive skill. And it also has the potential to be applied to other prediction systems and variables.