Short-Term Wind Power Forecasting Experiment Based on WRF Model and Adapting Partial Least Square Regression Method

Based on the Weather Research and Forecasting Model(WRF) with high temporal and spatial resolutions and the Adapting Partial Least Square Regression(APLSR) method,the short-term wind power forecast system coupled with the wind field and wind power forecasting models is developed in this paper.In order to assess the accuracy of wind field forecasts objectively,the wind speed and wind direction in the specified wind field in Gansu Province in January,April,July,and October 2009 were forecasted by WRF model,which were used to compare with the observed data at 50 m and 70 m heights on two wind masts closed to the wind field.Based on the relative accurate forecasts of wind field,the nonlinear statistic forecast models every 15 min for 200 wind turbines using APLSR method and single wind turbine technique were constructed based on the actual wind power recorded data and wind speed,wind direction,temperature,relative humidity and pressure forecasting values on the hub height from January 2008 to April 2009.So as to assess forecasting effects of wind power,forecast experiments from January to December 2008 were carried out.The results showed that:(1) The probability distributions of wind direction forecasts are consistent with the observed values in January,April,July,and October 2009,and the forecasting effects of static wind is also good.(2) The correlation coefficients between the wind speed forecasts and observed values at 50 m and 70 m heights on two masts in January,April,July,and October 2009 are 0.6~0.8.Root mean square error(RMSE) of forecasting error of wind speed is between 1.5~2.6 m·s-1.And WRF model can well forecast the diurnal variation characteristic of wind speed.(3) The correlation between total wind power forecasts and actual wind power recorded values for the wind field every 15 min from January to December 2008 are remarkable.Correlation coefficients are 0.58~0.90 and have passed the confident level of 99.9%.(4) The forecasting errors of total wind power every 15 min from January to December 2008 which compared with the total rated installed capacity are relatively small and RMSE of forecasting errors of total wind power are 2.76%~12.89%.