Medium-Range Wind Power Ensemble Forecasting for Texas

Generation of wind power is subject to significant variability due to the intermittent nature of weather conditions. Increases of wind power penetration into energy grids and decreasing reserve margins has created the need for forecast information with projection windows beyond a few days, particularly during significant weather events. The utility of daily 100 m wind forecasts made by the European Centre for Medium-Range Weather Forecasting (ECMWF) for estimating power generation from wind turbines in the state of Texas is investigated. The 15-day Variable Ensemble Prediction System (VarEPS) forecasts during 2011-2012 are compared to operational analyses and observations from sonic detection and ranging (SODAR) towers used in the Wind Forecast Improvement Project (WFIP). Statistical methods are used to compensate for diurnal model biases and seasonal wind variability at extended lead times. Historical power forecasts at the model grid point level (~25 km) are generated based on power curve extrapolations. These forecasts are spatially aggregated for weather zones for the Electric Reliability Council of Texas (ERCOT) grid and evaluated versus observed power records to explore the effects of calibration on the ensemble forecast product. The ability of the ensemble forecast system to identify ramping and cutoff events is assessed.