The Impact of Public Meteorological Observations on Short Term Regional Wind Power Forecasting

In this study, our hypothesis is that short term (hours ahead) wind energy forecast accuracy may be improved for a region-wide aggregate by incorporating meteorological station information (not necessarily co-located at a wind project) into the forecast model. Networks of meteorological measuring stations that report on an hourly basis (or more frequently) are operating across the United States. Many of these observations are well calibrated and used by government weather agencies for surface weather analyses and Numerical Weather Prediction model data assimilation. These measurements are largely untapped for short-term wind energy forecasting due to various reasons including data quality concerns, timeliness, and low correlation to turbine hub height winds. Previous studies have demonstrated the value of off-site measurements for improving site-specific short-term forecast accuracy (e.g, Westrick and Larson (2005), Gneiting et. al. (2006), Hering and Genton (2009)). We use a similar approach to predict aggregate power, but over entire regions.. We choose several regions within the U.S. where the aggregate wind power production is reported on a frequent and regular basis. Each region is characterized by a unique hub-height wind climatology which is one of several factors that may contribute to the potential value public met observations add for improving the statistically-trained wind power forecasts.