Powering the Future: Wind Power Forecasts for Solano, California

Alternative energy sources have become increasingly important. The drawback to these sources is they do not provide consistent amounts of power. For successful power grid management, utility companies need accurate power production forecasts at a variety of horizons. Wind power depends primarily on wind speed and to a lesser extent on wind direction, barometric pressure, and temperature. This talk will explore parametric, semi-parametric, non-parametric, and combination models to create wind power forecasts for a variety of horizons. Due to natural restrictions on power values, we consider a link function that relates the response, power, to its predictors. These models were fit to data collected every ten minutes during all seasons. Forecasts were created in ten minutes intervals for horizons up to six hours. The results of these model fits will be discussed. In addition, the talk will address the need for accurate prediction intervals and the corresponding difficulties.