Statistical Wind Power Forecasting for U.S. Wind Farms: Preprint

Electricity markets in the United States are evolving. Accurate wind power forecasts are beneficial for wind plant operators, utility operators, and utility customers. An accurate forecast allows grid operators to schedule economically efficient generation to meet the demand of electrical customers. The evolving markets hold some form of auction for various forward markets, such as hour ahead or day ahead. This paper describes several statistical forecasting models that can be useful in hourahead markets. Although longer-term forecasting relies on numerical weather models, the statistical models used here focus on the short-term forecasts that can be useful in the hour-ahead markets. The purpose of the paper is not to develop forecasting models that can compete with commercially available models. Instead, we investigate the extent to which time-series analysis can improve simplistic persistence forecasts. This project applied a class of models known as autoregressive moving average (ARMA) models to both wind speed and wind power output. The ARMA approach was selected because it is a powerful, well-known time-series technique and has been used by the California Independent System Operator in some of its forecasting work. The results from wind farms in Minnesota, Iowa, and along the Washington-Oregon border indicate that statistical modeling can provide a significant improvement in wind forecasts compared to persistence forecasts.