Extended Abstract (364K)1.3
Statistical Wind Power Forecasting for U.S. Wind Farms
Michael Milligan, National Renewable Energy Laboratory, Golden, CO; and M. Schwartz and Y. H. Wan
The benefits of accurate wind power generation forecasts for wind plant operators, utility operators, and utility customers are becoming apparent as wind energy penetration increases on the United States grid. Accurate forecasts make it possible for grid operators to schedule the economically efficient generation to meet the demand of electrical customers. Hour and day ahead forecast periods are common periods of interest for organizations in the evolving electricity markets. Statistical forecasting techniques are relatively inexpensive and can be used quite effectively for short-term (one to six hours) forecasts. ARMA models, a powerful and well-known time series technique already used by some power system operators for their forecasting work were applied to both wind speed and wind power output recorded at operating wind power plants (wind farms) in several areas of the United States. The initial results from wind farms in Iowa and Minnesota indicate that using ARMA model can result in a significant improvement over persistence forecasts at short-time frames. However, the results also indicate that different ARMA models work the best in different months and that seasonal influences may also have a large bearing on which ARMA models to use for forecasts at certain times of the year. The results from the Iowa and Minnesota wind farms will also be compared to results from wind farms in the northwestern United States. Finally, we will present some recommendations for future work on statistical wind generation forecasts, including possible use of ensemble ARMA models for short-term forecasts.
Session 1, Forecast Systems (Room 602/603)
Monday, 12 January 2004, 9:00 AM-4:30 PM, Room 602/603
Previous paper Next paper
Recorded presentation