Verification of wind power forecasts provided in real-time to the Irish Transmission System Operator

• Main Authors: Lang, S, McKeogh, E
• Format: Conference Proceeding
• Language: English
• Subjects: Error analysis; Forecasting; Power systems; Predictive models; weather forecasting; Wind farms; Wind forecasting; Wind power generation
• ISSN: 1932-5517
• DOI: 10.1109/PES.2010.5590032

Wind power forecasts for 21 representative wind farms provided in real-time to the Transmission System Operator in Ireland have for the first time been verified against actual wind generation data. Forecast data were generated using an ensemble prediction system. For forecast lengths up to 23 hours, the mean absolute error for individual wind farms was typically about 13% (normalised to installed capacity). The error increases to ~ 14% for day-ahead forecasts (24 - 48 hours), and to ~ 17% for 2 - 3 day forecasts. The error growth for the aggregated total of the wind farms was substantially lower, increasing from 8% (day-ahead) to 10% (2 - 3 days). Forecasting for the aggregate of 51 wind farms only reduced these errors by 1%. Using 21 representative wind farms is therefore a reasonable balance between up-scaling from a smaller number of wind farms, and the lower errors that result from monitoring and forecasting a much larger number.