An efficacious model for predicting icing-induced energy loss for wind turbines. (1st January 2022)

• Record Type: Journal Article
• Authors: Swenson, Lauren; Gao, Linyue; Hong, Jiarong; Shen, Lian
• Is Part Of: Applied energy. Volume 305(2022)
• Publication Date: 2022-01-01
• Subjects: Wind turbine icing -- Meteorological icing -- Weather Research and Forecasting (WRF) -- Numerical weather prediction (NWP) -- Energy loss; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2021.117809
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: We propose a fast and efficacious framework for wind farm icing loss forecasts. This approach bridges meteorological icing and turbine icing-induced energy loss. This approach reduces data and computation requirements for icing loss forecasts. Abstract: The wind industry in cold climates has shown strong growth in recent years, but turbine icing in these regions can cause significant energy loss leading to a reduction in reliability of wind energy. Previous studies on estimating wind turbine icing (WTI) generally rely on complex physical models, and many only model the ice growth itself while failing to correlate ice growth with energy loss. It is the estimation of icing-induced energy loss that is critical for power grid management to cope with energy deficits associated with extreme weather conditions. This study focuses on bridging this modeling gap through developing an efficacious methodology for predicting icing-induced energy losses for wind turbines in cold weather events. Specifically, this study uses measurements of 11 WTI events between 2018 and 2020 from a 2.5 MW wind turbine (Eolos site, University of Minnesota) to create a statistical correlation between meteorological conditions and icing-induced energy loss. Meteorological icing parameters generated from a Weather Research and Forecasting simulation are used as inputs to the model. The model is validated against in-situ data for all events, and against two additional 1.65 MW wind turbines for one event (Morris site, University of Minnesota). When comparing average estimated energy loss to measured loss, it shows a relative mean absolute error of 37% at Eolos and 2.9% at Morris (after power curve scaling). The new model is additionally implemented for 30 large-scale wind farms in the Midwest region of the United States for estimation of WTI energy loss. The method proposed in this study enables fast and accurate prediction of WTI energy loss for wind turbines.