An algorithm for wind turbine icing de-rated capacity prediction

Abstract During the winter of 2023-2024, Hunan province experienced four rounds of freezing disaster weather processes (Dec 15-20 th in 2023, Jan 18-25 th , Feb 1-11 th , and Feb 20-29 th in 2024). The maximum wind power de-rated capacity due to icing in the province reached 7850 MW (on Feb 25 in 2024), accounting for 82.3% of the total installed wind power capacity in Hunan province. Using a wind turbine icing and standby withdrawal prediction model, the average prediction accuracy rates for turbine icing de-rating across the four rounds were 83.8%, 78.4%, 83.3%, and 92.2%, respectively. Further error analysis of the prediction results revealed that the average temperature predicted is higher than the actual temperature, which may be the reason for the prediction error of wind power capacity. Therefore, using the Qiujialun wind farm as a sample, which had the largest prediction error, a temperature correction algorithm for individual turbines was established. The meteorological data at the model layer closest to the actual turbine hub altitude are used to reduce the deviation between the model altitude and the actual altitude in this algorithm, which is critical for mountainous wind farms like those in Hunan province. The improved algorithm was applied during the third and fourth rounds. The average absolute deviation (AAD) of the de-rating prediction for Qiujialun decreased from 22~24.1 MW in the first two rounds to 4.7~7.5 MW in the last two rounds, indicating improved prediction accuracy. Based on this correction method, recalculations were performed for the first two rounds. The absolute deviation for the first round decreased from 22 MW to 11.9 MW, and for the second round from 24.1 MW to 13.8 MW, representing an error reduction of approximately 50%.