A novel combination model for ultra-short-term wind speed prediction

Accurate and rapid ultra-short-term wind speed prediction is crucial for wind farm operation control. This paper proposes a wind speed prediction model that integrates Variational Mode Decomposition, Sparrow Search Algorithm, and Long Short-Term Memory to address the issue of insufficient precision in current ultra-short-term forecasts. To tackle the complexity of wind speed datasets and the challenge of selecting parameters for Variational Mode Decomposition, the Archimedean Optimization Algorithm is employed to optimize modal component values and the penalty factor, enabling adaptive selection of these parameters. Additionally, the Sparrow Search Algorithm is utilized to optimize parameters for Long Short-Term Memory, including the number of hidden layer units, iteration count, and initial learning rate. The combined model of Variational Mode Decomposition, Sparrow Search Algorithm, and Long Short-Term Memory accurately forecasts ultra-short-term wind speed based on historical data. Validate the prediction model with wind speed data from the National Data Buoy Center (NDBC) in the United States. The experimental results demonstrate that this model achieves a prediction accuracy of approximately 97%, providing a novel approach to ultra-short-term wind speed forecasting in wind farms.