Performance Evaluation of Hybrid Wind Speed Prediction in Response to Rapid Changes

Accurate wind speed forecasting is crucial for optimizing power generation, improving grid stability, and ensuring efficient wind farm operations. In this study, we used hybrid prediction methods using Kernel Extreme Learning Machine (KELM), Long Short-Term Memory (LSTM), and Support Vector Regression (SVR) algorithms to predict wind speed, as well as particle swarm optimization (PSO) and differential evolution (DE) as parameter optimizers. Unlike traditional evaluation based solely on global accuracy, our investigation extends to examining the behavior of these algorithms under different wind conditions, particularly rapid changes in wind speed. The results highlight the critical impact of these variations, revealing a degradation of the algorithm's performance during sudden changes in wind speed. These results provide valuable information to improve the reliability and robustness of wind speed prediction algorithms under dynamic environmental conditions.