Modern day grid reliability and security are highly dependent on accurate wind speed and power forecasts. Whereas the nonlinear nature of wind speed poses challenges in forecasting via traditional methods, machine learning-based hybrid models adequately address this issue. In this chapter, we use a hybrid model based on wavelet decomposition technique and several variants of SVR. To determine the best model, performance metrics are calculated for real-time wind speed datasets. ε-twin support vector regression (ε-TSVR) and twin support vector regression (TSVR) are used to forecast short-term wind speed in tandem with classical SVR and LSSVR, and its performance is compared with the benchmark persistence model. We also evaluate the performance of these models for a larger dataset based on wind sites in USA and India.
Supervised Machine Learning in Wind Forecasting and Ramp Event Prediction, Wind Energy Engineering, 2020, Pages 75-99,