A novel short-term wind speed prediction method based on hybrid statistical-artificial intelligence model with empirical wavelet transform and hyperparameter optimization

The short-term wind speed prediction is important for the connection of wind power to the grid. However, it is challenging due to the stochastic and intermittent characteristics of wind speed. In this paper, a novel wind speed prediction method named as EWT-ARIMA-LSSVM-GPR-DE-GWO method is proposed, where the empirical wavelet transform (EWT) is applied to decompose the original wind speed signal into multiple intrinsic mode functions (IMF), and various prediction models including the statistical model of autoregressive integrated moving average (ARIMA), artificial intelligence models of least squares support vector machine (LSSVM) and Gaussian process regression (GPR) are used for predicting the IMF in sequence from low frequency to high frequency based on the superiority of each model. Differential evolution-grey wolf optimizer (DE-GWO) algo-rithm is used to optimize the hyperparameters of artificial intelligence prediction models of LSSVM and GPR. The proposed short-term wind speed prediction method shows superior performance in both accuracy and stability with validation by the measurement data. Both the EWT signal decomposition and DE-GWO hyperparameter optimization are effective to promote the prediction accuracy for wind speeds with large non-stationary.