Hourly day ahead wind speed forecasting based on a hybrid model of EEMD, CNN-Bi-LSTM embedded with GA optimization

In this paper, a novel hybrid model of decomposition and deep learning embedded with GA optimization was proposed to forecast 24-hour ahead wind speed. The historical wind speed time series was pre-processed and then decomposed into intrinsic mode functions (IMEs) using Ensemble Empirical Mode Decomposition. Each IMEs then was trained and tested through a models of CNN-Bidirectional LSTM model. The hyperparameters of the hybrid CNN-Bi-LSTM model was optimized using GA. CNN can extract the internal characteristics of the time series directly meanwhile Bi-LSTM network can utilize the information in both forward and backward directions completely. The forecasting results of each IMEs were reconstructed to obtain the final forecast. The proposed method was applied to real WS dataset in Hanoi compared with 6 other methods. The result shows that the proposed method has demonstrated much better performance than the other methods. (C) 2022 The Author(s). Published by Elsevier Ltd.

Automated summary

In this paper, a novel hybrid model combining decomposition and deep learning, embedded with GA optimization, was proposed for wind speed forecasting. By decomposing and training the historical wind speed time series, better forecasting results than other methods were obtained.