EEMD-based Wind Speed Forecasting system using Bidirectional LSTM networks

Abstract

Wind energy is enticing attention worldwide due to its renewable nature. For the stable functioning of wind turbines in wind power generation, wind speed needs to be predicted accurately. However, accurate wind forecasting is a challenge due to its flexible and intermittent nature. The proposed system merges the features of Ensemble Empirical Mode Decomposition (EEMD) and Bidirectional Long Short Term Memory (BiDLSTM) networks to forecast wind speed. Presently, Data Denoising models are comprehensively applied for forecasting wind speed to enhance the forecast accuracy of the systems. Hence, in this paper, EEMD is employed to divide the input wind data into many high and low-frequency signals. Bidirectional LSTM networks are employed to predict the high and low-frequency subseries separately, and the forecasting outcomes of each subseries are combined to get the ultimate outcomes. The simulation outcomes confirm that the proposed EEMD-based hybrid system outperforms the models used for comparison in terms of accuracy.