A Comprehensive Approach to Wind Power Forecasting Using Advanced Hybrid Neural Networks

Abstract

Wind power prediction is important in successfully integrating renewable energy sources into the grid. This study is focused on a sub-domain of wind power prediction and compares Bidirectional Long Short Term Memory (BiLSTM) and Bidirectional Gated Recurrent Unit (BiGRU) architectures. Additionally, these models are enhanced by advanced pre-processing techniques, including such methods as Discrete Wavelet Transform (DWT) and Fourier Synchrosqueezed Transform (FSST), as well as hybrid models involving Convolutional Neural Network (CNN) and Random Forest (RF) together with BiLSTM and BiGRU Models. It was found that the hybrid model consisting of CNN and BiGRU performed better than other hybrids by returning an R2 score of 0.9093, RMSE of 0.1095, MSE of 0.0120 and MAE of 0.0466; this shows that our model had a much greater level of accuracy compared to others ones developed before. These model performance indices demonstrated its better trustworthiness and error level for further utilization in wind energy forecast applications required for efficiency improvements and reliability enhancement in wind energy management. The current study emphasizes the usefulness of combining deep learning approaches like BiLSTM and BiGRU for more accurate wind power predictions hence improving reliability and effectiveness in managing wind energy resources.

Keywords

• Bidirectional long short term memory (BiLSTM)
• bidirectional gated recurrent unit (BiGRU)
• convolutional neural networks (CNN)
• Discrete Wavelet Transform (DWT)
• hybrid deep learning
• wind power