A Hybrid LSTM-LightGBM Model for Precise Short-Term Wind Power Forecasting

Authors

Md Omer Faruque, Dhaka University of Engineering and Technology, Gazipur
Md Alamgir Hossain, Griffith University
S. M.Mahfuz Alam, Dhaka University of Engineering and Technology, Gazipur
Md Rashidul Islam, Rajshahi University of Engineering and Technology
Md Rabiul Islam, University of Wollongong
Youguang Guo, University of Technology Sydney

Publication Name

2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023

Abstract

This paper proposes a novel hybrid model that combines long short-term memory (LSTM) with light gradient boosting (LightGBM) to achieve precise onshore wind power prediction in Australia. The intermittent nature of renewable energy sources like wind power necessitates accurate forecasting to minimize operational costs and enhance power system reliability and security. The hybrid model showcases promising results in forecasting 15-minute interval short-term wind generation data. Performance is evaluated using root mean squared errors (RMSE) and means absolute errors (MAE), and compares the proposed model with neural network (NN), gated recurrent unit (GRU), and standalone LSTM. The comparative analysis demonstrates the superiority of the proposed hybrid model, showing lower error indices with exceptional forecasting capability. This hybrid LSTM-LightGBM model holds great potential for optimizing renewable energy integration into power systems, facilitating cost-effective and reliable energy generation.

Open Access Status

This publication is not available as open access