RLIDT: A novel reinforcement learning-infused deep transformer model for multivariate electricity load forecasting
Document Type
Conference Proceeding
Publication Title
Proceedings of the 16th International Conference on Electronics, Computers and Artificial Intelligence, ECAI 2024
Publisher
IEEE
School
School of Science
Funders
National Science Foundation (ECCS-2223628)
Abstract
Electricity load forecasting plays a crucial role in the management of electricity power grids, enhancing operational efficiency, ensuring network reliability, facilitating infrastructure planning, and promoting energy sustainability As the complexity of energy consumption patterns increases traditional forecasting techniques struggle to accommodate the intricate and nonlinear temporal dynamics characteristics present within the data. This paper introduces a novel hybrid model called RLIDT that merges reinforcement learning (RL) with the deep transformer architecture to address the complicated challenges associated with load forecasting effectively. The integration of RL for hyperparameter optimization within the transformer framework not only utilizes their respective advantages but also provides a dynamic, adaptive model that exhibits versatility, robustness, and enhanced predictive accuracy through continuous learning. The experimental investigations conducted on real-world datasets have clearly demonstrated the remarkable advantage of the proposed RLIDT model when compared to traditional methods.
DOI
10.1109/ECAI61503.2024.10607469
Access Rights
subscription content
Comments
Jalali, S. M. J., Fahiman, F., & Khodayar, M. (2024). RLIDT: A novel reinforcement learning-infused deep transformer model for multivariate electricity load forecasting. In 2024 16th International Conference on Electronics, Computers and Artificial Intelligence (ECAI) (pp. 1-5). IEEE. https://doi.org/10.1109/ECAI61503.2024.10607469