Real-Time Federated Learning for Smart Energy Management in an Educational Building at RMUTSV
Article Sidebar
Main Article Content
Abstract
Efficient electrical energy management in the Industrial Technician School Building at RMUTSV is crucial due to the complex and continuously fluctuating load characteristics. This research proposes a real-time framework based on Federated Learning (FL) for multi-phase load forecasting (Phases A, B, and C) and adaptive anomaly detection. The system integrates deep learning techniques, including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Autoencoder networks. Experiments using real-time time-series data collected over a four-month period (December 2024 - March 2025) revealed that the FL model achieved a Mean Absolute Error (MAE) of 1.24A, comparable to the centralized model (MAE = 1.15A). Phase B exhibited the highest volatility, with 41 anomalies detected (54.7%). A positive correlation between load and temperature was observed (r = +0.68). The total loss graph decreased within 50 training rounds, indicating strong learning capability. After implementation, the system reduced peak load by an average of 6.3% and significantly decreased abnormal events, enhancing stability and minimizing energy loss. This study demonstrates that FL is an effective, secure, and sustainable approach for deploying Artificial Intelligence (AI) in smart grid systems.
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Copyright belongs to Srinakharinwirot University Engineering Journal
References
C. Briggs, Z. Fan and P. Andras, “Federated learning for short-term residential load forecasting,” arXiv preprint arXiv:2105.13325, May 2022.
A. Taïk and S. Cherkaoui, “Electrical load forecasting using edge computing and federated learning,” arXiv preprint arXiv:2201.11248, Jan. 2022.
J. Sievers and T. Blank, “Secure short-term load forecasting for smart grids with transformer-based federated learning,” arXiv preprint arXiv:2310.17477, Oct. 2023.
S. Bose and K. Kim, “Federated short-term load forecasting with personalization layers for heterogeneous clients (PL-FL),” arXiv preprint arXiv:2309.13194, Sep. 2023.
R. Rahman, N. Kumar and D. C. Nguyen, “Electrical load forecasting in smart grid: A personalized federated learning approach,” in Proc. IEEE Consumer Communications & Networking Conf. (CCNC), Las Vegas, NV, USA, Jan. 2025, pp. 1–6.
R. Rahman, P. Moriano, S. U. Khan and D. C. Nguyen, “Electrical load forecasting over multihop smart metering networks with federated learning,” arXiv preprint arXiv:2502.17226, Feb. 2025.
Z. Zhang, S. Rath, J. Xu and T. Xiao, “Federated learning for smart grid: A survey on applications and potential vulnerabilities,” arXiv preprint arXiv:2409.10764, May 2025.
A. Wallis, S. Hauke, H. Jörg and K. Ziegler, “Federated learning for distributed load forecasting: Addressing data imbalance in smart grids,” in Proc. ENERGY 2025 Conf., 2025, pp. 13–18.
M. Zhang, Y. Li, H. Wang, X. Chen and J. Wu, “Enhancing smart grid load forecasting: An attention-based deep learning approach,” Energy Reports, vol. 12, pp. 567–578, Mar. 2024.
Y. Liu, Z. Dong, B. Liu, Y. Xu and Z. Ding, “FedForecast: A federated learning framework for short-term individual load forecasting,” International Journal of Electrical Power & Energy Systems, vol. 148, p. 108964, Oct. 2023.
American National Standards Institute, ANSI C12.20-2010: American National Standard for Electricity Meters—0.1, 0.2, and 0.5 Accuracy Classes. Washington, DC, USA: ANSI, 2010.
Z. Tang, Y. Yan, D. Wu, T. Yang, R. Dong, S. Hao, W. Wang, Y. Chen and Y. Tian, “A survey of integrating federated learning with smart grids: Application prospect, privacy preserving and challenges analysis,” in Proc. Int. Conf. Big Data and Security (ICBDS), Singapore, Dec. 2022, pp. 227–240.
H. Chen, Y. Zhang, Z. Wang, J. Zhao, and X. Guan, “Federated online learning for adaptive load forecasting across dynamic smart grid environments,” International Journal of Electrical Power & Energy Systems, vol. 170, p. 109482, Feb. 2025.
