Performance Testing of a Hybrid Imitation Learning and Rule-Based Approach for Real-Time Microgrid Energy Management
- Authors
-
-
Taheni Swibki
National Engineering School of Carthage
-
Dhaker Abbes
-
Kais Ouni
-
Lilia El Amraoui
-
- Abstract
- Abstract: This paper investigates a hybrid approach that combines Imitation Learning (IL) and Rule Based methods for real-time energy management in a real-scale grid-connected microgrid (MG). This approach eliminates the need for forecasting models, which can be complex and unreliable. The primary objective of this approach is to minimize energy costs by reducing dependence on the external grid through optimal control of the storage system and photovoltaic (PV) power generation of the Lille Catholic university demonstrator. The proposed method integrates a Deep Learning (DL) model trained to mimic the behavior of a Linear Programming (LP)-based energy management strategy. The optimal control outputs of LP, derived from real historical data, serves as ground truth expert demonstrations for the IL process, which enable the DL model to learn how to replicate optimal energy management decisions. Since the DL model cannot inherently enforce operational constraints (e.g., storage limits, energy balance, etc), a Rule-Based post-processing module is used to ensure that the microgrid operates within its operational limits. Numerical experiments show the effectiveness of the proposed method in enhancing responsiveness and efficiency in real-time energy management.
- References
- Cover Image
-
- Downloads
- Published
- 2026-09-29
- Issue
- Vol. 28 No. 3 (2026)
- Section
- Articles
- License
-
Copyright (c) 2026 Journal of Control Engineering and Applied Informatics
This work is licensed under a Creative Commons Attribution 4.0 International License.
