Distributed Constrained Real-Time Control of Multiple Batteries in Green Networks
Eleni Stai, Giorgos Gotzias, Marianna Papadostefanaki, Symeon Papavassiliou · IEEE Networking Letters · 2025
Energy sustainability is endeavored in all networked systems and renewable energy sources (RESs) will play a fundamental role towards its achievement. Due to their uncertainties, RESs integration for powering any type of network requires support for handling energy surplus or shortage due to e.g., unexpected weather conditions or generation-demand lags. Such support can be offered by batteries that have become crucial for a large scale integration of RESs to networked systems and when multiple batteries serve one network, their coordinated control will allow for maximizing the benefits of their operation. In this paper, we propose a real-time control scheme for multiple batteries based on a multi-agent reinforcement learning scheme that is distributed in execution but centralized in training so as to achieve a high level of coordination but in a scalable way. Most importantly, it is safe in the sense that it accounts for binding lookahead constraints on the batteries state-of-energy, which express long-term feasibility and cost efficiency goals in the network operation. We have performed a case study in smart grids with the network goal of following a dispatch plan in real-time. Numerical results showcase the effectiveness of the proposed scheme both in terms of cost performance and of constraints violation avoidance versus baseline approaches.