A unified and adaptive real-time graph-theoretic optimization strategy for reconfigurable battery storage systems

Reconfigurable battery storage (RBS) systems enable dynamic cell-level connections to enhance flexibility and fault tolerance. However, developing scalable and adaptive real-time control strategies for arbitrary RBS topologies remains a significant challenge, as most existing methods are customized to specific configurations and lack a generalizable solution. To bridge this research gap, this paper proposes a unified, graph-theoretic strategy for modeling and optimizing RBS reconfigurations. Firstly, the proposed strategy leverages undirected graphs to effectively model the RBS system by capturing the bidirectional behavior of the switches, providing a unified modeling approach applicable across diverse RBS topologies. Meanwhile, a sneak circuit identification method is proposed to enhance the safety and reliability of the model. Secondly, based on the graph model, the Path-SMART algorithm (Path-searching and Steiner-tree-based Method for Arbitrary Reconfigurable Topologies) is proposed. In this algorithm, series connection optimization is achieved using Dijkstra’s shortest path algorithm, while parallel search is modeled as a Steiner-tree problem solved using Prim’s algorithm. The algorithm effectively reduces the number of switches required for the desired connections while maintaining a lower computational complexity compared to existing search methods. Third, Path-SMART is implemented in real time to enable adaptive reconfiguration that maintains cell balancing during both charging and discharging. It also incorporates a graph update mechanism to preserve system functionality in the presence of switch open-circuit faults and cell failures. Validated through simulations and hardware experiments, the proposed framework demonstrates strong applicability, robust operational safety, and effective cell balancing.