Target-Mean State-of-Charge Control for Maximum Utilization of Heterogeneous Reconfigurable Battery Systems Under Constant-Bus Constraints

Cell degradation in second-life battery packs introduces heterogeneous capacity and internal resistance mismatch, reducing the effectiveness of conventional balancing approaches and limiting available pack runtime. Although equal state of charge (SoC) does not necessarily imply equal usable capacity, SoC-based control remains attractive for runtime-oriented operation. This paper proposes a target-mean controller for heterogeneous reconfigurable battery packs under constant-bus constraints that aims to improve runtime and achieve the cutoff-defined theoretical maximum capacity utilization limit. Using only real-time cell SoC measurements and legal switching actions, the controller selects the configuration that best reduces deviation from the pack-average SoC while preferentially loading cells above the mean. The online action selection requires no active balancing hardware, no explicit capacity or state of health (SoH) estimation, and no offline optimization; experimentally measured capacities are used only for calibrated Coulomb-counting SoC estimation. Simulation results on a heterogeneous five-cell reconfigurable battery pack show that the proposed controller reaches the cutoff-defined 90% theoretical utilization limit in the full-initial-SoC cases, while also extending runtime and reducing switching activity by up to 11.75% relative to the comparison methods. Hardware validation on a five-cell prototype further confirms this trend, achieving 89.12% experimental utilization, zero final SoC spread, and higher delivered energy than both comparison methods. A stepped-load hardware test further achieved 88.19% utilization from current integration, corresponding to 97.99% of the cutoff-defined 90% theoretical limit. The results suggest that, for heterogeneous second-life packs, SoC-based reconfiguration control can achieve both runtime improvement and near-maximum utilization without the added complexity of explicit SoH-aware balancing.