A Modified Multi-Port Half-Bridge Circuit with Data-Driven Predictive Voltage Control for Battery Balancing and Multi-Level Output

Battery-balancing circuits are essential for improving the performance, safety, and service life of lithium-ion battery packs in electric vehicles and energy storage systems. This paper proposes a modified multi-port half-bridge DC–DC circuit with a reconfigurable port network and its control method for battery balancing and multi-level DC voltage output. The circuit evolves from traditional inductor-based balancing units, while a new sequential turn-off switching strategy is introduced so that only one switch is turned off at any moment, achieving precise voltage distribution by adjusting the duty cycle. To improve control accuracy, a dual closed-loop voltage-current control strategy with adaptive gain scheduling and nonlinear compensation is employed. Furthermore, a predictive voltage control strategy based on Mamba-Multilayer Perceptron optimized by the Crested Porcupine Optimizer (CPO-Mamba-MLP-PVC) is proposed. This data-driven approach predicts a target voltage that considers battery and circuit losses, thereby optimizing the balancing path. Experimental results obtained from a hardware prototype verify both battery equalization and multi-level DC output functions. Compared with conventional methods, the proposed CPO-Mamba-MLP-PVC strategy reduces the balancing time by 18.03% and increases the energy utilization rate to 90.7%.