Optimized power management of multi-parallel energy storage systems via virtual impedance: Accurate power sharing and circulating-current suppression

This is the authors' abstract. We don't add key points for this paper.

Accurate power sharing and circulating-current suppression are essential for reliable operation of multi-parallel power conversion systems in large-scale energy storage applications. Conventional droop control is sensitive to feeder-impedance mismatch, causing unequal voltage drops, aggravating active–reactive coupling, and driving converter outputs away from rated-capacity ratios. This paper proposes a generalized power-management strategy based on virtual-impedance shaping. Feeder resistance is obtained from known cable parameters or calibrated from fundamental voltage/current measurements during commissioning, and a virtual-resistance component compensates the feeder resistance while retaining a prescribed residual resistance. This avoids continuous online tracking of the full feeder impedance and keeps the residual resistance positive under bounded estimation errors. Coordinated relationships among droop coefficients, virtual inertia, damping coefficients, equivalent impedance, and rated PCS capacities are derived to improve proportional power sharing and transient synchronization. A reduced-order small-signal model evaluates eigenvalue distribution and damping under different residual-resistance conditions and bounded feeder-resistance estimation errors. Dual-PCS systems with 1:1 and 1:2 capacity ratios are studied in Matlab/Simulink and on an RT-LAB platform. Time-domain simulations use the ideal complete-compensation condition to isolate the power-decoupling and circulating-current-suppression mechanisms, with comparisons against conventional droop and adaptive virtual complex-impedance control under identical severe feeder-mismatch conditions, plus a sensitivity study over ±5% feeder-resistance estimation error. Under the nominal comparison condition, the proposed method reduces steady-state power-sharing errors to 0.12% (active) and 7.43% (reactive), and lowers RMS circulating current to 0.11 A. Results demonstrate improved power-sharing accuracy, circulating-current suppression, and dynamic performance, while revealing the method's sensitivity to feeder-resistance estimation accuracy.