An identifiability-guided dual-mode current-conditioned equivalent circuit model for supercapacitors in urban rail transit energy storage systems

Abstract

Accurate supercapacitor equivalent circuit models (ECMs) are essential for voltage reconstruction, control design, and energy-efficiency evaluation in urban rail transit (URT) energy storage systems. Under practical URT profiles, supercapacitors are subjected to high-current bidirectional pulses, where charge–discharge asymmetry and current-conditioned apparent capacitance variation may reduce the accuracy of conventional fixed-parameter models. To address these issues, this paper develops a dual-mode current-conditioned supercapacitor ECM and an identifiability-guided parameterization framework. The charging and discharging dynamics are independently parameterized from their respective multi-current capacitance characteristics and integrated into a unified dynamic model through a hysteresis-based transition mechanism. The apparent capacitance is expressed as a function of voltage, current, and operating mode while retaining a compact single-state circuit structure. For parameterization, resistance and baseline capacitance are first directly extracted, and the mode-specific capacitance-law coefficients are initialized from experimentally extracted capacitance characteristics. Sensitivity and normalized correlation analyses are subsequently used as pre-optimization tools to restrict poorly separable parameter directions, after which only the retained parameter subset is refined through constrained optimization. Repeatability and measurement-noise tests are further conducted to evaluate parameterization stability and prediction robustness. Compared with representative RC, ladder, and branch ECMs under the same multi-current datasets and an independent URT-based dynamic profile, the proposed ECM achieves RMSE, MAE, and maximum absolute error of 13, 11, and 38 mV. It also achieves minimal errors in charge-discharge energy and remaining-capacity estimation, enabling the model for system-level voltage reconstruction and regenerative-energy evaluation.

The authors' abstract, as published at the source. Journal of Energy Storage, 2026 · DOI ↗