FRA‐Based Parameters Estimation for Locomotive DC‐Link Capacitors

Accurate capacitance estimation supports condition monitoring of locomotive DC‐link capacitors. This paper develops a voltage‐based estimation framework by integrating the fast recursive algorithm (FRA) with a linear‐in‐the‐parameters precharging model. Capacitor current is reconstructed from source‐side and capacitor‐side voltage measurements and known circuit resistances, allowing capacitance to be identified without an additional current sensor or separate signal injection. FRA is applied to estimate the regression coefficients from a selected precharging data window. Laboratory experiments using actual locomotive DC‐link capacitor components with reference capacitances of 6216, 3108, and 1554  μ F evaluate the effects of sample count and sampling frequency. The reported estimation errors remain below 1% across the tested frequency range of 200 Hz–2 kHz. Comparisons with an implemented recursive least‐squares baseline show more consistent estimation performance across the investigated acquisition settings. The framework enables quasi‐online capacitance assessment during precharging events, providing a basis for periodic condition monitoring. Its feasibility is demonstrated under laboratory conditions; field validation remains necessary.