Aluminum electrolytic capacitors are widely used in power electronic converters. However, their high failure rate presents significant challenges to system reliability. This paper aims to propose a method utilizes a quasi-online identification algorithm to measure the instantaneous capacitance (C) and equivalent series resistance (ESR) in real-time. By continuously tracking the variation trends of these parameters under actual operating conditions, the method effectively monitors the capacitor’s aging process over its lifetime to ensure system reliability.
The equivalent-circuit model of the capacitor at low-to-medium frequencies is first simplified and a discretized matrix equation is then derived to formulate the recursive least squares (RLS) algorithm. An adaptive forgetting factor is introduced to improve both tracking capability and steady-state performance. To suppress high frequency noise and extract additional system information, the measured signals are pre-processed by an empirical mode decomposition (EMD) filter before being fed into the RLS estimator. The proposed identification scheme is finally implemented on Buck and Boost converters to demonstrate its validity and accuracy.
Using the proposed identification method to estimate the C parameters of the Buck and Boost converter prototypes, the obtained errors for both C and ESR are less than 3.5%, demonstrating high accuracy. The accuracy of this method is higher than that of the stochastic Newton method and Kalman filter approaches employed in previous literature.
This paper proposes a quasi-online method that integrates EMD with adaptive forgetting factor recursive least squares (AFFRLS) to estimate the C and ESR of capacitors in direct current to direct current converters.
