Battery AC Impedance Measurement Using Combined Median and Gaussian Filtering

Abstract

Battery state estimation is essential for ensuring efficiency and safety. Among various parameters used for battery state estimation, AC impedance is widely regarded as a highly reliable diagnostic indicator, as it can be measured directly from the power conversion system during battery charging and discharging. By controlling the switching operation, a pseudo-random sequence (PRS) signal with a wide frequency spectrum is injected into the battery. Frequency Domain Analysis is then applied to analyze the frequency components and calculate the battery AC impedance. However, the PRS signal inherently exhibits spectral attenuation at high frequencies, which reduces the signal-to-noise ratio. Additionally, signal leakage introduces impulsive noise and impedance-trajectory distortion, making post-processing essential for reliable impedance measurement Previous studies often involve high implementation complexity and additional post-processing burden to enhance noise-filtering performance, which can lead to end effect and response delays. To address these issues, this study proposes a post-processing method that combines a median filter for effective impulsive noise removal and a Gaussian filter for signal smoothing. The proposed method is implemented on a passive cell balancing circuit. Its performance is verified through comparison with results obtained using a single Gaussian filter and a median-moving average filter.