An improved physics-free genetic impedance fitting technique for precise high-frequency modeling of two-terminal electrical systems

There is an increasing industrial demand of using impedance fitting techniques to enable an equivalent circuit to approximate the electrical behavior of an electrical system. However, due to the high dependence on the prior physical information of the electrical system, previous impedance fitting techniques have long suffered from unacceptable implementation complexity and unsatisfactory fitting precision. To realize easy, cheap, and precise impedance fitting, a Physics-free genetic impedance fitting technique (PFGIFT) is proposed in this paper. Firstly, PFGIFT utilizes a five-component RLC-parallel (FCRLCP) equivalent circuit, which gets rid of expensive physics-based equivalent circuit modeling, to approximate the impedance of a two-terminal electrical system. Then, a novel bounded gaussian mixture model based genetic algorithm (BGMMBGA) is proposed to fully leverage the fitting capability of the FCRLCP equivalent circuit by embedding parameter tuning expertise into the evolutionary process. Under the combined effects of FCRLCP equivalent circuit and BGMMBGA, PFGIFT achieves precise impedance fitting performance without the use of expensive knowledge about physical information. The effectiveness of PFGIFT is verified separately on a human-designed impedance fitting task and an industrial application of fitting the common-mode impedance of a motor drive system. © 2023 Elsevier GmbH.