Feature-Engineering Enabled Multiobjective Evolutionary Impedance Fitting Technique

The impedance fitting technique aims to make the impedance of an equivalent circuit the same as the measured impedance of the real system through parameter tuning. Since only the magnitude and phase of the impedance are measurable, the single-objective optimization algorithm, where the objective is to minimize the weighted summation of the fitting errors on magnitude and phase, has been widely used to achieve automatic impedance fitting. However, this scheme is easy to be trapped in the local minima, which, leads to poor fitting precision. A feature-engineering enabled multi-objective evolutionary impedance fitting (FEMEIF) technique is proposed in this paper. By implementing several effective feature engineering techniques to fully leverage the limited impedance information, and by utilizing these synergetic features to enable multi-objective impedance fitting for realizing better hierarchy on individuals, FEMEIF much improves the impedance fitting performance with higher fitting precision and lower optimization variance compared with the traditional single-objective impedance fitting schemes. FEMEIF provides a novel methodology for improving the impedance fitting task, requires no additional measurements or expenses, and is universal for various optimization algorithms or equivalent circuits. Besides, it provides an open topic for exploring other potential features to achieve further improvement. The effectiveness of the FEMEIF is verified separately on fitting the impedance of a generic RLC-parallel circuit and on an industrial application of fitting the common-mode impedance of a motor system. © 2023 IEEE