In power electronic devices, high-speed switching often causes serious electromagnetic interference (EMI) problems. For compliance with electromagnetic compatibility standards, EMI filters are widely used. This article develops an efficient modeling and optimization method of EMI filter by employing artificial neural network (ANN). A partly connected ANN, which is accurate and time saving in training procedure, is proposed to model an EMI filter. Moreover, genetic algorithm and the EMI filter model based on the partly connected ANN are applied to optimizing the EMI filter. Through comparing simulated and measured results of insertion loss of EMI filter, the proposed modeling and optimization method is validated. Compared to equivalent circuit model and electromagnetic model, the proposed method can establish a more accurate EMI filter model and can optimize the design of the EMI filter more efficiently.

In variable speed drive (VSD) system, pulse width modulation (PWM) technology often causes serious common-mode (CM) interference. In order to suppress the CM interference efficiently, it is necessary to analyze and forecast the CM interference theoretically. As an important element of CM interference propagation path, the CM impedance of induction motor needs to be modeled. In this paper, a novel modeling method of the CM impedance from 9 kHz to 30 MHz is proposed. Online measurement is performed to get the spectra of the CM voltage and the CM current. And an improved measurement method of CM voltage is developed. This method uses a high-pass filter to measure the high-frequency components of CM voltage. Based on measured results, the CM impedance in working state is extracted. Then, based on physical structure of induction motor, a high-frequency model is established to fit the extracted CM impedance. Nonlinear programming and Monte Carlo algorithm are used to calculate the model parameters. Through comparing measured and calculated results of the CM current on motor side, the CM impedance model is validated.