A new approach for determining the common mode (CM) impedance parameters of EMI emissions with the fast backpropagation algorithm of neural networks for low frequency band

Abstract

Predicting and solving EMI emissions related to common mode (CM) ones in pulse with modulation (PWM) inverter induction motor drive systems requires various frequency-dependent effects to be considered. Thus, the application of active circuitry to increase the performance per size and prevent dependency on some parameters of the motor for electromagnetic interference (EMI) filter is considered. The major problems are power line frequency rejection and the compensation of the feedback loop, which is influenced by the wide-ranging utility impedance. While analysis and simulations show, in the literature, that these problems prevent the practical application of active filtering to power supplies especially less than 100 kHz, the approximation with artificial neural networks demonstrate a good promise to take into account the parameters of the motor impedance for independence of the filter easily. Thus, the proposed method allows another way to prevent the switching noise applying to the conducted electromagnetic emissions generally caused by power supply firstly to switching conditions, then secondly to common mode condition of the motor drive system.