Offline Quantitative Evaluation of Electrical Machine Insulation Through Common-Mode Impedance Resonance Frequency Perturbation

Identifying high-frequency common-mode parameters is crucial for assessing the insulation health of inverter-fed machine systems. However, this task presents significant challenges due to the frequency-dependent effects on winding inductance and resistance. To overcome these difficulties, this article proposes a method for common-mode parameter identification based on resonance frequency perturbation. First, the common-mode resonant frequency is marginally adjusted around the initial value by introducing a series of minor capacitors between the machine terminals (either line-end or neutral-point) and the ground. Subsequently, the frequency-independent capacitances are computed by resolving the resonance frequency minor perturbation equations, utilizing the particle swarm optimization algorithm. Lastly, frequency-dependent inductance and resistance are identified via resonance frequency sweeping. Experimental work was conducted on a 380 V/3 kW induction machine using both star and delta connections. Despite the frequency-dependent inductance and resistance, the parasitic capacitances of the stator winding were accurately extracted. This allows for a quantitative health evaluation of turn and groundwall insulation, with a relative error of approximately 1%.