Permanent magnet generator turn fault detection using Kalman filter technique

In this paper, a stator turn fault detection strategy is developed for a permanent magnet (PM) generator system. Unlike conventional power generation systems, the output of the PM generator is directly rectified by an uncontrolled diode bridge. The only accessible signal is the DC link current/voltage. As a result, most existing detection techniques based on the phase current/voltage signals are not applicable. Instead, the 2nd and 6th harmonics of the DC link current are exploited for turn fault detection and they are extracted by a Kalman filter. It is shown that the phase unbalance caused by a turn fault gives rise to significant increase in the 2nd harmonic DC link current. Consequently, the dominant harmonic under healthy and fault conditions are of the 6th and 2nd orders, respectively. Hence, the turn fault can be detected by comparing the magnitudes of the two harmonics. The detection method is assessed by extensive simulation under various fault scenarios. It is shown that the developed Kalman filter method exhibits significant advantages in response time and computation effort than online fast Fourier transform (FFT) based techniques.