Sensorless control of DFIGs based on active and reactive power computation for encoder fault tolerance

A new detection and isolation technique for encoder faults is presented for doubly-fed induction generators. This method compares the measured rotor position with an estimation from a new MRAS (model-reference adaptive system) observer based on the computation of both active and reactive power transferred via the air gap. The difference between measured and estimated position is monitored by a CUSUM (Cumulative Sum) algorithm, while a second observer, based on an internal state-space model, detects anomalous changes in the measured position. A position fault is flagged when both algorithms react. The combination of sensorless algorithm and fault detection technique is tested experimentally on a 3-kW 4-pole doubly-fed induction generator (DFIG) for different working points. The starting performance of the MRAS observer and its behaviour in case of transients is tested during the synchronization process of the DFIG.