Fault Diagnosis on a Wound Rotor Induction Generator Using Probabilistic Intelligence

Wound rotor induction generators are commonly used for wind applications. Although this technology is mature and in widespread use, there has been relatively little research on online condition monitoring thereof towards improving overall reliability of the system in which it is applied. This paper presents a method for diagnosing incipient faults on a wound rotor induction generator. The proposed method uses a probabilistic intelligence technique Bayesian classification together with voltage signature analysis for the fault diagnosis which has yet to be presented for wound rotor induction generators. A model of a three-phase wound rotor induction generator is constructed using finite element modelling. The behaviour of the generator is investigated under healthy, stator fault and rotor fault conditions. The proposed method is then implemented and tested for the task of diagnosing these faults. Results indicate that the Nave Bayes classifier was successfully trained and yielded 94% test accuracy which indicates the potential suitability of the method in enhancing predictive maintenance for wound rotor induction generators.