Application of fuzzy Q learning (FQL) technique to wind turbine imbalance fault identification using generator current signals

In this paper, Fuzzy Q Learning (FQL) based an adaptive self learning wind turbine fault diagnostic model is proposed using generator current signals. Proposed FQL classifier is capable to achieve very high classification accuracy with small number of samples. This is our first attempt to design such type of classifier using reinforcement learning for fault identification of wind turbine. The beauty of proposed approach is to diagnose the faults without prior knowledge of the system or target value corresponding to input samples. Moreover, proposed approach can also represent the success rate with respect to the number of samples. Raw data of permanent magnet synchronous generator (PMSG) stator current is preprocessed through empirical mode decomposition (EMD) method to generate IMFs. Classifier employs decision tree to further prune these IMFs to most relevant input variables which serve as input to FQL fault classifier. We compare performance of proposed FQL classifier with other AI based classifiers such as ANN and SVM. Imitation results and performance comparison shows that our proposed FQL based classifier could serve as an important tool for wind turbine fault diagnosis.