Machine learning classifier for fault classification in photovoltaic system

This paper performs fault classification by training a machine learning classifier with several faults and normal operation of photovoltaic (PV) system. The proposed fault classification algorithm is developed by adapting the signal processing properties of wavelet transform for feature extraction, and the learning algorithms of supervised learning models for classifier training. The data required to train the classifier is obtained by simulating a two stage PV system operational under several faults and normal operation. The data tabulated and the features extracted depicted improvement in the classifier training accuracy (99.12%), which is better when compared with literature based conventional training methods.