Fault Classification for Single Phase Photovoltaic Systems using Machine Learning Techniques

Photovoltaic monitoring is essential in all kinds of systems for its efficient and optimal working. Even after development of modern protection devices, a lot of faults remain undetected. This paper develops a monitoring and detection algorithm for various faults and operating conditions in single phase PV system. The algorithm is developed using wavelet transform and pattern recognition and classification abilities of machine learning techniques. Discrete wavelet transform is used for extracting features of a given signal at pre-processing stage while multi-layer perceptron neural network is used for fault classification process. The data required for developing the algorithm is obtained by simulating PV system for various operating conditions in MATLAB/Simulink environment. The developed algorithm depicts 99.3% training accuracy and 100% testing accuracy for a randomly divided data set.