Photovoltaic energy is the one amid the other non-conventional energy resources extensively used. The convenient conversion process makes it more advantageous. The unusual situation like partial shading and faults are major cause for limiting the power below maximum possible power from the photovoltaic array. In order to operate and maintain a photovoltaic plant safely, effective fault detection and diagnosis are essential. Recent advances in fault diagnosis have been made through the use of machine learning methods. Although, some limitations remain such as feature extraction depends on expert expertise and is not automated. The second problem is that artificial characteristic extraction ignores some features that could be useful. Third, shallow network structures can't be used to learn the nonlinear characteristics of current-voltage curves. These issues are addressed by applying supervised deep learning methods that can automatically extract features, but a lot of categorized data is needed. This article proposed a fault detection scheme based on the voltage and current parameters. The voltage and current ratios are introduced to measure the threshold values according to various faults behaviour. This technique requires less data to detect the fault and also characterise the faults automatically. The wavelet packet transform is used to analyse and measure the energy and standard deviation (STD) of the proposed PV fault parameters. The simulated results have also been analysed using wavelet packet transform (WPT). The performance evaluation and the testing of proposed fault detection scheme is done using 4×4 PV array of 1596 W in MATLAB/Simulink.

___________________________________________________________________________________

Keywords: Fault Detection, Fault Diagnosis, Photovoltaic Array, Partial Shading, Wavelet Packet Transform.