Abstract: Solar energy is most abundant energy source in the universe, PV technology is experiencing a rapid growth over the past few years. But abnormal conditions such as faults, low irradiance etc. lead to the reduction in the available output power from a photovoltaic array. To ensure performance and safety of the PV system, it is necessary to develop techniques that can efficiently localize the faults occurred. This paper presents a detection scheme for faults in the Solar system. If undetected, faults can considerably lower the output of solar systems, damage the panels, and potentially cause fire hazards. The presented fault detection scheme employs Multi-Resolution Signal Decomposition (MSD) technique and two machine learning algorithms namely Fuzzy Logic and K-Nearest Neighbour (KNN) to classify the fault and determine its location. Simulation results verify the accuracy, reliability and scalability of the presented scheme.

Keywords: Fault detection, K-Nearest Neighbour (KNN) algorithm, Maximum Power Point Tracking (MPPT),Machine Learning algorithm, Photovoltaic (PV) array, ground-fault protection devices (GFPDs.

| DOI: 10.17148/IJIREEICE.2021.9523