Evaluation of Fault Detection Algorithms for Photovoltaic Array Using Distributed Machine Learning Platform

The protection of solar photovoltaic (PV) systems has become immensely important in the last several decades due to the exponential growth in the World's PV power capacity. As a result, fault analysis in PV arrays has evolved into a crucial task for protecting PV modules from damage and minimizing the possibility of occupational hazards. However, PV systems are covered by standard protection mechanisms, although some faults with low mismatch and high fault impedance levels may go undetectable. It might be challenging to discern an anomaly from normal operation due to the often-minute changes in electrical signal magnitude caused by malfunctioning photovoltaic components during such faults. In such cases, datadriven machine learning methods give reliable detection and classification results. This paper presents a workflow to use a machine-learning library of a distributed computing framework, PySpark. PySpark, a Python API for Apache Spark, is a powerful computational engine that efficiently handles enormous data volumes. The key characteristics of I-V curves under various fault occurrences and standard conditions are extracted from a MATLAB simulation of the PV Module for fault detection on the dc side of the PV array. The ML library in PySpark is then used to examine these attributes and detect faults. A confusion matrix addressing soft accuracy, precision, recall, etc., is used in a comparative analysis of several classification methods.