An IoT and Semi-Supervised Learning-Based Sensorless Technique for Panel Level Solar Photovoltaic Array Fault Diagnosis

The performance of solar photovoltaic (SPV) systems is significantly affected by faults in the SPV arrays. Changes in environmental conditions and nonlinear characteristics of solar cells make it challenging to identify faults by conventional protection schemes. Existing learning-based fault diagnosis approaches require voluminous data and sensors, making them costly and challenging to integrate with existing SPV systems. Moreover, the lack of panel-level monitoring in existing approaches renders them incapable of pinpointing the precise location of the faulty panel. To address these issues, this article proposes a methodology for detecting faults, localizing the exact faulty panel, and accurately classifying faults in SPV arrays. The methodology involves a sensorless electronic circuit using a bipolar junction transistor (BJT) and Zener diode for fault detection, an Internet of things (IoT)-based web application for monitoring individual solar panels for fault localization, and a deep autoencoder (DAE)-based semi-supervised learning module followed by a hybrid support vector machine and logistic regression (SVM-LR) for fault classification. The proposed methodology is a low-cost and less data-intensive solution that results in precise fault detection, exact faulty panel localization, and fault classification with an overall accuracy of 99.67%. The methodology is validated in a laboratory-scale real-time setup of a grid-connected SPV system under varying electrical and environmental conditions.