Abstract Fault diagnosis and detection are essential for ensuring the dependability and operational efficiency of solar photovoltaic (PV) systems. This research introduces an innovative machine learning-based fault diagnosis and detection methodology implemented on a 33 kW solar PV system located at P S R Engineering College, Sivakasi. The comprehensive study leverages five years of real-time operational data, encompassing 23,000 data instances that capture seven distinct fault categories: sensor faults, inverter faults, converter faults, grid irregularity faults, PV array mismatch faults, open circuit faults, and maximum power point tracking (MPPT) controller faults. Six machine learning (ML) models were rigorously evaluated to determine the most effective fault detection approach: the random forest classifier, decision trees, support vector machine (SVM), K-nearest neighbors (KNN), logistic regression and the naive Bayes classifier. Performance metrics, including accuracy, precision, and standard deviation (SD), were systematically analysed. The random forest classifier emerged as the best model, achieving an exceptional accuracy of 98.7% with a remarkably low SD score of 0.003; consequently, it was selected as the primary fault prediction algorithm. The implementation utilized the Python programming language and successfully validated the model through comprehensive test case scenarios. Additionally, a sophisticated prototype of a smart PV system was developed, which demonstrated advanced fault detection, data collection, and diagnostic capabilities via NodeMCU ESP32 controllers and an integrated sensor array. The key innovative features of the prototype include real-time monitoring of the PV panel voltage, current, and temperature; automated data logging to Google Sheets via the IFTTT server; an autonomous water spray mechanism for panel dust removal to optimize system efficiency; intelligent camera activation for capturing and documenting abnormal parameter conditions; and comprehensive fault diagnosis and reporting capabilities. The proposed approach significantly enhances the capabilities of PV system operators and maintenance personnel by enabling timely, accurate fault identification and localization, ultimately improving system reliability and operational performance. This technically sophisticated methodology represents a substantial advancement in predictive maintenance strategies for solar photovoltaic infrastructure, offering a robust framework for comprehensive system health monitoring and management.
Pillai et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: