Effectiveness of supervised machine learning models for electrical fault detection in solar PV systems

Area of Science:

• Electrical Engineering
• Renewable Energy Systems
• Artificial Intelligence

Background:

• Photovoltaic (PV) systems are crucial for renewable energy but face integration challenges.
• Faults significantly impact PV plant production and operational lifespan.
• Reliable fault detection is essential for grid stability and PV system efficiency.

Purpose of the Study:

• To develop and evaluate Machine Learning (ML) algorithms for detecting and classifying diverse PV system faults.
• To assess the performance of Decision Tree (DT), Naïve Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), and XGBoost algorithms.
• To analyze the impact of faults on PV system parameters using MATLAB/Simulink.

Main Methods:

• Utilized ML algorithms (DT, NB, RF, SVM, XGBoost) for fault classification.
• Conducted simulations in MATLAB/Simulink to model fault scenarios.
• Analyzed voltage, current, and power variations under fault conditions.

Main Results:

• Achieved high classification accuracies: XGBoost (98.0%), NB (97.60%), SVM (97.40%), DT (97.20%), RF (97.20%).
• Validated classification effectiveness using confusion matrices and correlation heatmaps.
• Demonstrated ML's capability in identifying Short Circuits (SC), Open Circuits (OC), Ground Faults (GF), and Mismatch Faults (MF).

Conclusions:

• ML algorithms provide a robust solution for PV electrical fault detection and classification.
• Intelligent monitoring, IoT-based detection, and predictive analytics are vital for enhancing PV system reliability.
• The study underscores the importance of advanced ML techniques for secure and efficient PV energy integration.