[Paper Review] Detection and classification of faults aimed at preventive maintenance of PV systems

Edgar Hernando Sepúlveda-Oviedo, Louise Travé-Massuyès+3 more|arXiv (Cornell University)|Jun 13, 2023

Renewable energy and sustainable power systems4 citations

TL;DR

This paper proposes a novel Random Forest-based approach for detecting and classifying fine faults—particularly snail trail defects—in photovoltaic (PV) systems, using an advanced feature extraction and selection pipeline. The method achieves high classification accuracy with reduced computational time, enabling effective preventive maintenance in PV power plants.

ABSTRACT

Diagnosis in PV systems aims to detect, locate and identify faults. Diagnosing these faults is vital to guarantee energy production and extend the useful life of PV power plants. In the literature, multiple machine learning approaches have been proposed for this purpose. However, few of these works have paid special attention to the detection of fine faults and the specialized process of extraction and selection of features for their classification. A fine fault is one whose characteristic signature is difficult to distinguish to that of a healthy panel. As a contribution to the detection of fine faults (especially of the snail trail type), this article proposes an innovative approach based on the Random Forest (RF) algorithm. This approach uses a complex feature extraction and selection method that improves the computational time of fault classification while maintaining high accuracy.

Motivation & Objective

• To address the challenge of detecting fine faults in PV systems, especially those with subtle signatures like snail trails.
• To improve fault classification accuracy and computational efficiency in PV system diagnostics.
• To develop a specialized feature extraction and selection process tailored for fine fault detection.
• To support preventive maintenance by enabling early identification of incipient faults in PV power plants.

Proposed method

• The study employs the Random Forest (RF) algorithm as the core classifier for fault detection and classification.
• A complex feature extraction pipeline is designed to capture subtle electrical and thermal signatures from PV panel data.
• A feature selection technique is applied to reduce dimensionality and improve computational efficiency without sacrificing accuracy.
• The method processes time-series or real-time monitoring data from PV systems to identify fault patterns.
• The approach is validated on a dataset from a real PV power plant, focusing on snail trail and other fine fault types.
• Hyperparameter tuning and cross-validation are used to optimize the RF model's performance.

Experimental results

Research questions

• RQ1 How can fine faults in PV systems—particularly snail trail defects—be effectively detected using machine learning?
• RQ2 What feature extraction and selection methods significantly improve fault classification accuracy and speed?
• RQ3 To what extent does the proposed Random Forest-based approach outperform existing methods in detecting subtle PV faults?
• RQ4 Can the proposed method support practical preventive maintenance in large-scale PV power plants?

Key findings

• The proposed method achieves high fault classification accuracy, particularly for fine faults such as snail trails, which are difficult to detect with conventional methods.
• The feature selection process reduces computational time significantly while maintaining high detection performance.
• The Random Forest model demonstrates robustness and generalization capability across diverse fault types in PV systems.
• The approach enables early detection of incipient faults, supporting effective preventive maintenance strategies.
• The study shows that specialized feature engineering is critical for distinguishing subtle fault signatures from normal operational variations.
• The method is scalable and suitable for integration into real-time monitoring systems for utility-scale PV plants.