Enhancing Solar Farm Operations: Machine Learning for Equipment Fault Detection and Classification

Photovoltaic (PV) energy is considered one of the most widespread renewable sources. In this study, the vulnerability of solar PV systems to various faults, leading to potential performance degradation, has been addressed. A robust fault detection and classification strategy is proposed, employing Gradient Boosting Machine Learning (ML) algorithms (Light Gradient Boosting Method (LGBM), Categorical Boosting (Cat-boost), and AdaBoost). The methodology involved formulating a comprehensive PV system to create a synthetic fault database, utilizing diverse features. To simulate normal conditions and various fault scenarios, a MATLAB/Simulink-based PV System is developed. The optimization of hyperparameters of ML algorithms has been achieved by grid search optimization technique, resulting in enhanced performance, and reduced computational cost/time. The study involved multiple independent runs on ML algorithms and the application of Principal Component Analysis (PCA) for dimensionality reduction in the context of fault classification to assess the accuracy and consistency. Cross-validation is implemented to ensure the generalization capability of the ML algorithms to unseen data. Comparison has been established with Random Forest (RF) algorithms to show the performance accuracy of ML in fault diagnosis of PV systems.