Fault Detection in Solar PV Systems Integrated with the Power Grid: Evaluating Logistic Regression through Confusion Matrix Analysis

Abstract

The fast-paced adoption of solar photovoltaic (PV) technologies has been a double-edged sword for power grids in terms of system reliability, fault identification, and grid stability. The issue of being able to tell the faults in the solar plants correctly and fast is of utmost importance to keep the power supply uninterrupted and operating losses at a minimum. The present review considers using logistic regression as a machine learning method in the detection and classification of faults in grid-tied PV systems. Historical and real-time PV operation data serve as a model input for the logistic regression models to predict faults with high accuracy, still keeping the process computers efficient. The assessment of the models’ performance is done using the confusion matrix, which gives comprehensive views of true positive, true negative, false positive, and false negative predictions. The review through this method emphasises the main metrics like precision, recall, and F1-score, thereby providing a complete evaluation of the model in telling apart normal and faulty system states. Moreover, an investigation is made into how the detection accuracy is affected by different operational parameters such as voltage, current, irradiance, and temperature. The findings suggest that logistic regression, once proper training and validation are done, can become a trustworthy, clear-cut, and economical technique for detecting faults in PV systems, thus being an adjunct to the more sophisticated machine learning methods. Besides, the issues of data imbalance, measurement noise, and real-time implementation are brought up along with the techniques to boost detection performance. This review offers a unified viewpoint on the various fault detection techniques for PV systems and shows the logistic regression plus confusion matrix analysis approach to better grid dependability and operational resilience.