Improved fault detection and classification in PV arrays using stockwell transform and data mining techniques

Abstract

The growing integration of photovoltaic (PV) systems into the power grid necessitates reliable fault detection and classification mechanisms to ensure operational efficiency and safety. Fault detection in photovoltaic (PV) arrays is crucial for maintaining optimal system performance and ensuring the reliability of solar power generation. This paper proposes a novel approach for fault detection in PV arrays by employing the Stockwell transform in combination with various data mining techniques. The Stockwell transform is an advanced time-frequency analysis tool that allows for enhanced feature extraction from time-series data. By applying the Stockwell transform to the PV array's operational data, valuable frequency-domain information is extracted, enabling the identification of subtle fault signatures. To effectively detect and classify various faults, different data mining techniques, such as support vector machines, decision trees, random forests, and k-nearest neighbours, are applied to the transformed data. Each technique's effectiveness in identifying faults is evaluated and compared, enabling the selection of the most suitable algorithm for the specific application. Experimental results demonstrate the effectiveness of the proposed fault detection approach, exhibiting high accuracy, sensitivity, and specificity in identifying various types of faults in PV arrays. Extensive simulations and experimental validations were conducted on various fault conditions, including partial shading, open-circuit faults, and degradation. The results demonstrate the proposed method's superior performance, achieving an accuracy of 99.61 %, precision of 99.75 % and F1 score of 98.73 %. These metrics significantly surpass traditional fault detection techniques, highlighting the method's potential for real-world deployment. The approach not only enhances the reliability of PV systems but also contributes to reducing maintenance costs and improving system efficiency. The combination of the Stockwell transforms with data mining techniques proposed here provides a robust and efficient framework for early detection of faults, enabling timely maintenance and minimizing energy losses.