Enhancing fault detection reliability in PV systems through a novel approach based on interval-valued kernel PCA

Abstract

• Interval-valued kernel PCA is proposed for robust PV fault detection under uncertainty. • Interval information is preserved to improve robustness to noise and nonlinearity. • Experimental validation is performed on a grid-connected PV system with real faults. • Results show reduced false alarms, missed detections, and detection delays. • The method enables reliable uncertainty-aware monitoring of PV energy systems. This paper proposes an interval-valued kernel principal component analysis framework for fault detection in photovoltaic systems operating under uncertain conditions. The proposed approach preserves the interval-valued structure of measurement data throughout the entire monitoring process, allowing an explicit representation of data variability and uncertainty while enhancing fault sensitivity. Experimental validation is carried out using real measurement data acquired from a grid-connected photovoltaic system subjected to three representative fault types: a one-phase sensor fault exhibiting ramp behavior, nonhomogeneous partial shading corresponding to an intermittent fault, and open-circuit faults in the photovoltaic array characterized by step variations. The proposed method is comparatively evaluated against conventional principal component analysis, kernel principal component analysis, and the vertex-based kernel principal component analysis extension. The results demonstrate that the interval-valued framework consistently reduces false alarm rates, improves fault detection accuracy, and ensures timely fault identification. Quantitatively, the proposed approach achieves improvements of approximately 61% for the Hotelling statistic, 80% for the squared prediction error, and 94.5% for the combined monitoring index across the aggregated loss function, confirming its robustness and effectiveness for reliable fault detection in photovoltaic systems.