PV-FSI: A Physics-Aware Thermographic Severity Framework for UAV-Based Photovoltaic Degradation Assessment

Abstract

Photovoltaic (PV) systems require fault-severity assessment methods that move beyond qualitative thermal inspection toward quantitative, physically meaningful evaluation. This study presents a benchmark-validated thermographic framework, the Photovoltaic Fault Severity Index (PV-FSI), for continuous fault-severity quantification on a normalized 0-100 scale. PV-FSI combines three complementary sub-indices: Hotspot Severity Index (HSI), capturing localized critical heating; Thermal Entropy Index (TEI), representing structural disorder in thermal fields; and Thermal Non-Uniformity Index (TNI), describing spatial temperature variability. Using the DTU Drone Infrared Thermography dataset as a controlled benchmark, PV-FSI is evaluated across multiple fault classes, UAV altitudes, and acquisition conditions. Results show significant fault separability (Kruskal-Wallis, p < 0.001 ), stable severity ordering under altitude variation, and controlled sensitivity to noise and scaling perturbations. A comparison between fixed physics-based weights and PCA-based data-driven weights yields nearly identical rankings, supporting index coherence and robustness. Thermoelectrical analysis further indicates a modest but significant association between PV-FSI and power degradation, consistent with the role of PV-FSI as a thermal severity indicator rather than a direct power-loss estimator. Overall, PV-FSI provides an interpretable, physics-aware degradation descriptor that supports maintenance prioritization and practical monitoring workflows, while motivating longitudinal field validation on diverse PV assets before prognostic deployment.