Year: 2026
Type: article
Abstract: Reliable defect identification in photovoltaic modules is essential for protecting energy yield, ensuring warranty compliance, and enabling scalable inspection of rapidly expanding solar fleets. Existing automated systems perform classification but operate as opaque predictors, offering little diagnostic insight for high-stakes infrastructure. Here, we introduce a reasoning-aware vision-language framework that embeds domain-specific diagnostic logic into learning across heterogeneous electroluminescence, thermal, and red-green-blue (RGB) imagery. By linking visual evidence to plausible defect mechanisms before classification, the framework produces structured diagnostic reports aligned with professional inspection practice. Evaluated on 1,867 real-world photovoltaic images, comprising 1,098 cell-level and 769 module-level images, across eight defect categories, the framework reaches 92.8% accuracy, generates interpretable rationales, and remains robust under realistic image corruptions. A blind concordance study with three independent photovoltaic inspectors shows strong agreement between model diagnoses and expert assessments. Reasoning-aware multimodal learning offers a general paradigm for trustworthy AI-assisted inspection of photovoltaic energy infrastructure.
Source: Cell Reports Physical Science
Language: English
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $5,620
OpenAlex ID: W7148178051
