The rapid expansion of photovoltaic (PV) deployment has made monitoring, diagnostics, degradation assessment, and predictive maintenance central to long-term solar asset management. This review critically examined conventional and advanced approaches for PV performance monitoring, fault diagnosis, reliability modeling, and maintenance decision support. Supervisory Control and Data Acquisition (SCADA)-based monitoring, string-level supervision, Current-Voltage (I-V) tracing, infrared thermography, electroluminescence, photoluminescence, Unmanned Aerial Vehicle (UAV) inspection, module-level power electronics, AI-based fault classification, and digital twin frameworks were compared based on diagnostic resolution, scalability, cost, implementation complexity, field applicability, and maintenance value. The review shows that conventional monitoring is scalable and inexpensive but insufficient for early module-level fault localization. In contrast, imaging and artificial intelligence (AI)-based methods improve fault resolution but remain constrained by dataset scarcity, limited transferability, inspection-condition sensitivity, and limited real-world validation. Reliability models, including empirical degradation analysis, Weibull modeling, Bayesian updating, survival analysis, and remaining useful life (RUL) estimation, are discussed as tools for maintenance-oriented decision-making. The principal novelty lies in operationalizing PV diagnostics as a staged engineering pathway that links scalable monitoring, field inspection, high-resolution validation, AI-based fault interpretation, reliability modeling, and maintenance prioritization. Particular emphasis is placed on study-quality assessment, domain shift in AI models, false-alarm implications, diagnostic validation, uncertainty-aware reliability modeling, and measurable engineering outcomes such as fault localization, inspection lead time, downtime avoidance, maintenance-cost reduction, and yield retention. The review identifies benchmark datasets, standardized fault taxonomies, multimodal diagnostics, explainable AI, climate-adaptive reliability modeling, and reliability-centered digital twins as essential priorities for future PV asset management.
G et al. (Fri,) studied this question.