Predictive Operations and Maintenance in Photovoltaic Systems: A Systematic Review and Technique-Scale Taxonomy for Energy Management

Rojas J.J.H.; Bonifacio R.F.A.; Chumbimuni M.S.Q.; Travezaño J.M.C.; Carhuamaca S.S.S.; Milla B.Y.B. · Journal of Robotics and Control (JRC) (2025) · DOI: 10.18196/jrc.v6i5.28485

Resumen

The accelerated expansion of photovoltaic (PV) generation increases the demands on operations and maintenance (O&M), where early fault detection and timely decisions are essential to sustain availability and yield. Reported techniques remain fragmented across electrical signal analytics (current-voltage (I-V), power-voltage (P-V), inverter telemetry), thermography and computer vision, reflectometry (spread spectrum time domain reflectometry, SSTDR; phasor measurement units, PMU), Internet of Things (IoT) and supervisory control and data acquisition (SCADA) platforms, and their integration into energy management systems (EMS). The main research contribution is the development of an integrative and verifiable framework that maps diagnostic techniques to application scales and decision goals (detect, prioritize, localize), aligns fault modes with the most suitable signals, and connects diagnostics with state-aware EMS dispatch. Following PRISMA guidelines, a systematic search from 2020 to 2025 identified 195 peer-reviewed articles after screening and removal of duplicates. Data were extracted on the application scale (module, string, inverter, building, microgrid), primary signals, analytical methods, deployment settings, validation procedures, and performance metrics. A PRISMA flow diagram was also added to ensure process transparency. The findings show that non-intrusive electrical diagnostics form the backbone of continuous monitoring, machine-learning models perform best when grounded in physical variables and calibrated to site conditions, thermography and unmanned aerial vehicles (UAVs) accelerate visual triage, reflectometry enhances physical fault localization, and statistical process control detects subtle drifts. Integrating these techniques into EMS enables temporary derating, optimized maintenance scheduling, and co-optimization with storage and electric vehicles. Remaining gaps include the scarcity of multi-site and longitudinal validations, as well as the lack of standardized operational metrics. The proposed framework provides a reproducible pathway to scale predictive O&M in both urban and non-urban contexts.

Instituciones: Universidad Continental