Aerial Infrared Health Monitoring of Solar Photovoltaic Farms at Scale

TL;DR

This work tackles scalable health monitoring for large-scale solar photovoltaic farms by building a georeferenced, multi-sensor infrared dataset across North America and an end-to-end ML pipeline for real-time anomaly detection and loss estimation. It integrates table and panel localization, hotspot and safety hazard detection, and post-processing to produce actionable GeoJSON outputs and a standardized PV health rating system. The study demonstrates high segmentation and detection performance, and provides insights on energy/revenue losses and asset health across diverse geographies, mounting configurations, and ages. The proposed framework enables predictive maintenance and scalable analytics, offering significant potential to improve reliability and economic performance of renewable energy infrastructure.

Abstract

Solar photovoltaic (PV) farms represent a major source of global renewable energy generation, yet their true operational efficiency often remains unknown at scale. In this paper, we present a comprehensive, data-driven framework for large-scale airborne infrared inspection of North American solar installations. Leveraging high-resolution thermal imagery, we construct and curate a geographically diverse dataset encompassing thousands of PV sites, enabling machine learning-based detection and localization of defects that are not detectable in the visible spectrum. Our pipeline integrates advanced image processing, georeferencing, and airborne thermal infrared anomaly detection to provide rigorous estimates of performance losses. We highlight practical considerations in aerial data collection, annotation methodologies, and model deployment across a wide range of environmental and operational conditions. Our work delivers new insights into the reliability of large-scale solar assets and serves as a foundation for ongoing research on performance trends, predictive maintenance, and scalable analytics in the renewable energy sector.

Figures (9)

• Figure 1: Samples of solar sites varied by capacity in MW and panel mounting type. Top row, left-to-right: 1, 25, 100, and 400 total capacity in MW. Bottom row, left-to-right: Canopy, Rooftop, Ground, and Mixed. Infrared orthomosaics are visualized in grayscale and overlaid onto RGB orthomosaics. Brighter pixels represent higher temperatures.
• Figure 2: Geographic locations of airborne infrared inspections in our dataset. Our dataset consists of high-resolution infrared orthomosaics of a diverse set of 6,155 solar farms across North America captured by fixed-wing aircraft.
• Figure 3: Distribution plots of dataset statistics of (a) total site capacity in MWDC, (b) solar plant site age in years, (c) solar panel module types, and (d) solar panel mounting types.
• Figure 4: Solar Panel Defect Categories and their Descriptions. For Anomaly Detection in infrared imagery, we detect 6 types of defects, including hotspots, multi-hotspots, diode bypass, single panel outage, string outage, and misaligned panels due to faulty trackers.
• Figure 5: Architecture of our Infrared Solar Farm Inspection Pipeline.