Electroluminescence (EL) inspection is widely used to identify microcracks, inactive areas and other defects in photovoltaic modules. While the technique is established in manufacturing, applying it efficiently to installed modules presents a different challenge.
MBJ Solutions argues that artificial intelligence (AI) can help make field EL inspection faster and more consistent, particularly when hundreds or thousands of modules need to be assessed.
From laboratory analysis to field inspection
Field inspections have to contend with restricted access, changing conditions and limited maintenance windows. The challenge is therefore not simply capturing useful images, but converting them into practical maintenance decisions.
MBJ Solutions uses three systems for different levels of inspection:
AI can assist by identifying suspicious areas, applying consistent defect criteria and reducing the amount of repetitive manual image review required.
Automatic defect detection can highlight anomalies across an inspected module.
Inspecting modules while they remain in operation
Quickcheck is designed to inspect installed modules without electrically disconnecting each module from the PV system.
The portable measurement head combines a camera, battery, LED illumination and onboard computer. Operators view live images through a smartphone mounted on a telescopic guide bar, which can also help reach modules further away.
Instead of relying on one image, the module can be scanned section by section. Individual captures are then stitched together to create a coherent representation of the inspected area.
This gives AI analysis greater context, allowing suspicious regions to be located across the module rather than assessed only as isolated images.
How the measurement works
Quickcheck uses controlled local illumination to generate luminescence information in field conditions.
A reference image is first recorded with the LED switched off. An integrated 850 nm near-infrared source then excites the cells, with the resulting luminescence captured by an InGaAs camera using a 1150 ± 25 nm bandpass filter.
Alternating LED-on and LED-off images are compared to produce an image where electrically active areas appear bright and damaged or inactive regions appear dark.
MBJ Solutions says this allows EL-like diagnostic information to be collected during daylight-oriented field inspections.
AI supports, rather than replaces, expertise
The value of AI becomes clearer as inspection volumes increase. A trained model can repeatedly apply the same criteria when screening large numbers of images.
Potential benefits include:
MBJ Solutions stresses that final technical judgement remains with experienced engineers and service specialists, particularly in borderline cases.
Field data also differs from factory data. Ageing, soiling, installation stresses, wind and snow loads and long-term operation create defect patterns that may not appear during production. AI models therefore need to be trained and refined using field-relevant information.
Stitched module imaging can show automatically detected defects across the complete inspected area.
As the installed PV base ages, inspection demand is likely to increase. AI-supported EL inspection could help service teams extend diagnostic coverage without a corresponding increase in manual image assessment, allowing specialist expertise to be concentrated where further investigation is required.
Read the complete insight from MBJ Solutions on field electroluminescence inspection, AI-assisted defect detection and PV diagnostics in PES Solar: https://pes.eu.com/exclusive-articles/supporting-electroluminescence-inspection-in-the-field-with-ai