Solar Panel Inspection
Inspect photovoltaic cells for defects
Graphical programming environment for deep learning-based industrial image analysis
Solar panels typically operate in the field for 25 to 30 years. Small defects in the solar photovoltaic (PV) cells comprising each panel decreases the efficiency with which they convert sunlight into usable electricity or lead to premature failure. These defects can impose a significant cost in lost power generation over the panel’s operational lifespan, making it imperative to reject cells with even small defects before final assembly.
Each PV cell has several layers, including front and back metallic electrodes, a silicon layer, and a textured surface with an anti-reflective coating. PV cells can vary in visual texture and shade without these variations having any effect on performance. Scratches, cracks, bubbles, inclusions, and contact forming errors all affect final efficiency. These are detected through various combinations of electroluminescence (EL) imaging, photoluminescence (PL) imaging, and visible light imaging.
Time-consuming manual inspection processes can be a production bottleneck. Conventional machine vision has trouble disregarding all acceptable color and texture appearance variations, while the wide range of defect types, sizes, and possible locations makes it difficult to program a set of rules for finding them.
Cognex Deep Learning is an ideal technology for solving solar cell inspection. It trains on a set of images showing the full range of acceptable PV cells, and a set of images showing the full range of possible errors. The defect detection tool learns to ignore all background texture and color variations, and identifies even tiny defects, no matter what they look like or where they appear in the cell. It is both more accurate and significantly faster than manual inspection.