Electroluminescence technology allows the inspection of photovoltaic panel components that are not visible to the naked eye or detectable in other forms of solar panel inspections, write Dr Michelle McCann and Lawrence McIntosh of PV Lab Australia, with Liam O’Duibhir, Carsten Eckelmann, Aaron Tranter and Yan Wang
Electroluminescence testing is one of the techniques at the heart of our solar panel testing business at PV Lab Australia. This test allows us to see deep inside a solar panel to see things that we cannot see with the naked eye and that cannot be seen with other tests.
Electroluminescence (EL) is used by system owners and insurers to search for damage after an extreme weather event; system owners or builders to take care of their investment by regularly checking for unexpected panel degradation in large solar parks; by builders or installers to check for damage during panel installation; and by buyers to check for damage during shipping.
Electroluminescence is performed in two parts. The first task is to capture the image of a solar panel. The second part is to analyze this image.
Image acquisition is essentially done by operating the solar panel in reverse: applying a current and letting the light out, similar to an LED since a solar cell is also like a diode. It is quite easy to do this in the dark, at night for a large solar park or in a dark room in a laboratory, but there are also methods for daytime electroluminescence.
In the old days …
For a large solar park, the most rudimentary method of image capture is driving at night with a tripod-mounted camera in the rear of a vehicle with a power source. The power source can be used to electrify entire strings or individual panels (as needed) and the camera captures an image. It takes a few minutes for each panel.
Image analysis is performed after image capture and is a desk-based job that takes a few minutes for each panel. A person is required to zoom in on the image, focusing on only one cell at a time to assess whether it has any noteworthy features. If it does, these are noted in a spreadsheet. This process is then repeated for the 60 to several hundred cells in the panel.
Inspections with drones
To accelerate image acquisition at PV Lab Australia, we partnered with Quantified Energy Labs (QE Labs) to bring drone-based electroluminescent imaging to Australia for the first time.
QE Labs is based in Singapore and is a spin-off of the National University of Singapore. The company’s autonomous drone mapping technology is up to 20x faster and 10x cheaper than tripod-based electroluminescence. In a recent project that also appears to be one of the largest EL inspections in the world, they imaged 120,000 modules in a floating photovoltaic plant in just two weeks.
The system provided by QE Labs is very practical. Using the sitemap as input, the flight path for the drone is planned independently and the drone flies on autopilot. The images captured by the drone can be georeferenced to allow the reliable identification of any defective modules.
However, speed is rarely free, and there’s a slight drop in image quality compared to tripod-based inspections. Depending on the detected panel failure, this may not be relevant, but in cases where it is relevant, the ability to scan an entire solar park and then narrow down the panels or regions of interest more than compensates adequately in terms of time and money. spent.
There’s no point in capturing images of 120,000 forms in two weeks if you need minutes to analyze them. Even with just one minute of analysis per panel, 120,000 minutes is almost three months of analysis time.
Enter machine learning (ML). Machine learning is particularly suited to the image analysis task as solar panels contain distinct and homogeneous cells, enabling automatic, fast and accurate detection of defects.
Image recognition through machine learning has demonstrable success in many other areas such as medicine, advanced manufacturing, and satellite imagery. Additionally, these algorithms can continuously improve and monitor their accuracy, dramatically reducing the need for human intervention.
Machine learning algorithms can also scale based on demand, exponentially increasing the throughput of the current process. Accurate and automated defect detection would significantly reduce analysis time and costs for the customer.
For a machine learning approach to be successful, three things are critical:
- A data set large enough to accurately train the system.
- Smart algorithms.
- Models built and trained to run in real time in what is effectively a production environment.
To this end, PV Lab Australia partnered with two local companies, Aqacia and 2pi Software. This work is supported by Innovation Connect in Canberra.
Aqacia provides machine learning and artificial intelligence (AI) solutions for high-tech industries, addressing the toughest real-world problems and building solutions to enable seamless integration of AI into manufacturing and research settings.
2pi Software is an AI / ML enablement company that provides a digital platform that reduces the heavy load and software complexity involved in entering and exiting large volumes of data to and from neural networks. This allows data scientists to focus on the main domain problem being addressed.
For the project, PV Lab Australia brings a dataset of many tens of thousands of panels and quadrillion of thousands of cells that have been individually analyzed. Aqacia is developing automatic image processing and ML inference models to automate defect identification and 2pi Software will create a solution that leverages the high-performance processing and parallelization capabilities of the Amazon Web Services (AWS) cloud to enable streams of flexible data processing work to be built to optimally train AI / ML models.
This partnership is ultimately to support the work with drones. Depending on the size of the dataset, human analysis can take weeks, while AI technology can process hundreds of panels in minutes.
The increased speed means that the panels can be evaluated in real time both in the PV Lab production environment and in a drone-aboard environment. Real-time assessment allows for quick feedback to system owners and iterative control of the inspection process if more detailed assessment is required in certain situations.
“The 2pi Software team is very honored to be involved in such a groundbreaking innovation in the PV industry and applying emerging AI / ML techniques to help monitor the continuous health and life efficiency of solar panels. – a very fundamental aspect of the PV promise, ”says 2pi software director Liam O’Duibhir.
Bringing everything together
This work is almost (but not quite) too good to be true. The final product achieves the triplet of being:
- Faster, thanks to drones and machine learning.
- More accurate because human error is reduced thanks to machine learning.
- Cheaper for the customer, thanks again to drones and machine learning.
This future is nothing to be afraid of. Sure, these are machine learning enhanced drones with night vision on a search and destroy mission to eliminate underperformance in solar parks, but we think that’s a good thing.