The most significant part of solar plants is PV module (Panel). They shall work at least 25 years with more than 80% of their nominal power. But in operation years, various problems may occur which result in reduction of electricity production of PV modules. For example, one operated by pass diode will reduce PV module production to about 67% of its nominal.
Most of these problem can be found by Infrared (IR) inspection. Utility scales solar plants have large number panels for example a 10 MW plant may include 15000 to 37000 panels in area about 10 Hectares. So manual IR inspection takes time and is not accurate sometimes according to requirements of the IEC 62446-3 standard. As an example manual IR inspection of 10 MW solar plant needs about 10 days. But equipped drones with thermal camera can reduce this time to about 4 hours. On the other hand, results ore accurate and Geotagged. But big Data produced during drone based inspection is a real challenge in this method. For example, drone IR inspection of a 10 MW solar plant will produce about 4000 thermal images.
This is where Artificial Intelligence can help to find defects automatically. Various categories of anomalies that can be found using AI are as below:
- Single cell hotspot.
- Multi cell hotspot.
- Operated bypass diode.
- Open circuited string.
- Short circuited string.
- Disconnected inverter.
We have inspected several utility scaled solar plants with total capacity of about 128 MW in the recent years. Thanks to large datasets gathered in this projects, we have trained our AI algorithms to found anomalies with high precision.
Making operational reports and giving recommendation for each defect to solar plant owner/operator is the final stage of our work.
