Identifying panel-level faults using edge computing & AI enabled IoT hardware

Module-level faults can have serious affect on the performance and reliability of any solar PV plant. These faults, if not detected timely, can pose safety hazards like fire accidents on site and eventually lead to significant asset damage and/or under-performance. A better fault detection technology through the use of Edge Computing IoT hardware and machine learning frameworks is the need of the hour.?

A Solar PV system could encounter several faults at module-level. These faults include electrical faults like line to line, line to ground, open circuit or non-electrical faults like glass breakage, bird pooping etc.. In our research we have identified and presented how better and real-time fault detection is possible using cost-effective Edge devices and Machine Learning frameworks.

The purpose of the research is to identify and develop module-level fault detection frameworks with classification techniques to build low-cost edge-devices (IoT Sensors) that could be deployed at large scale in low-power-output PV arrays.?

Our research classifies the impact of non-electrical faults (glass breakage, delamination, bird poop etc.) on the current and voltage patterns of the panels. This has helped us to detect these non-electrical faults accurately based on their signature patterns and impact on electrical and thermal parameters of PV modules.??

Through the research we were able to classify more than six types of non-electrical module-level faults in a solar PV system with 93% accuracy using the low power edge devices. The research paper gives a clear idea with concrete results on the effectiveness of machine learning frameworks and cost-effective edge computing for sustainable, profitable and reliable solar PV systems.?