Development of Smart Solar Panel Monitoring and Protection System from Hot-Spot Effects by utilizing Machine Learning
Jihadsyah Faa^iz Ahmad, Daniel Kurnia, Rahmat Hidayat

Institut Teknologi Bandung, Physics of Magnetism and Photonics Research Division, Physics Program Study, Faculty of Mathematics and Natural Sciences,
Jl. Ganesha 10, Bandung, West Java, 40132, Indonesia

Abstract

The use of solar panels as a low-carbon alternative for electricity generation is increasing as the demand to overcome the climate change problem increases. Nowadays, solar panels are now prospective to be used on a small and individual scale such as in residential houses or small offices. However, in such individual usage, solar panels are often placed in an environment with many surrounding objects that may cause sunlight obstructions forming shadows on the solar panel surface. The shadows can cause hot-spot effects that lead to solar cell panel degradation. Besides regular periodic shadows, there are also incidental sunlight obstructions due to falling leaves, bird droppings, debris, etc. Therefore, the usage of solar panels such as in this environment requires a system for controlling and monitoring the healthiness of the solar panels. Here, we develop such a system by implementing bypass circuits and machine learning. By using a data classification artificial intelligence algorithm, an early warning system can be created and remove the need for the individual to monitor the solar panels. In this presentation, the algorithm was trained on the patterns of current generated by the cells from the 8 hours during daylight. It is then used to determine whether the next-day pattern of the generated current is unusual or not. When the system decides that the pattern is unusual then it will raise an alert showing the presence of unusual sunlight obstructions or shadows on the solar panels. With such a smart system, solar panel health monitoring will be more effective and friendly to the users.

Keywords: solar panel, hot spot effect, health monitoring, machine learning

Topic: System and Applications