News #20

Optimally Tracked PV Systems with Deep Learning

Photovoltaic systems equipped with solar trackers show a 20 to 30 percent gain in energy yield compared to fixed ground-mounted systems. Besides this, the layout design and alignment can take other criteria into account, such as the light requirements of certain plant varieties underneath agrivoltaic and biodiversity-PV systems or also the grid feed-in at certain times of the day. In the research project "DeepTrack", Zimmermann PV-Tracker GmbH, part of the Zimmermann PV-Steel Group, and the Fraunhofer Institute for Solar Energy Systems ISE are improving tracking algorithms with a digital twin that uses deep learning to calculate optimized control approaches. Among other things, the digital twin learns from the data of its “real” twin, a PV tracker built by Zimmermann PV-Tracker, which is located at the Fraunhofer ISE’s Outdoor Performance Lab in Merdingen near Freiburg.

© Fraunhofer ISE
Tracking PV system of Zimmermann Tracker GmbH at Fraunhofer ISE’s Outdoor Performance Lab in Merdingen near Freiburg.

According to the International Technology Roadmap for Photovoltaics published by the German Engineering Federation (VDMA), 60 percent of all PV power plants worldwide will operate with a tracker system in the future. In countries with high levels of solar radiation, such as Spain, such systems already account for most newly built ground-mounted PV systems. Following the adoption of “Solar Package I” into the German Renewable Energy Sources Act (EEG), strong growth in agrivoltaic (APV) systems with trackers is also expected in Germany. "For APV systems in particular, with its wide variety of crops and systems, we see great potential for tracking PV systems with optimized tracking algorithms," says Hannes Elsen, product manager at Zimmermann PV.

In the "DeepTrack" research project, the company installed one of its tracking PV systems on Fraunhofer ISE’s outdoor test field to obtain measurements under real conditions. Based on the measurement results, the project consortium developed a digital twin that couples PV monitoring and modeling tools with weather forecasts thanks to deep learning. The optimal tracking positions of the PV modules are mapped for different situations.

© Fraunhofer ISE
Low-mounted PV modules with trackers enable agricultural land use between the module rows in the APV system.

"As a first step, we developed control sequences that were geared towards the optimal electricity yield of bifacial solar modules or the best conditions for the plants underneath the APV system," explains Dr. Matthew Berwind, team leader at Fraunhofer ISE. "The next step is to combine the two approaches so that both aspects are maximized as much as possible. Calculating this sweet spot is challenging but possible with our AI-based approach."

The "DeepTrack" research project, supported by the InvestBW funding program of the Baden-Wuerttemberg Ministry of Economic Affairs, Labor and Tourism, is scheduled to run until early 2025. During this period, the researchers will continue to refine and validate the digital twin model by continuously comparing it with the actual performance data to ensure the reliability and effectiveness of this innovative technology. 

Last modified: