Digital Twins Can Reduce Operating Costs for Photovoltaic Parks

By 2035, the electricity supply is to come almost entirely from renewable energies. This was decided by the German government in April 2022. In 2022, almost 2.2 million photovoltaic systems installed in Germany with a total capacity of 67.4 gigawatts covered 11.6% of gross electricity consumption. By 2040, this figure is expected to be almost six times as much, at 400 gigawatts.

This requires increased expansion of solar parks. In order for solar power to be produced cost-effectively, the performance of the parks must be permanently ensured, technical failures reduced, and the aging of the systems slowed down. Monitoring the systems plays an important role here. When large-scale plants are built, the construction planning data is often entered into the system as initial data, not least for cost reasons. Deviations and changes that arise during construction are not recorded at all, or only incompletely or incorrectly. This results in more or less serious discrepancies between the plan and the actual construction project.

During operation, the lack of alignment between the planned and actual states leads to delayed identification of damage and thus to unnecessary technical failures of various system components and faster aging of the systems.

This is where the VR4PV research project came in. Together with industry partners, Fraunhofer CSP in Halle (Saale) and Fraunhofer IFF in Magdeburg developed a virtual environment and a digital image of photovoltaic systems for future analysis, inspection, and maintenance of photovoltaic parks. A digital image of an entire solar plant was created, combining both the geographical and electrical context of the individual components and enabling the recording and display of data and files relating to these components.

“The automated data acquisition and visualization of systems enables faster detection of technical problems and yield losses in systems and forms the basis for efficient and reliable commissioning, as well as for the subsequent development of operational management and maintenance strategies and the associated services,” says Dr. Matthias Ebert, group manager for PV systems and integration at the Fraunhofer Center for Silicon Photovoltaics CSP.

As part of the project, suitable imaging techniques were used to record the PV systems as a whole, as well as at the component level, in combination with the necessary geolocation. Deep learning methods were used to develop and apply data routines for the automated recording and allocation of plant-relevant variables. In addition, a data management concept was developed for structuring a database of the acquired and modeled data (health records) based on AutomationML. With the help of drone and detailed images and AI-based evaluation, a digital image of the PV system under investigation was created in the form of a 3D model, which can be viewed as a virtual reality application in the Elbedome of the Fraunhofer IFF. “We have thus created the prerequisites for a digital twin of PV systems, the benefits of which will now become apparent in its application in operation and in expansions and conversions,” says Dr. Frank Ryll, group manager “Maintenance and Service Management” at the Fraunhofer Institute for Factory Operation and Automation IFF.

Thanks to the digital structure, the planned status could be compared with the actual status of the PV system. Information was bundled through the detailed recording and organization of digital data, and information that was previously not digitally accessible was documented during installation and maintenance. This allows for more cost-effective operation and reduces energy generation costs. As a result, the running costs for electricity generation from photovoltaics are reduced, the systems are less prone to failure, and are therefore more efficient over a longer period of time.