April 4, 2023

Solar thermal power plants and process heat plants: Higher yields using AI

Artificial intelligence (AI) is already firmly established in many areas of everyday life, work and manufacturing, and is developing at a rapid pace. One of the strengths of AI is its ability to automate processes and analyse large amounts of data much faster than was previously possible. In the recently launched AuSeSol-AI research project, DLR researchers are working with the companies CSP-Services and Industrial Solar, as well as AI experts from the Technical University of Munich, the Jülich Super Computing Centre (JSC) and the fortiss research institute, to develop new AI-based applications for solar power plants and solar process heating systems. The aim is to further reduce the cost of producing electricity and process heat from Concentrated Solar Power (CSP) plants.

AI is particularly good at discovering patterns, repetitions and regularities in large amounts of data. Solar thermal power plants generate a large amount of data that has not yet been systematically analysed. This includes all the operational data collected, such as temperature values from solar radiation receivers, focusing signals from solar mirrors or meteorological cloud camera data. With the help of AI, systems or machines will in future be able to analyse this data automatically, make recommendations for operating decisions or even make decisions themselves.

DLR researchers are developing AI-based models for thousands of tracked mirrors in solar thermal tower systems. These are designed to provide highly accurate predictions of the radiation power in the solar receiver within seconds. The power plant can use these to quickly adjust the solar field settings to the current irradiation situation. 

Another group at the DLR Institute of Solar Research wants to automate drone-based optical measurement technology for solar thermal power plants using AI methods. In the future, image recognition algorithms will speed up and improve the evaluation of aerial photographs.
Tobias Hirsch, project manager at the DLR Institute of Solar Research: "The AuSeSol-AI project brings together researchers from the solar thermal and AI fields with companies from the solar power plant industry that want to integrate AI into their service portfolio. In my view, this is an ideal combination to develop ideas that will lead to new or improved products and services for the industry.”

In the future, CSP-Services plans to use AI-based algorithms in addition to its drone-based optical measurement systems to quickly and automatically analyse other power plant operating data. Klaus Pottler of CSP-Services says: "We want to use the enormous potential of AI to provide power plant operators with even more comprehensive and precise information about the condition of their parabolic troughs or the mirror fields of tower power plants in the future.“

Solar thermal process heating systems are another application focus of the project. Even today, such systems are mainly operated autonomously on the basis of specific settings, i.e. with little or no control by technical personnel. AI should make autonomous operation more robust and efficient. In the future, the software will decide which operating settings the system should use on the basis of evaluated measurement data. One such setting, for example, is the output of the pumps that transport the heat transfer fluid, water, through the plant's evaporator tubes. For a more precise radiation forecast, cloud camera systems with AI-based evaluation algorithms are being developed specifically for use in process heating plants.

The requirements for fully autonomous operation of a process heating system are particularly high: the AI must be able to solve any problems that arise completely independently. How should the control system react if a temperature sensor fails? Should it keep the system running in a safe mode or shut it down? The AI experts are tackling these and similar problems together with Industrial Solar, a company specialising in the development and construction of solar thermal process heat plants.

The three AI-focused research institutes are involved in the project with different groups and experts:

  • TU Munich: Department of Computer Science, Information and Technology
  • Research Centre Jülich, Jülich Supercomputing Centre (JSC)
  • fortiss GmbH, State Research Institute of the Free State of Bavaria

In the course of the project, researchers and companies will test how the various methods work in practice at the Andasol 3 power plant in southern Spain, at a process heat plant in Jordan and at the DLR Solar Tower in Jülich.

Contact

Elke Reuschenbach

Head of Communications
German Aerospace Center (DLR)
Institute of Solar Research
Linder Höhe, 51147 Köln-Porz
Germany
Tel: +49 2203 601-4153