A vertical packed bed regenerator is a sensible thermal energy storage for storing high-temperature heat from industrial processes. A wide variety of bulk materials can be used as the storage material. The laboratory setup consists of an upright cylindrical metal container, which is filled with rocks. In order to store heat in the packed bed, hot air flows around it. The hot air cools down and transfers the heat to the bulk material. To discharge the storage, cold air can flow through the heated bed. This heats up and could be used in a real plant, for example, to preheat process air.

For testing the methods developed in the research area Industrial Energy Systems (simulation models, optimization models, fault detection algorithms, ...) on the test setup, the packed bed regenerator is equipped with a large number of sensors and actuators. In order to simulate and analyze the behavior of the storage in the future, the packed bed regenerator is equipped with a digital twin. The digital twin is there to completely transform the behavior of the packed bed regenerator into the virtual space and to make decisions there by comparing it with simulation results. With the help of modern process control software and standardized communication protocols, the behavior of the packed bed regenerator can be precisely measured and automatically transformed into the virtual space. A 3D thermal simulation model forms the virtual entity of the digital twin. This model can reproduce the thermal behavior of the packed bed regenerator very accurately and also predict it arbitrarily far into the future. In addition to synchronizing the physical and virtual entity, the digital twin is also responsible for managing measurement and simulation data and for knowledge representation. The knowledge representation in the digital twin is realized by so-called ontologies. The ontologies store expert knowledge and semantic information, such as the exact positions of the sensors in relation to each other or the occurrence of errors in a machine-readable way. Based on this infrastructure, researchers and students can develop specialized services and algorithms that communicate with each other via the digital twin and benefit from existing knowledge from the ontology.

In a collaboration between the Research Units Fluid Flow Machinery and Industrial Energy Systems, a digital twin of the reversible pump turbine is currently being designed in the lab of the Institute of Energy Systems and Thermodynamics.

A digital twin platform is used to test how machine learning can be used for optimal process control. Therefore, a virtual simulation model is coupled to the real model machine via a SCADA system, which allows a suitable machine learning algorithm to communicate directly with the machine and learn a strategy for optimal control of the pump turbine through direct interaction.