Comparison of a Physical and a Data-Driven Model of a Packed Bed Regenerator for Industrial Applications

Thermal Energy Storage systems are promising technologies to match intermittent heat supply with demand and improve the energy efficiency of industrial processes. To optimally integrate these energy storage systems in industry, reliable and industrially applicable models are required. This work examines two different modeling approaches for a Sensible Thermal Energy Storage device, namely a Packed Bed Regenerator. A physical 1Dmodel using finite difference methods and a data-driven grey box model using Recurrent Neural Networks are described. Experimental data from a Packed Bed Regenerator test rig is used to create the data-driven model and to compare the results of both models with real measurements. A quantitative and qualitative comparison of the data-driven and the physical model is conducted. The results of the quantitative investigation show, that both models are able to capture the complex behavior of the Packed Bed Regenerator. With the qualitative analysis, the features of the different models are highlighted and advantages and limitations are discussed. Thus, it provides an orientation in the decision-making process for the choice of an appropriate modeling approach. The findings of this work can support the creation of physical, as well as data-driven models of sensible energy storage systems and strengthen their implementation to industrial processes. The generic grey box modeling approach and the findings of the qualitative comparison of the models can be also applied to other modeling tasks.