Thermocline thermal storage modeling towards its predictive optimal management

The increasing penetration of renewable energies brings into sharp relief the potential of thermal storages, particularly in multi-energy networks. Indeed, they offer low storage cost without environmental concerns. But since physical phenomena during the operation of thermocline thermal storages are complex, associated physical models come with long calculation times. Consequently, their integration within the live management of multi-energy networks is challenging at the current time. Hence, this work proposes meta modeling based on simulation data from previous physical modeling as an alternative to physical modeling. The developed meta model is compared to three different physical models with varying complexity. Compared to a typical physical model with the same spatial dimension, the meta model calculation time is reduced by a factor of up to 1000 while losses as well as internal temperature distribution estimations are equally provided. In an experimental validation based on four different thermal storage configurations with storage capacities between 450 and 2900 kWhth, a normalized root mean square deviation lower than 3% between the meta model and the physical model results is observed. Moreover, physical models often fail to depict real-life behavior of thermal storages, for example due to manufacturing defects or highly dimension-dependent phenomena like flow channeling. In order to overcome this practical challenge, the construction and use of a meta model on the basis of existing experimental data is successfully demonstrated with a normalized root mean square deviation between model and experiments below 4%.