Analysis of a horizontal flow closed loop thermal energy storage system in pilot scale for high temperature applications – Part II: Numerical investigation

Abstract

The storage system investigated in this work, namely the CellFlux system, consists of a regenerator type thermal energy storage volume which is coupled to a heat exchanger by a circulating intermediate working fluid. The numerical simulations in this work are based on experiments conducted with a large scale pilot plant having a bed length of more than 10 m. The storage volume is of a novel design with horizontal flow direction and is filled with hollow bricks as sensible heat storage material. So far, most publications focus on packed bed storage systems, often with molten salt or oil, only few consider regularly shaped channels with gaseous flow. For the investigation in this part of the publication, a one-dimensional dispersion concentric model for channel flow is implemented in MATLAB/Simulink.The analysis in part I of this publication has revealed a radial flow maldistribution, which can be predicted by a correction function correlated beforehand. The commonly used assumption of plug flow, however, will be afflicted with an error. Thus, the aim of this work is to determine the accuracy of the model both under the plug flow assumption and the application of aforementioned correction function. Three single blow experiments, where the storage is charged or discharged from a uniform initial temperature, are compared to the numerical model. The average temperature deviation between experiment and simulation reaches up to 10% without the correction function, but improves significantly through its application and then ranges between 0.4% and 4.8%.Moreover, studies based on larger experiments have either not observed or generally not addressed the effect of a flow maldistribution. This appears to be an issue for small scale experiments where typically more sophisticated models are applied. If it was possible to model the entire storage volume with the present 1D model, the flow maldistribution effects could average out. Also, computing effort would be low and allow simulations on system level. This approach was taken only by few authors so far, particularly for packed bed systems and was also not experimentally verified. In order to correctly depict the experimental set up, the model considers the different composition of the interior, such as flow distributors and inlet/outlet cones as well as the thermal heat capacity of the walls and losses to the environment. As a result the model shows good agreement with the experiments: the mean temperature difference of the exit temperature between experiment and simulation during cyclic operation remained below 5%.