CFD Modeling of the Hydrogen Fast Filling Process for Type 3 Cylinders and Cylinders Lined With Phase Change Material

Abstract

A simulation of the fast filling of a 195-liter type 3 tank with hydrogen was completed with ANSYS Fluent as a baseline case for developing a CFD model capable of accurately modeling the hydrogen cylinder filling process. 141-second profiles of mass flow and temperature of the incoming hydrogen flow into the cylinder were prescribed from experimental data previously collected at the Gas Technology Institute (GTI) in Des Plaines, IL. All the simulations were completed with the coupled pressure based algorithm with the K-Omega SST turbulence model and real gas NIST properties (REFPROP) to capture the effects of compressibility of hydrogen during the filling process. Gravity was enabled in the axial direction of the cylinder. The initial pressure and temperature in the cylinder were 124 bar and 292.3 K, respectively, with a target, experimental pressure of 383 bar at the end of the filling. For the initial case, the walls of the cylinder were modelled as adiabatic to reduce the computational effort. The final pressure and temperature of the adiabatic wall case matched the experimental pressure and temperature within approximately 30 bar and 6 degrees, respectively. The overall pressure and temperature profiles over the course of the filling process also provided a good match between the simulation results and experimental data. A conjugate heat transfer case with the aluminum liner as part of the domain and an adiabatic outer wall was attempted in order to capture the heat transfer to the liner. The conjugate heat transfer case provided promising results but was taxing in the computational time needed to simulate the entire filling process. A User Defined Function (UDF) for a simple lumped heat capacitance model was applied at the wall to model the wall temperature and capture the heat transfer occurring to the wall while reducing the time needed to complete the simulation. The final pressure prediction for this case was excellent, within 3 bar of the experimental value, and matched it accurately for the duration of the fill process. The final temperature prediction worsened and exceeded the experimental value by 16 degrees Celsius. The UDF model also allowed the ability to easily explore more exotic liners such as Phase Change Materials (PCM) which were also simulated in this work.