OpenFCST is an Open Source simulation code based on the popular dealii finite element library. The current main applications are 2D and 3D simulations for Proton Exchange Membrane Fuel Cells (PEMFCs). The focus of OpenFCST is component development and operating strategies, hence it is based on a very detailed physical description of mass/charge transport, electrochemistry and heat transfer. Numerical accuracy is of paramount importance for this field of application, hence speed has been sacrificed in favour of accuracy.

OpenFCST feature overview. From nanometre to millimetre. Local performance, MEA simulations, flowfield design.

Development Team:

The core developer team for OpenFCST is the Energy Systems Design Laboratory at the University of Alberta, led by Dr. Marc Secanell. The AFCC Automotive Fuel Cell Cooperation Corp. was an early and long time sponsor of this work as well as the Carpe-FC fuel cell network.

As this is an Open Source project, anyone is welcome to join the development team.

Community

Group photo at the 2016 OpenFCST development workshop.

OpenFCST has a small, but very dedicated user and developer community. Thanks to the Carpe-FC network, we could meet in person annually in Vancouver or Edmonton. In addition to the workshops, there are regular teleconference meetings and of course frequent email conversations. We are also trying out hangouts and google groups.

 

 

 

 

Software Development

Source Code Access

OpenFCST source code access is available through:

Release Procedure

  1. Ask Marc for access to the development version on Bitbucket
  2. Decide on new feature to develop
  3. Create a new feature branch, by either forking the project or by creating a new branch on the main repository
  4. Develop …
    • Make sure to merge recent changes of the development branch into your work.
    • Commit frequently.
    • Execute test-suite from time to time.
    • Write some demo cases for the example section.
    • Write regression and unit tests.
  5. Submit pull request

The necessary acceptance criteria for acceptance of the pull request are:

  • Test suite passes.
  • Documentation is readable.
  • Accepted at coders meeting.

Operating system requirements

The easiest way to run OpenFCST is on a Linux system. Most users run either OpenSuSE or Ubuntu. A virtual machine image is available for Windows users.

Capability Overview

Meshing

Several default fuel cell geometries. IO for external meshes. Adaptive mesh refinement based on local solution error.

2D MEA model

Wide selection of MEA models (anode, cathode, full MEA). Several multi-scale sub models for the catalyst layers.

Electrochemistry

Implementation of Butler-Volmer, Tafel and more complex anode and cathode electrochemistry.

3D Microscale

OpenFCST offerers simulation capability on reconstructed porous materials. 3D imaging datasets can be used as simulation domain.

3D Multiphase

OpenFCST offerers simulation capability on reconstructed porous materials. 3D imaging datasets can be used as simulation domain.

Ex Situ Models

QT based graphical user interface. Web applications through python and Pyramid.

GUI

QT based graphical user interface. Web applications through python and Pyramid.

Plug Points

The Plug Point architecture allows easy addition of new materials, layers and algorithms. Automatically integrates with the GUI and the parameter handler.

MEA Models

OpenFCST comes with several fuel cell models built in. The two main model lines are two dimensional cross sections of an MEA (most dominately half channel – half landing geometries) and 3 dimensional porous media reconstructions. The last two years were devoted to add a liquid phase to the models.

 

MEA Mode Showcase: Non-Isothermal

The non-isothermal fuel cell model is the current standard baseline model of OpenFCST. It has been extensively tested and benchmarked in conditions where liquid water is not a primary concern.

This model explicitly solves for the temperature field, electronic and protonic potential, oxygen, hydrogen and nitrogen, liquid water vapour and ionomer water content. The governing equations for all these quantities are coupled in a weakly nonlinear fashion. A finite element discretization of these equations is implemented through the dealii framework, and a Newton solver is used to linearlize the problem.

Governing equations for the non-isothermal MEA model implemented in OpenFCST.

The primary unknowns and their domain of existence for the non-isothermal MEA model.

 

Why do we need a non-isothermal model ?

Specifically at high current densities, a significant amount of heat is produced in the catalyst layer. This can significantly change the water content of the ionomer and hence the protonic conductivity. In addition, de-lamination of the catalyst layer or other anomalies can create hot-spots within the MEA causing significant damage.

Comparison of polarization curves using the isothermal and non-isothermal modules of OpenFCST.

Maximum temperature with the MEA under several operating conditions.

Heat based loss breakdown

The  source terms in the heat transport equations are the loss terms of the transport processes (Ohm’s heating) and electrochemical reactions. Hence a physical performance loss breakdown can be created:

Heat loss based voltage loss breakdown.

This is a significant advantage over using a inappropriate superposition of linear terms in the polcurve.

MEA Mode Showcase: 2-Phase Flow

The past two years we have (well mostly Jie) have implemented a liquid water transport and saturation component for OpenFCST. This extends the non-isothermal drastically, as water management is one of the major design and operating lever for PEMFCs.

Simulations and experimental results under dry and wet conditions.

In contrast to most fuel cell simulation codes which use a saturation based formulation, OpenFCST uses a pressure based approach. Due to the change of the hydrophobic and hydrophilic properties of different porous media within the fuel cell, the saturation is typically discontinuous between layers. Using a saturation based approach, one would have to allow for these jumps by e.g. using discontinuous finite elements. This is very seldom implemented. In liquid and gas pressure based approach, both primary unknowns remain continuous, hence no special treatment at the interfaces is needed. The saturation is calculated based on the contact angles and pore size distributions, and displays a natural discontinuity if the physics demands it,

 

MEA Mode Showcase: Multiscale Catalyst Models

One of the original design criteria for OpenFCST was to allow for multi-scale models, i.e. that we could run analytical or numerical sub-scale models of each integration point of the global mesh. Traditional macro-homogeneous MEA model do not account for additional losses incurred by the nano-scale structure of the catalyst layer. With multi-scale models we can account for some of these additional contributions.

Does it matter ?

Left: Current density density plots (Multi-Scale). Center: Polarization Curves, Right: Current Density Plots (Macro-Homogeneous)

The above simulation show the behaviour of a multi-scale model and a macro-homogeneous model. In case of the macro-homogeneous model, the limiting current is independent of the catalyst layer thickness, whereas the multi-scale model shows a strong dependence, as all of the catalyst layer has to contribute to the current generation. This is the behaviour we expect based on experimental results.

Microscale Effectiveness

The effectiveness of the micro-scale can be probed by comparing the current produced with and without additional microstructure. This allows to probe the effectiveness locally within the catalyst layer:

Distribution of local properties within the cathode catalyst layer.

 

MEA Mode Showcase: ORR Reactions

In addition to the standard correctly implemented Butler-Volmer equation for the cathode ORR reaction, OpenFCST features a Double-Trap model. This accounts for the experimentally observed doubling of the Tafel slope.

Does it matter ?

A change in the kinetics is observable at a global level by a significant change in the polarization curve:

Tafel vs Double Trap model for the ORR reaction on the cathode.

The doubling of the Tafel slope leads to a significant dropoff in the polarization curve at the higher current densities.

On the local scale of the catalyst layer, we observe that more of the catalyst layer has to be utilized to create a given current:

GIF sequence of the local current density in the catalyst layer. Left: Tafel kinetics; Right: Double-Trap kinetics.

MEA Mode Showcase: 3D Microscale Simulations

Multiscale landscape executable through OpenFCST.

Ex Situ Models

In situ validation for fuel cells has always been problematic. We are creating a suite of ex-situ experiments and models, to validate and parameter fit individual parts of a fuel cell model with respect to a controlled ex-situ experiment. Parameter estimation can then be used to define the material properties and appropriate transport model to be used within the OpenFCST in-situ models.

Ex-Situ Showcase: Permeability of Porous Media

Example of ex-situ modeling workflow. Left: Experimental setup to measure permeability and diffusivity. Centre: Numerical simulation setpu. Right: Model validation and parameter estimation.

Ex-Situ Showcase: 2 Phase Flow in Channels

Over the past two years, Alex Jarauta has implemented the two phase flow capability within OpenFCST.

Droplet on inclined Plane

Experimental setup and comparison between experimental and numerical results.

Video: Droplet on inclined plane

Ex-Situ Showcase: Membrane Oxygen trasnport properties

Experimental setup

Experimental setup.

Results

Numerical vs analytical results of the parameter fitting routine.

 

Changelog

Action Date Description
Create 2017.08.31 Draft
Publication 2017.09.01
Status in progress

Disclaimer

I am doing this as a hobby, so take all of this with care …

License to graphs and photos

Unless otherwise stated all figures are published under the most non restrictive Creative Commons License:
CC0
To the extent possible under law, Andreas Putz has waived all copyright and related or neighbouring rights to the graphs and images without explicit copyright statement..

%d bloggers like this:
Title
Caption
File name
Size
Alignment
Link to
  Open new windows
  Rel nofollow