Process Systems Enterprise Limited
email this page print this page
pdf overview

Fuel cell modelling, simulation and optimisation

Approach for true predictive modelling

The challenge: many complex effects need to be considered simultaneously

Effects that need to be considered simultaeously – click to enlarge

 

 

 

 

 

 

First principles modelling – creating the framework

First-principles modelling – click to enlarge

 

Model validation – adjusting parameters to match reality

Model validation – click to enlarge

 

A major challenge of fuel cell design – and the reason that experimental techniques alone are inadequate – is that so many complex phenomena and interacting components need to be considered simultaneously.

Why first-principles modelling is essential

It is no use designing fuel or air flow channels without understanding the how they interact with the chemical and electrochemical phenomena occurring in the electrolyte.

Similarly it is pointless designing a fuel cell stack, with all its complex interactions, without considering the system within which it operates and the likely dynamic load requirements.

Advanced process modelling was specifically designed to handle such complex interacting systems. This is the reason that so many leading fuel cell companies use gPROMS and PSE's Advanced Model Library for Fuel Cells (AML:FC).

The power of first-principles models validated with experimental data

Advanced process models have two attributes that make them capable of a very high degree of predictive accuracy, within a flexible framework.

The first is a detailed first-principles representation of all the fundamental physical and chemical phenomena that characterise the process.

The gPROMS AML:FC fuel cell models typically include all chemical and electrochemical reactions and interactions, the micro-scale diffusion of molecules across and within the layers, the heat transfer across the layers and to the surroundings, and cell deactivation.

Because they involve first-principles modelling they are equally applicable to SOFC, PEM and other fuel cell types.

The second ingredient is accurate model parameter information – for example, reaction kinetic constants, heat transfer coefficients and other material properties – derived from "real-world" laboratory, pilot or operating data.

These are determined using gPROMS's powerful parameter estimation techniques, which use the model in conjunction with experimental data to fit the most accurate possible model parameters from sets of experimental data.

It is the combination of first-principles theory with real-life data that provides advanced process models with a highly-accurate predictive capability over a wide range of design and operating conditions.

In addition to the high-fidelity predictive capability, associated model-based data analysis provides a means to analyse and effectively deploy experimental data, as well as to determine where future experimentation is required.

Hybrid modelling and co-simulation

Hybrid modelling –
Fluent and gPROMS

FLUENT® model of flow channels

Hybrid modelling – Fluent and gPROMS

gPROMS model of membrane-electrode assembly

In some cases it may be advantageous to use gPROMS models in combination with other modelling software, in order to take advantages of unique software features and strengths. For example:

  • it is possible to use gPROMS to handle the complex chemistry and electrochemistry of the anode-electrolyte-cathode structure, and ANSYS Inc.'s FLUENT® for detailed modelling of flow channel hydrodynamics (see AML:FC–FLUENT Hybrid Modelling Interface (AML:FC HMI) for more information).

    The AML:FC HMI includes a number of proprietary model reduction techniques that speed up execution by orders of magnitude over the equivalent CFD-only run, while providing more detailed modelling of the underlying phenomena.
  • gPROMS models can be used with Mathworks MATLAB® and Simulink® environments for control analysis and design (see gO:Simulink for more information).
  • gPROMS models can be used within CAPE-OPEN compliant steady-state flowsheeting packages such as Aspen Technology's Aspen Plus® or Hysys® (see gO:CAPE-OPEN for more information).

Such "hybrid simulation" or "co-simulation" is well-established. Much of the pioneering work in this field was done on fuel cell applications because of their need to take into account complex interacting phenomena simultaneously.

Process optimisation vs. simulation

When working with a high-fidelity predictive modelling approach, the objective of experimentation becomes to maximise the accuracy of key model parameters, rather some aspect of the design or operation.

Once accurate values have been determined for all key parameters, the model is used to optimise many different aspects of design and operation.

Once a fully-validated fuel cell model is available it can be used in conjunction with formal mathematical optimisation techniques to answer questions directly rather than by many trial-and-error simulations.

Examples of a complex optimisation are "choose the flow channel architecture (from a selection of three possible candidates) that will minimise the temperature differential across the anode for a range of typical operating scenarios".

With a well-formulated predictive model, this question can be answered in a single run rather than many trial-and-error simulations coupled with physical testing.

Modelling vs. physical testing and experimentation

The use of high-fidelity predictive modelling and model-based innovation techniques do not eliminate the need for physical testing.

However modelling can reduce the need for testing significantly and make the testing that is performed much more effective.

The application of a model-based innovation approach means that:

  • experimentation and physical testing are directed at improving the model, providing a mechanism to leverage the investment in experimental or pilot data many times over
  • many different design options can be explored with a high degree of confidence, with only the leading candidates going forward to physical testing.

    This allows more design options to be analysed in a shorter timeframe than would otherwise be possible, and allows unsuitable designs to be rapidly eliminated.