Authors: - Karin Fröjd, Modelon AB, Scheelev. 17, 223 70 Lund, Sweden - Karin Axelsson, PowerCell AB, Ruskvädersgatan 12, 418 34 Göteborg, Sweden - Ivar Torstensson, Modelon AB, Scheelev. 17, 223 70 Lund, Sweden - Erik Åberg, Modelon AB, Scheelev. 17, 223 70 Lund, Sweden - Erik Osvaldsson, PowerCell AB, Ruskvädersgatan 12, 418 34 Göteborg, Sweden - Gregor Dolanc, J. Stefan Institute, Jamova cesta 39, 1000 Ljubljana, Slovenia - Bostjan Pregelj, J. Stefan Institute, Jamova cesta 39, 1000 Ljubljana, Slovenia - Jonas Eborn, Modelon AB, Scheelev. 17, 223 70 Lund, Sweden - Jens Pålsson, Modelon AB, Scheelev. 17, 223 70 Lund, Sweden In this article a real-time model for dynamic simulation of a fuel processor is presented. The model is intended for HIL testing of the PLC for a truck Auxiliary Power Unit (APU) system. The APU comprises a PEM fuel cell and fuel processor to enable direct utilization of on-board diesel. The system is under development in FCGEN, an EU project under the FP7 program FCH JU [1]. One critical challenge is to design the control system (PLC) to ensure failsafe and environmental friendly startup and operation. The startup phase of the fuel processor is the most critical part, since it is a highly dynamic process involving several complex reactors. It is advantageous to verify the control system before the fuel processor is assembled to avoid possible breakage of components. Such verification can be done with a real-time model representing the physical system. In this study such a model is created using Modelica and Dymola. It is shown that it is possible to load and execute a real-time Modelica model capable of realistically mimicking the system response on a HIL platform. The model runs in real time using a first order explicit (Euler) solver with a time step size of 25 ms. Full text article: http://www.ep.liu.se/ecp/096/071/ecp14096071.pdf http://www.modelon.com/news/news-display/artikel/modelica-conference/

1. 1
DEVELOPMENT OF A REAL-TIME FUEL
PROCESSOR MODEL FOR HIL SIMULATION
Karin Fröjd1, Karin Axelsson2, Ivar Torstensson1, Erik Åberg1, Erik Osvaldsson2, Gregor Dolanc3,
Bostjan Pregelj3, Jonas Eborn1, Jens Pålsson1
1 Modelon AB, Ideon Science Park, Scheelev. 17, 223 70 Lund, Sweden
2 PowerCell AB, Ruskvädersgatan 12, 418 34 Göteborg, Sweden
3J. Stefan Institute, Jamova cesta 39, 1000 Ljubljana, Slovenia

2. • Introduction
 The FCGEN project
 The fuel processor
 Motivation for real-time model
• Real-time model
 HIL Requirements
 Model description
• Results & Discussion
• Summary
OUTLINE

3. • Introduction
 The FCGEN project
 The fuel processor
 Motivation for real-time model
• Real-time model
 HIL Requirements
 Model description
• Results & Discussion
• Summary
OUTLINE

4. • FCGEN is a research project under the EU 7th
Framework programme1
• Goal: Truck Auxiliary Power Unit (APU)
powered by diesel fuel:
 Based on PEM Fuel Cell
 Fuel Processing Module to convert diesel fuel to
syngas
• Partners: Volvo GTT, PowerCell, Jülich, JSI,
IMM, JM, Modelon
• Modelon provides the simulation solutions
FCGEN PROJECT
1 FCGEN is financially supported by the European Union’s
Seventh Framework Programme (FP7/2007‐2013) for the Fuel Cells and
Hydrogen Joint Technology Initiative under grant agreement n° [277844]

5. • Conversion of diesel to syngas by steam
reformation
• Syngas cleanup and hydrogen enrichment by DS,
WGS, PrOx reactors
• Exhaust gas cleanup by CAB
• Startup help by CSB
FUEL PROCESSOR MODULE

6. 1. Use start burner to heat ATR and reactors
to sufficient temperature.
2. Use start burner to produce steam for ATR.
3. Ignite ATR, keep start burner in operation to
ensure sufficient fuel and emission
conversion.
4. Shut down start burner, start normal
operation.
5. Stop by-pass of fuel cell.
FPM STARTUP PROCEDURE

7. • FPM startup procedure is complicated:
 Duration: ~0.5-1 h
 6 chemical reactors:
Proper startup
Avoid poisoning
Avoid thermal damaging of catalysts
 Minimize emissions
 7+ compressors, 10+ pumps, 7+ valves, …
PLC design and testing is crucial and complicated
• Model based PLC testing:
 Cost: Avoid expensive system damage
 Availability: Available before system is assembled, and
during system reassembly.
MOTIVATION – REAL TIME MODEL
Goal: A realistic plant model which can be used in a HIL rig for PLC testing

8. • Introduction
 The FCGEN project
 The fuel processor
 Motivation
• Real-time model
 HIL Requirements
 Model description
• Results & Discussion
• Summary
OUTLINE

9. Aim: A model for PLC logic testing
• Robustness: No crashes
• CPU time: No overruns on the HIL computer;
Euler time steps of 25 ms
• Accuracy:
 Realistic response to changes in input
signals.
 Capture trends for temperatures and mass
flows. Actual numbers need not to be caught.
 Response times are allowed to deviate
slightly from reality.
HIL REQUIREMENTS

10. FCGEN-RT model
• Based on Fuel Cell Library from Modelon
• All relevant components (25+) included in the
model:
 Reactors
 Heat exchangers
 Valves
 Pumps
 Compressors
• Media for water, steam, reformate, air
• 217 continuous time states
• Control system model – for testing
MODEL DESCRIPTION

11. • Model simplifications: Discretized heat exchangers
and reactors replaced by lumped models.
• Non-linear equations: Simplified submodels, allowing
linear equation systems.
 Linear friction losses
 Simplified chemistry
• Event elimination using noEvent().
• Zero flow / back flow elimination: Minimum mass
flow applied when control signal is zero.
ACHIEVING REAL-TIME

12. • Chemical time scale << 25 ms: Replaced dynamic
equations with static:
 Equilibrium (QSS)
 Complete conversion (infinitely fast reactions)
• Flow time scale << 25 ms: Increased flow losses and
volumes.
• Media: Replaced high order NASA Glenn correlations
by linear correlations covering the full temperature
range.
ACHIEVING REAL-TIME

13. • Plant model imported in Simulink by the
Dymola Simulink Interface
• SpeedGoat connected to PLC by CAN buses
• Remote access for convenient testing by JSI
HIL SETUP
Simatic&HMI PC
(supplied by JSI,
located at Powercell)
Remote access PC
(remote access from
Slovenia)
Ethernet
Internet
PLC
(APU Control
System)
Real time
computer
(runs real-time model)
Host computer
(Dymola/Simulink plant
model)
Ethernet
Ethernet
CAN bus
data signals
PLC
with
CS
Simatic & HMI
computer
Host
computer
Real-time
computer

14. • Introduction
 The FCGEN project
 The fuel processor
 Motivation for real-time model
• Real-time model
 HIL Requirements
 Model description
• Results & Discussion
• Summary
OUTLINE

15. • Offline tests of startup in
Dymola
• Temperature traces reasonable
RESULTS
Startup sequence:
1. Start CSB, heat inlet air for ATR
2. Use CSB to produce steam for ATR.
3. Ignite ATR, continue CSB operation
4. Shut down CSB, start normal operation.
ATR inlet temperatures. The numbers below the x axis
indicate start of phase 1-4.
Reactor exhaust temperature. The numbers below the
x axis indi-cate start of phase 1-4.

16. • Offline testing in Dymola
• Comparison between detailed model and model with
increased flow losses and volumes
• Accuracy is kept
RESULTS
Temperatures before (*_orig) and after (*_realtime)
modifications of volumes and flow losses to eliminate time
constants below 25 ms
Mass flows before (*_orig) and after (*_realtime)
modifications of volumes and flow losses to eliminate
time constants below 25 ms.

17. • The model loads and starts successfully on
the Speedgoat
• CPU time
 68 times faster than real time using
DASSL on a laptop (Intel Core™ i7-3740QM
CPU @ 2.70 GHz)
 5.6 times faster than real time using Euler
steps of 25 ms. CPU time for one Euler
step: 4.5 ms
 CPU time on Speedgoat for one Euler step:
~23 ms
RESULTS

18. A challenge to get such a large system of
reactors real time capable. Toughest tasks:
1. Time constants: To achieve time
constants of 25 ms while keeping
accuracy.
2. Robustness: To ensure convergence and
no events under all conditions.
3. Event removal
DISCUSSION

19. • Introduction
 The FCGEN project
 The fuel processor
 Motivation for real-time model
• Real-time model
 HIL Requirements
 Model description
• Results & Discussion
• Summary
OUTLINE

20. • Real-time model of a FPM was created
 Euler steps of 25 ms
 Reasonable results
 Loads successfully on a Speedgoat
machine
• Next step: Use for control logics testing
SUMMARY AND OUTLOOK