Fuel Cell Systems: A Challenge of Multiphysical Simulations - MATLAB
Video Player is loading.
Current Time 0:00
Duration 20:04
Loaded: 0.82%
Stream Type LIVE
Remaining Time 20:04
 
1x
  • Chapters
  • descriptions off, selected
  • en (Main), selected
    Video length is 20:05

    Fuel Cell Systems: A Challenge of Multiphysical Simulations

    Erik Hartmann, SEGULA Technologies GmbH

    The reduction in CO2 emissions leads to a change in drive systems in the mobility sector. While a clear trend towards electrification by means of battery electric systems (BEVs) can be seen in the passenger car segment, the picture is different for light and heavy commercial vehicles. With fuel cells, electrical energy is generated on-board. In contrast to the BEV, this reduces weight and charging time, which is important in the commercial vehicle segment. As part of this work, Segula Technologies developed a multiphysical simulation tool for fuel cell system specification and control. Explore this holistic approach to modeling mechanical, electrical, hydraulical, and control systems with potential technical limitations as well as simulating performance. See examples of how to master these challenges using this tool.

    Published: 30 May 2022

    [MUSIC PLAYING]

    Welcome, everybody. My name is Erik Hartmann. And I would like to show you today our fuel cell development activities, especially in terms of challenges of multiphysics simulation within The MathWorks environment.

    At Segula Technologies, in Russelsheim, we develop vehicles as well as next Gen powertrains where I, as a team leader, be responsible for system simulation and 1D-3D CFD topics. Just a few words about my coauthors at this point. Nils Rohde is responsible for modeling detailed physical systems. Vimal Muthusamy is mainly working on code simulation topics, which includes the control algorithms within the model. And Stephan Schnorpfeil is our main contact person for any hydrogen and fuel cell related application. A special Thanks, at this point, goes to Erin McGarrity and Eva Pelster from MathWorks who supported us last month in terms of troubleshooting and giving advice to improve our model.

    So let's have a quick look at the agenda of this presentation. After a short description of our motivation, I will show you the main circuits and components of a fuel cell system. Regarding the development and simulation approach, I will present you the way we used to build out the model, as well as the main challenges based on a few examples. Finally, a short summary will be provided at the end as well as an outlook where we want to extend and improve our model in the future.

    Our main motivation was to build a model which we can use to make a proof of concept of a complete fuel cell system in early development stages. So typical main questions are is a system capable of deliver the required electrical power, are the balance of plant components right sized, is the control logic working as intended. In general, the model should be able to answer these questions within an acceptable simulation performance. And due to the fact that we are already doing controls development with materials, products, we wanted to achieve that within The MathWorks environment.

    We mainly used MATLAB Simulink, the Simscape tool boxes for physical modeling, and Stateflow for the control strategies. As a starting point, we used a basic fuel cell plant model, especially the stack model, which is already available as documented example case from MathWorks, and we modified it for our needs.

    Why do we focus on fuel cells? Well, there are a lot of reasons in terms of the general reduction of CO2 emissions, especially in the mobility sector. Whereas the battery seems to be the future technology for passenger cars, there's another trend for heavy duty and marine applications with different requirements to the powertrain system. And this is a fuel cell. Especially, high-power demands, short charging times, high payload, and less weight are important in logistics industry as well as transportation of passengers. Nobody wants to lose payload due to a battery which weighs a few tons.

    For us, it means that we need to provide right-sized components of the fuel cell system with focus on efficiency, as well as the ability to scale the whole system based on the final application, from light, heavy duty, to train, marine, and aerospace applications. On the next slide, I'll give you an overview of a basic fuel cell system layout.

    For transportation applications, the proton exchange membrane fuel cell, PEM, also known as polymer electrolyte membrane, is commonly used due to low temperatures, pressures, and stackability. In general, the fuel cell system is a complex one with different physical domains. The stack represents the heart of the system as it collects all input from the different domains, and finally, delivers the electrical energy. Then we have thermal fluids, starting with the hydrogen supply, which represents the anode side, highlighted in green. And for the oxidation process, we need an air supply on the cathode side, which consists of different components, like compressor, heat exchanger, and humidifier.

    Where we have oxidation processes, cooling is required for the components. So we have, in general, two circuits-- a low temperature circuit, which is used for power electronics, heat exchanger, and compressor e-machine. The high-temperature circuit on the other side is mainly used for the fuel cell stack itself to keep the operating temperature at approximately 80 degrees Celsius.

    In addition to the thermal fluids, we have the electrical circuits for the low and high voltage system to follow up the electrical energy output from the stack to the drive train, for example, as well as converting voltage levels and feed the e-machines of compressor and pumps. The interface of these e-machines to compressor and pumps are modeled within the mechanical domain. And finally, a subsystem for controlling several components, as well as valves is needed.

    With that short overview of the different domains, we switch to the modeling approach on the following slides. In general, there is a classical V approach consisting of the left path for system definition, starting from vehicle targets to a single module specification, and the right part for testing on different system levels until the approval for the final concept. At the end, we have successfully reached the targets, or not, and the process goes into the next round.

    Focus of the presentation is the left part regarding the system definition. It is an iterative process where targets may change during development or defined limits cannot be met, like package dimensions of the fuel cell system, which may have an influence on subsystems and modules. Additionally, testing will also influence the system definition side as tests may fail or unexpected problems occur.

    The first step is the definition of targets for the vehicle, for example, acceleration and top speed for a specific driving range. Also drive cycle for certification is an important target, as well as a field of operation. So a high payload for heavy duty trucks, for example. These targets represent the input data for the power train system.

    From the eco perspective, the targets will be transformed into the following typical requirements to the power train system, like power output, transient behavior, vehicle integration, and train capacity. In addition to that, the operating strategy is also important, as you have, in general, two different approaches. The load follow up, which means that the fuel cell will be operated dynamically to provide the required electrical energy, and the battery assists in parallel for high loads and accelerations only.

    In contrary, the range extender approach is similar to a serious hybrid where the fuel cell will be mainly used in a static operating point to constantly charge a battery. The dynamic behavior will finally be covered by the battery itself. So with this information and requirements, we go one step deeper into the fuel cell subsystems.

    For the subsystems, the main question is, how the required system power can be achieved, what kind of components do we need, and how do you place them in the different subsystem circuits. For example, for high power, you will need a compressor. Compressed air needs a heat exchanger, and so on. How many cooling circuits do you need, how many valves-- these are the questions. So in this stage, the basic layout with the required components will be created.

    The final step is a component specification. So how to deliver a required air mass flow, for example, how much pressure drop is acceptable, and so on. So after specifying the components, it is possible to start supplier sourcing in order to find suitable components, which are already available. As I said earlier, the stack itself is the heart of the fuel cell system. It is recommended to select the stack first. In ideal case, you already have measurement data from stack, for example, polarization curves for different conditions, because the stack finally defines the subsystem components, mainly in terms of temperature, pressure, and humidity requirements.

    If the stack modeling is not correct, all other modeled components are not specified correctly as well. For our model, we used the basic stack model for MathWorks and modified it in order to be sensitive to humidity, lambda, and other parameters, based on the measurement data. In order to calibrate the components correctly, it is recommended to use separate subsystems in Simscape as test harnesses.

    With that, you avoid the influence of other components, and you can use fixed in and outputs first, which are mainly driven by the stack. These test harnesses are needed for each component and should be used as reference models, in order to keep the complete system model updated with the calibrated components. So this sums up the general approach, which is an iterative one where a change of the compressor, for example, may have an influence on the whole cathode air delivery side, as well as the adjacent unknown path and ruling circles. So a small change-- big impact. From the general approach, we come to the modeling challenges, where I will show you a few examples on the next slides.

    During the modeling process of different modules, the jet pump was a quite challenging one, as it contains the primary flow from hydrogen storage and the secondary flow from recirculation path, which will be mixed in the middle section of the pump and finally provide it to the stack. The handling of supersonic conditions was a challenge because they occur in the nozzle section of the primary flow and additional math was required, as these conditions are not natively supported via Simscape.

    Another challenge was the system complexity, so that we try to avoid overdetermined system. Physical domains have been marginalized and calculated individually to handle that situation. Purging was a further problem. Purging means that hydrogen from recirculation path will be directly transported to the exhaust when you rent down or shut down the fuel cell. So you have a purge path in parallel to the recirculation path and this leads to a break in the algebraic loop where hydrogen is being extracted. So additional math was also required to avoid stability issues.

    The passive humidifier was a further challenge during the development of the model. Passive means that water from the cathode website from stack outlet to exhaust will be used to humidify the air from the kettle dry side, from ambient air inlet to stack inlet. In general, a high yield humidification level is required for a high stack efficiency. The water transport was a challenge, in order to enable the transport itself, as well as remove the transport water from the wet side.

    In addition, thermal coupling was required to improve the accuracy of the simulation. And finally, a special bypass design was needed as we have a passive humidifier. So only a defined mass flow should be transported directly through the humidifier. And the remaining mass flow is being bypassed on the dry and the wet side.

    As a complete model is also being used for controls development and providing an initial calibration for the fuel cell control unit, there are some challenges as well. A model discretization was required from Simscape continuous domain to simulate discrete domain in order to implement the control algorithms. So some sensitivity studies were necessary to identify the best trade off, in terms of performance and accuracy. For the discretization process, unit delays and quantiles have been used.

    The tuning of the control parameters was a challenge too, as we have a cascaded architecture. So one controller will influence the next sub controller, and so on. The bottom-to-top approach was used to, for example, tune the cooling pump speed control first, and then the coolant mass flow control for overall maintaining the stack temperature. For some controllers, overshooting was allowed, for some, not, which ended in further sensitivity studies to find the right gains, as you can see, basically, in the picture on the right side.

    A third challenge was the parallel ongoing process of improving the simulation model as well. So we also used reference subsystems for the different fuel cell circuits to use the most up-to-date model for the co-simulation. But that means some additional iterative steps during this parallel development, as I mentioned earlier, for the general development process based on the V approach.

    As a final summary of the described process, the fuel cell model accuracy has been improved with detailed implementation of subsystems and components, compared to the simplified model at the beginning. Finally, we were able to size the components, as well as optimize control algorithms, and therefore, provide an initial calibration, ready to use on fuel cell test bench. Besides controller tuning, we use code simulation also in combination with 3D CFD, especially [INAUDIBLE] to provide with drier boundary conditions from the Simscape model. In addition, we got a deeper understanding of the Simscape environment, due to our collaboration with MathWorks in this time.

    The next steps we consider for model improvement and extension will start with the validation on test bench for the complete system, including the balance of plant components. Furthermore, we will increase the flexibility as we add variant models for different components. So it will be easier to match the model to other fuel cell system layouts and parts, for example, if you have an additional turbine or an active humidifier. On top of that, it is also planned to move from the moist air to a custom multi-species domain in Simscape for tracking more species, for example, N2, O2, H2, which allows the implementation of additional system functions and control strategies.

    So I hope I could give you a few insights into multiphysics domains and challenges of fuel cell system simulation. If you have any question, feel free to ask in the following Q&A session or contact me afterwards based on the listed contact information. Thank you very much for your attention. Bye.

    View more related videos