Dissertations / Theses on the topic 'Fuel cells'

This thesis deals with control of fuel cells, focusing on high-temperature proton-exchange-membrane fuel cells.

Fuel cells are devices that convert the chemical energy of hydrogen, methanol or other chemical compounds directly into electricity, without combustion or thermal cycles. They are efficient, scalable and silent devices that can provide power to a wide variety of utilities, from portable electronics to vehicles, to nation-wide electric grids.

Whereas studies about the design of fuel cell systems and the electrochemical properties of their components abound in the open literature, there has been only a minor interest, albeit growing, in dynamics and control of fuel cells.

In the relatively small body of available literature, there are some apparently contradictory statements: sometimes the slow dynamics of fuel cells is claimed to present a control problem, whereas in other articles fuel cells are claimed to be easy to control and able to follow references that change very rapidly. These contradictions are mainly caused by differences in the sets of phenomena and dynamics that the authors decided to investigate, and also by how they formulated the control problem. For instance, there is little doubt that the temperature dynamics of a fuel cell can be slow, but users are not concerned with the cell’s temperature: power output is a much more important measure of performance.

Fuel cells are very multidisciplinary systems, where electrical engineering, electrochemistry, chemical engineering and materials science are all involved at various levels; it is therefore unsurprising that few researchers can master all of these branches, and that most of them will neglect or misinterpret phenomena they are unfamiliar with.

The ambition of this thesis is to consider the main phenomena influencing the dynamics of fuel cells, to properly define the control problem and suggest possible approaches and solutions to it.

This thesis will focus on a particular type of fuel cell, a variation of proton-exchange-membrane fuel cells with a membrane of polybenzimidazole instead of the usual, commercially available Nafion. The advantages of this particular type of fuel cells for control are particularly interesting, and stem from their operation at temperatures higher than those typical of Nafion-based cells: these new cells do not have any water-management issues, can remove more heat with their exhaust gases, and have better tolerance to poisons such as carbon monoxide.

The first part of this thesis will be concerned with defining and modelling the dynamic phenomena of interest. Indeed, a common mistake is to assume that fuel cells have a single dynamics: instead, many phenomena with radically different time scales concur to define a fuel-cell stack’s overall behaviour. The dynamics of interest are those of chemical engineering (heat and mass balances), of electrochemistry (diffusion in electrodes, electrochemical catalysis) and of electrical engineering (converters, inverters and electric motors). The first part of the thesis will first present some experimental results of importance for the electrochemical transient, and will then develop the equations required to model the four dynamic modes chosen to represent a fuel-cell system running on hydrogen and air at atmospheric pressure: cathodic overvoltage, hydrogen pressure in the anode, oxygen fraction in the cathode and stack temperature.

The second part will explore some of the possible approaches to control the power output from a fuel-cell stack. It has been attempted to produce a modularised set of controllers, one for each dynamics to control. It is a major point of the thesis, however, that the task of controlling a fuel cell is to be judged exclusively by its final result, that is power delivery: all other control loops, however independent, will have to be designed bearing that goal in mind.

The overvoltage, which corresponds nonlinearly to the rate of reaction, is controlled by operating a buck-boost DC/DC converter, which in turn is modelled and controlled with switching rules. Hydrogen pressure, being described by an unstable dynamic equation, requires feedback to be controlled. A controller with PI feedback and a feedforward part to improve performance is suggested. The oxygen fraction in the cathodic stream cannot be easily measured with a satisfactory bandwidth, but its dynamics is stable and disturbances can be measured quite precisely: it is therefore suggested to use a feedforward controller. Contrary to the most common approach for Nafion-based fuel cells, temperature is not controlled with a separate cooling loop: instead, the air flow is used to cool the fuel-cell stack. This significantly simplifies the stack design, operation and production cost. To control temperature, it is suggested to use a P controller, possibly with a feedforward component. Simulations show that this approach to stack cooling is feasible and poses no or few additional requirements on the air flow actuator that is necessary to control air composition in the cathode.

This thesis addresses the dynamic behaviour of proton exchange membrane fuel cells (PEMFCs) and alkaline fuel cells (AFCs). For successful implementation in automotive vehicles and other applications with rapidly varying power demands, the dynamic behaviour of the fuel cell is critical. Knowledge of the load variation requirements as well as the response time of the cell at load change is essential for identifying the need for and design of a buffer system.

The transient response of a PEMFC supplied with pure hydrogen and oxygen was investigated by load step measurements assisted by electrochemical impedance spectroscopy and chronoamperometry. Using an in-house designed resistance board, the uncontrolled response in both cell voltage and current upon step changes in a resistive load was observed. The PEMFC was found to respond quickly and reproducibly to load changes. Two transient processes limiting the fuel cell response were identified: i) A cathodic charge transfer process with a potential dependent response time and ii) a diffusion process with a constant response time. The diffusion transient only appeared at high current densities and was offset from the charge transfer transient by a temporarily stable plateau. Transient paths were plotted in the V-i diagram, matching a predicted pattern with overshooting cell voltage and current during a load step.

The transient response of a PEMFC was measured for various cathode gas compositions and gas utilisations (fraction of supplied reactant gas which is consumed in the fuel cell reaction). For a PEMFC operated on pure hydrogen and oxygen, the cell voltage response to current steps was fast, with response times in the range 0.01-1 s, depending on the applied current. For a PEMFC supplied with air as cathode gas, an additional relaxation process related to oxygen transport caused a slower response (appr. 0.1-2 s depending on the applied current). Response curves up to appr. 0.01 s were apparently unaffected by gas composition and utilisation and were most likely dominated by capacitive discharge of the double layer and reaction with surplus oxygen residing in the cathode. The utilisation of hydrogen had only a minor effect on the response curves, while the utilisation of air severely influenced the transient PEMFC response. Results suggested that air flow rates should be high to obtain rapid PEMFC response.

The load-following capability of a single PEMFC was studied by measuring the cell voltage response to a sinusoidal current load with large amplitudes. Effects on the cell voltage response when varying the DC value, amplitude and frequency of the current load were recorded. The load-following capability of the PEMFC was excellent in the operating range where changes in cell voltage were dominated by ohmic losses. No hysteresis in the cell voltage response was observed in this range for frequencies up to 1-10 Hz. In the operating range where changes in cell voltage were dominated by activation losses, hysteresis appeared at lower frequencies (>0.1 Hz) due to sluggish response in the voltage range near open circuit voltage. The increased mass transport limitation imposed when supplying the PEMFC with air caused hysteresis to appear at lower frequencies than for oxygen (above 0.1 Hz, compared to 1-10 Hz for oxygen).

The dynamic behaviour of an AFC supplied with pure oxygen and hydrogen was investigated by load step measurements assisted by electrochemical impedance spectroscopy (EIS). Load step measurements were carried out using an in-house designed resistance board which gave step changes in a purely resistive load. Resistive load steps between various operating points along the polarisation curve were carried out and the corresponding transient response in cell voltage and current was measured. The transient cell response consisted of an initial ohmic drop followed by a relaxation towards the new steady state. The observed response was slower at higher cell voltages. Measured response times varied on a time scale of appr. 10 ms to 10 s, depending on the initial and final voltages. Results from EIS measurements suggested that the potential dependent response time stemmed from the charge transfer reaction at the cathode. Transient response curves were plotted in the V-i diagram and shown to follow a pattern determined by the load resistance and ohmic resistance of the AFC. Results showed that when supplied with pure oxygen and hydrogen, the AFC responded sufficiently fast for automotive applications.

An iso-thermal, one-dimensional, transient model of an AFC cathode was developed, based on mass balances for oxygen and ionic species and floodedagglomerate theory. Model results show the coupled effects of oxygen diffusion, ion transport and propagation of local electrode potential on the response in current density to a cathodic step in electrode potential. For a set of base case parameters, oxygen diffusion and potential propagation, with characteristic time constants of 0.30 and 0.11 ms, respectively, dominated the current response up to appr. 1 ms, while the slower ion diffusion with time constant 5.0 s controlled the final relaxation towards steady state at appr. 60 s. A smaller agglomerate radius and electrode thickness and a smaller double layer capacitance gave faster electrode response. For a cathodic step in electrode potential, an overshoot in faradaic current appeared around 0.5 ms. This overshoot was related to an initially higher oxygen concentration in the agglomerates, but was masked by capacitive current for base case parameters. Simulated response in oxygen concentration profiles suggested that the potential dependent response time found in previous studies can be related to consumption of surplus oxygen in the catalyst layer.

QC 20141028