Fuel processing for he1 cell power systems By Shabbit Ahmed, Romesh Kumar and Michael Krumpelt, Argonne National Laboratory, USA

Fuel processors to produce hydrogen from conventional and alternative fuels are being developed for use in fuel cell power generators. The design of these fuel processors hinges on many factors that include the temperature and pressure required for the conversion, the type and level of by-product that the fuel cell can tolerate, and the duty cycle of the fuel cell power system. This article reviews the types of fuels being considered for fuel cell systems, the reformer technologies being pursued, and the suitability of the reformers for specific applications. The various components needed in the fuel processor have been identified, and some results obtained from the fuel processor development work being conducted at Argonne National Laboratory have been reported.

Fuel cells are being developed for distributed

and portable power generation and for

consumer applications. Recent advances in fuel

cell stack technology, such as the IOO-fold

reduction in the platinum content of the

electrodes and more economic bipolar plates,

have led to increased interest in the commercialisation of polymer electrolyte fuel

cells (PEFCs). As the stack technology matures, the choice of an appropriate fuel is becoming an

increasingly dominant issue, especially because

of the absence of a viable hydrogen storage

option and a hydrogen marketing infrastructure,

at least in the near term.

oxide fuel cell (SOFC) being developed for

operation above 500°C can readily oxidise

carbon monoxide electrochemically. At the

other end of the fuel cell temperature spectrum, the 80°C polymer electrolyte fuel cell needs a

fuel gas with a low (~50 ppm) content of carbon

monoxide and an even lower (co.1 ppm)

concentration of hydrogen sulfide.

Figure 1 is a schematic diagram of a fuel cell

power system, which highlights the different

fuel processor components that are needed for

the low-temperature polymer electrolyte fuel cell

system. The CO clean-up devices, and perhaps

the sulfur removal unit, downstream of the reformer will not be needed for an SOFC stack.

Fuel cell type determines extent of fuel processing required

Fuel of choice depends on application

The nature and extent of the fuel processing Fuel cells themselves operate on hydrogen,

required is determined by the type of fuel cell to which - at least in the near term - will not be be used and the type of application. At least six widely available. Thus, one or more of the

different types of fuel cells have been available fuels must be reformed to hydrogen (or

demonstrated for electric power generation. a fuel gas containing hydrogen) for use in fuel

Except for the direct methanol fire1 cell, all fuel cells. For automotive fuel cell systems, the

cell types require some degree of fuel processing. commercially available fuels are the

The extent of fuel processing required - that is, conventionaI gasolines and diesel, with limited the allowable levels of carbon monoxide and availability of alternative fuels, such as

other trace species (such as ammonia, hydrogen methanol, ethanol, natural gas and various other

sulfide) in the reformate - is determined fuel blends. Hauer et all’1 compared some of

primarily by the operating temperature of the the fuel choices for fuel cell vehicles and their

fuel cell. In general, the higher-temperature fuel effect on fuel processing. For stationary fuel cell cells have greater tolerance to non-hydrogen systems, natural gas is an attractive fuel because species in the fuel gas. For example, the solid of its ready availability in most urban areas.

Fuel for transportation fuel cells

Methanol is unquestionably the easiest of the

potential transportation fuels to convert to hydrogen. Methanol dissociates to carbon

monoxide and hydrogen at temperatures below

400°C and can be catalytically steam-reformed

at 250°C or less. This low conversion

temperature is advantageous for rapid start-up

of the reformer, a capability that is essential for

vehicular applications. Further, methanol can be

converted to hydrogen with efficiencies of >90%

(efficiency being defined as the lower heating

value of the product hydrogen, as a percentage

of the lower heating value of the methanol feed). Methanol is produced largely from natural gas.

The conversion process itself requires energy,

and thus methanol is less attractive than gasoline

on the basis of a well-to-wheel efficiency.[2]

Other comparisons with gasoline are based on

factors related to the production capacity

relative to the needs of the automotive sector,

the cost per mile, the environmental impact,

refueling infrastructure etc.

Gasoline has more than twice the energy

content of methanol, may be less expensive ($/mile) in the US, and has a well-established

infrastructure that would allow the fuel cell

vehicle a smoother entry into the market.

However, its conversion to hydrogen requires

temperatures >65O”C and tends to produce

considerable amounts of carbon monoxide,

methane and possibly coke. Furthermore,

petroleum-derived fuels (such as gasoline)

contain sulfur and trace amounts of metal which, if not effectively removed, can

irreversibly damage the fuel cell. Although these contaminants may be removed more effectively

at the refinery than in an on-board fuel

processor, that decision will eventually emerge

from market drivers and possibly government regulations.

Natural gas is not a strong contender for

fueling vehicles because of its relatively low

energy density. It requires reforming

temperatures of 700°C or higher. Typically,

natural gas contains over 90% methane, along

Fuel Cells Bulletin No. 12

with higher hydrocarbons, carbon dioxide and

nitrogen. Converting methane to hydrogen can

be highly efficient because of its high hydrogen-

to-carbon ratio. Methane is.also less prone to

coking during reforming, compared to other hydrocarbon fuels.

Fuel for stationary fuel cells

The existing natural gas infrastructure in most

urban areas makes it very attractive for use in

stationary fuel cells. Several US companies are

developing fuel cell power generators for residential

and small business use. Natural gas-fueled,

phosphoric acid fuel cell (PAFC) power plants that

can generate 200 kW have been commercially

available for some time. Higher-temperature fuel cells, i.e. the molten carbonate fuel cell (MCFC)

and the solid oxide fuel cell (SOFC), have also been

demonstrated for stationary power generation from

200 kW to 2 MW In almost all of these stationary

fuel cell applications, the natural gas fuel is

converted to hydrogen by catalytic steam

reforming, often within the fuel cell stack or

bundle, but not necessarily in the anode

compartment of the fuel cell.

Choice of reforming process affected by fuel and application The chemical process and petrochemical

industries use hydrogen in many different

processes, and have manufactured hydrogen for

decades. The conversion of hydrocarbons to

hydrogen is carried out using one of three major

techniques - steam reforming, partial oxidation

and autothermal reforming. It is perhaps

appropriate to review these technologies from

the perspective of the fuel cell applications,

which require one or more of the following:

l Hydrogen production levels that are several

orders of magnitude smaller than in

chemical plants.

l A fuel processor that is compact and

lightweight.

l The ability to cycle through frequent start-

ups and shutdowns (one or more per day).

l A hydrogen production rate that is responsive to the change in demand, which

can vary from 5% to 100% of the rated

processing rate.

l Meeting very strict cost targets.

l Maintaining performance reliably.

Steam reforming (C,H,O, + H,O 3

co2 + H,) is probably the most common

method for producing hydrogen in the chemical

Air

Fuel -J4 Reformer -w Sulfur Shifter -w

co Fuel Removal 7) Removal * Cell

230-l 200-C 350-C 230-35O’C 150-200-C 80%

t-

b

1

Water * Radiator * Burner

Exhaust

lrocess industry. In this process, steam reacts

rith the hydrocarbon (e.g. natural gas) in the lresence of a catalyst to produce hydrogen,

arbon monoxide and carbon dioxide. These eformers are well suited for long periods of

teady-state operation, and can deliver relatively

.igh concentrations of hydrogen (~-70% on a dry

asis). The carbon monoxide and carbon dioxide

re removed from the reformate gas stream by a

ariety of reactions and scrubbing techniques

uch as water-gas shift reaction, methanation,

10, absorption in amine solutions, and iressure swing adsorption. The primary steam

:forming reaction is strongly endothermic, and

eactor designs are typically limited by heat

ransfer, rather than the reaction kinetics.

:onsequently, these reactors are designed to sromote heat exchange and tend to be heavy and

urge. The indirect heat transfer (across a wall)

lakes conventional steam reformers less

ttractive for the rapid start-up and dynamic

:sponse needed in automotive applications.

Partial oxidation reformers react the fuel with

sub-stoichiometric amount of oxygen. The litial oxidation reaction (C,H,O, + 0, j

10, + CO + H,O) results in heat generation

nd high temperatures. The heat generated from

he oxidation reaction is then used to steam-

:form the remaining hydrocarbons (usually

methane and other pyrolysis products), by

ejecting an appropriate amount of steam into

his gas mixture. The oxidation step may be onducted with or without a catalyst.

Autothermal reformers combine the heat

ffects of the partial oxidation and steam

:forming reactions by feeding the fuel, water

nd air together into the reactor. This process is arried out in the presence of a catalyst, which

ontrols the reaction pathways and thereby

etermines the relative extents of the oxidation

nd steam reforming reactions. The presence of

team and the use of an appropriate catalyst

provide benefits, such as lower-temperature

operation and greater product selectivity to

favour the formation of H, and CO,, while

inhibiting the formation of coke.

The processing of hydrocarbons always has the potential to form coke. If the reactor is not.

properly designed or operated, coking is likely to

occur. Thermodynamic equilibrium calculations

can provide a first approximation of the

potential for carbon formation. Figure 2 shows

the amount of carbon that will be formed in the

reforming of isooctane under three different feed

conditions. If the feed consists of only isooctane

and oxygen (from air) at O/C = 1, then coke

would form at temperatures up to 1175°C. If water is added while maintaining the O/C ratio

at 1, the reactor temperature can be lowered to 1025°C before carbon formation occurs. If the

oxygen and water proportions are doubled such

that the O/C = 2, then coking can be avoided at

temperatures above 575°C.

The Figure illustrates the advantage of

autothermal reforming (which injects water in

with the feed) from the standpoint of coking. These equilibrium calculations also show that

hydrogen yields are higher when water is in the

feed stream, because more hydrogen enters the

reactor. Of course, the reaction energies have to

be accounted for in the reforming process, in

addition to the product distribution obtained

from thermodynamic equilibrium calculations. Regardless of the type of reformer, the initial

product invariably contains carbon monoxide. It

can be converted to additional hydrogen via the

water-gas shift reaction (CO + H,O 3

H, + CO,), which is conducted in two reactors: a high-temperature shift reactor followed by a

low-temperature shift reactor. Copper zinc

oxide, the most active low-temperature shift

catalyst, is very sensitive to the temperature and

environment; it deactivates above 250°C. To be

activated, the catalyst needs to be reduced in situ

Fuel Cells Bulletin No. 12 0

1 .E-04

p l.E-08

E

$ l.E-12

E i.E-16 P s

1 .E-20

$$ I.&24 C,H&(0,+3.76N,)

0 +aH,O

1 .E-28

1 .E-32

1 .E-36

100 300 500

:BH,8+2(0,+3.76NJ +4H,O

,!+I,+ 0,+3.76N,)

700 900 1100 1300 1500 Temperature, ‘C

and thereafter isolated from air. Rapid oxidation

of the copper can lead to very high (1OOO’C) temperatures that will at the very least deactivate

the catalyst. The shift reactors are usually the

largest reactors in the fuel processing train, and

can lower the CO level to - 1%. The final CO

reduction to ~10 ppm is presently approached

by a catalytic preferential oxidation (CO + O,a

CO,) step.

Alternatively, some reformers may incorporate

a membrane separator within the reactor, by

which pure hydrogen is extracted from the

reformate gas mixture using a palladium alloy

membrane. The removal of hydrogen helps to increase the fuel conversion and the hydrogen

selectivity of the reforming reaction. In addition,

almost pure hydrogen is made available to the

fuel cell. For high hydrogen recovery with the

membrane separators, the reforming must be

carried out at elevated pressures (20 bar or more). Steam reforming of liquid fuels can

employ this process more easily, because it takes

relatively little energy to pressurise the liquid fuel

and water to the necessary high pressures. With

partial oxidation or autothermal reforming,

compressing the air feed to the operating

pressure of the membrane separator would need

very significant amounts of energy, and result in high parasitic power consumption.

Effect of fuel cell application on fuel cell system The constraints of the fuel cell application have a

strong bearing on the design of the fuel

processor. For example, stationary fuel cell power

generators for residential use could operate on natural gas and generate 2-5 kWe. They are less

severely constrained with respect to size and

4 10 n 2

a 0

0:30 1 :oo I:30 2100

Experiment Time, hh:mmExperiment Time, hh:mm

weight, than those for automotive use. The

stationary units would operate continuously without requiring frequent start-up and

shutdowns, but with some load-following

capability. As these generators might be hooked

into the power grid, this dynamic load-following

need not be as severe as that needed for a typical

automotive driving cycle.

For light-duty fuel cell vehicles (passenger

cars), the automobile manufacturers appear to

have settled on the polymer electrolyte fuel cell

as the fuel cell of choice. Its low operating

temperature of 80°C makes it suitable for fast

start-up, and its high energy density can make it compact. However, the polymer electrolyte fuel

cell has a very low tolerance for carbon

monoxide, hydrogen sulfide and other

contaminants. Thus, the fuel processor for such

automotive fuel cell systems must incorporate

several reformate processing steps, including

sulfur removal, water-gas shift reactors for CO

reduction, and a final CO removal unit (e.g. a

preferential oxidiser). Each of these reactions occurs at distinctly different temperatures.

Furthermore, the shift reaction may need excess

water, and the preferential oxidiser requires oxygen (air) injection in a controlled manner.

The fuel processor therefore must include heat-

exchangers, and air and water injection devices and controls. To maintain high reforming

efficiencies, it is important to thermally integrate these various steps. However, high thermal

integration may compromise the rapid start-up

and dynamic load-following capability needed

for automobile use.

With the appropriate system design all of

these functions can be achieved in a relatively

compact fuel processor that can fit inside a car. For example, during start-up the fuel processor

can be heated rapidly to its normal operating

temperatures by burning a small amount of fuel

and passing the combustion products through

the fuel processor. Appropriate amounts of

excess air may be injected at key locations to

prevent overheating of components. This type of

start-up scenario can be examined in detail

with comprehensive system-level models.

Consideration must be given to any specific

limitations for the catalysts and materials in the fuel processor. For instance, the start-up protocol

described above will not work if a copper-zinc

oxide catalyst is used in the water-gas shift

reactor, for the reasons mentioned earlier. Issues

such as these are being addressed by research into

alternative shift catalysts, which would not be

damaged by exposure to air during start-up (or

during system shutdown between uses).

Catalytic autothermal reformer for gasoline A review of the state of reforming technology

conducted several years ago at Argonne National

Laboratory131 led to the conclusion that fuel

processors based on partial oxidation reforming

are the ones most suited for automotive

applications. This conclusion was based on the

fact that this type of reforming does not need

indirect heat transfer (heat transferred across a

wall), as is needed in steam reformers. This

feature of the partial oxidation system enables the reformer to be compact and lightweight,

since no heat transfer materials or media are

needed. It allows for rapid start-up capability

because the reactor can be heated directly by

feeding in fuel and air, but at a higher air-to-fuel

ratio than is used during normal operation.

Dynamic load-following is also easier, because as

long as the air/fuel ratio is constant, the heat generated in partial oxidation reformers is

directly proportional to the processing rate. In

Fuel Cells Bulletin No. 12

steam reformers the heating rate (across the wall) has to closely match the processing rate, without

which the catalyst can overheat or “quench”.

In addition to a simple partial oxidation process

as described earlier, Argonne has pursued the

catalytic autothermal reforming system because it

has a number of advantages. The use of an

appropriate catalyst allows lower-temperature

reforming, which offers many benefits including

higher system efficiency, 141 lower levels of carbon

monoxide produced in the reformer, and a wider variety of materials of construction and

fabrication options. The combination of water

injection and catalyst selectivity results in

inhibition of coke formation, a significant

advantage in hydrocarbon processing.

At Argonne, we have developed a new class of

catalyst materials that can convert hydrocarbon

fuels - including gasoline, naturaI gas and diesel

- by the autothermal reforming process.I51

Figure 3 shows the concentration of hydrogen,

carbon monoxide and carbon dioxide in the

reformate generated from gasoline in a

cylindrical reactor (3 inches (7.6 cm) in diameter, 14 inches (35 cm) long) filled with

pellets of the Argonne catalyst. The

corresponding hydrogen output was -40 l/min.

The 10% CO in the reformer effluent would

need to be converted to CO, and additional H,

in a water-gas shift reactor. Conventional shift

technology uses a high-temperature shift reactor

at 350-4OO”C, followed by a low-temperature

shift reactor at -2OOOC. With the objective of

using a single-stage water-gas shift reactor, where

the catalyst is much more thermally rugged than

copper-zinc oxide, Argonne has developed an advanced shift catalyst. This material is active at

25045O”C, and it appears to be very attractive for fuel cell applications because it can tolerate

both oxidising and reducing environments, as

well as temperature excursions.

If not removed, the sulfur present in

hydrocarbon fuels will poison catalysts in the

fuel processor and in the fuel cell. In the

petrochemical industry, organosulfur is converted to H,S by hydro-desulfurisation. The

H,S is then converted to sulfur by the Claus process. The sulfur is then recovered as either

elemental sulfur or, more generally, oxidised to

SO, and converted to H,S04.

In fuel cell systems, the approach has been to

trap the H,S with zinc oxide. In a fuel

processor, the sulfur must be removed ahead of

any catalyst that is not sulfur-tolerant. If the

reforming catalyst itself is not sulfur-tolerant,

then the fuel processor becomes complicated

because some hydrogen from the product gas

must be recycled to the front end of the fuel

processor for hydro-desulfurisation of the fuel. A sulfur-tolerant reforming catalyst is highly

desirable, because it would permit the

formation of H,S in the reformer, which can

then be removed in a zinc oxide bed (or other

suitable sulfur trap).

Preliminary calculations indicate that about

8 kg of ZnO are required for an operating

lifetime of 100,000 miles at 80 miles per gallon

(3.5 l/100 km), based on the assumptions that

the US average sulfur content of gasoline is 347

ppm and that the average fuel cell power will be

IO-12 kWe over the lifetime of the fuel

processor. If the sulfur content of gasoline is reduced to lower levels in the coming years, the

amount of zinc oxide needed would also drop.

Investigations are being conducted at Argonne

to verify that zinc oxide will be capable of

meeting this objective, to identify any potential

weaknesses in this method of sulfur removal, and

to develop alternative methods if needed.

Tests are being conducted on an engineering-

scale (6 kWe) integrated fuel processor that

includes the catalytic autothermal reformer, a

zinc oxide bed for the sulfur removal, and a water-gas shift section: Both the reformer and

the shift reactor use catalysts developed at

A rgonne. Preliminary results from this

integrated reactor show that isooctane (a

principal component in gasoline) can be

converted to a fuel gas stream containing 40%

hydrogen with less than 4% carbon monoxide

(see Table 1). With improved process control, the carbon monoxide level could be reduced to

less than 2%.

Conclusions Fuel processing technology for the generation of

hydrogen is receiving a fresh look because of new

applications in fuel cell power generation.

Depending on the type of fuel cell, the specific application and the type of fuel, the fuel

processor system design can change significantly.

The need for smaller, lighter, more responsive

fuel processors producing a high-purity, hydrogen-rich gas has created opportunities for

alternative technologies in reforming, scrubbing

and separation processes.

Argonne Nattonal Laboratory has a multi-

disciplinary programme directed towards the

development of fuel processing technology

suitable for fuel cell systems. Component

development and system integration work has

led to newer catalysts and reactor designs which

have demonstrated higher hydrogen

concentrations at lower temperatures in compact

and lightweight fuel processing hardware.

Acknowledgment This work was supported by the US

Department of Energy’s Office of Advanced

Automotive Technologies. Argonne National

Laboratory is owned by the United States

government, and operated by the University of

Chicago under the provisions of a contract with

the Department of Energy under contract W-

31-109-Eng-38.

References 1. K.-H. Hauer, 0. Duebel, H. Friedrich, W. Steiger, J. Quissek: “Technical requirements for

fuel cell powered electric vehicles”, Powertrain

InternationalZ( 1) (Winter 1999).

2. R.L. Espino, J.L. Robbins: “Fuel and fuel

reforming options for fuel cell vehicles”.

Proceedings of 30th International Symposium

on Automotive Technology & Automation

(ISATA), Florence, Italy, 1998.

3. R. Kumar, S. Ahmed, M. Krumpelt, K.M.

Myles: “Reformers for the production of

hydrogen from methanol and alternative fuels for fuel cell powered vehicles”. Argonne National

Laboratory Report ANL-92/31, Argonne,

Illinois, USA, 1992.

4. R. Kumar, R. Ahluwalia, E.D. Doss, H.K.

Geyer, M. Krumpelt: “Design, integration, and

trade-off analyses of gasoline-fueled polymer

electrolyte fuel cell systems for transportation”. 1998 Fuel Cell Seminar, Palm Springs,

California, November 1998; Abstracts Book,

226229.

5. S. Ahmed, M. Krumpelt, R. Kumar, S.H.D.

Lee, J.D. Carter, R. Wilkenhoener, C. Marshall:

“Catalytic partial oxidation reforming of

hydrocarbon fuels”. 1998 Fuel Cell Seminar, Palm Springs, California, November 1998;

Abstracts Book, 242-245.

For more information, contact: Dr Shabbir Ahmed, Group Leader for Fuel Processing, Argonne National Laboratory, 9700 South Cass Avenue, Bldg. 205, Argonne, IL 60439, USA. Tel: +l 630 252 4553, Fax: +1 630 952 4553, Email: ahmed@cmt.anl.gov

Fuel Cells Bulletin No. 12 0