Draft, July 16, 2002 1

Development of a 4kW System with PEM Fuel Cell and Electrolyzer with H2 and O2 Storage

Draft July 9, 2002 Report to:

University Research in Sustainable Technologies

Toxic Use Reduction Institute (TURI) University of Massachusetts, Lowell

By

Adarsh Das and John Duffy U Mass Lowell Solar Engineering Program

In cooperation with Michael Kimble

ElectroChem Inc., Woburn, MA

Draft, July 16, 2002 2

TABLE OF CONTENTS

1. ABSTRACT:.................................................................................................................3

2. OVERALL OBJECTIVES: ..............................................................................................4

3. PREVIOUS WORK:......................................................................................................5

4. PRIOR WORK ON THIS PROJECT...............................................................................7

1. Heat balance and Thermal Model for existing 400W electrolyzer.......................................... 7

2. Heat Balance and Thermal Model for (proposed) 1600W electrolyzer ................................ 18

3. Thermal model for existing 4kW fuel cell stack ................................................................... 21

5. DESIGN TOOLS BASED ON CALCULATIONS AND THERMAL MODELS...........ERROR! BOOKMARK NOT DEFINED.

6. CONCLUSION ...........................................................................................................26

7. APPENDICES: ...........................................................................................................27

Draft, July 16, 2002 3

1. Abstract:

The solar electrolyzer/storage/fuel cell system has the potential to meet the needs of billions of people without electricity in an environmentally sustainable way. In the overall system design, a proton exchange membrane (PEM) electrolyzer and a PEM fuel cell are combined with high-pressure hydrogen and oxygen storage in a novel way in a closed water system. When combined with renewable energy technologies, such as photovoltaic, wind, or hydro, the operation of such a system will result in no emissions whatsoever. The manufacture of such a system will have minimal impact on environmental and public health. Several Massachusetts companies have the potential to manufacture several of the components of such a system. Fuel cells generate approximately as much heat as useful electricity, while electrolyzers convert approximately half the electrical energy input to heat. The rejection of the heat generated, from the relatively small volume of the fuel cell, is the main objective of this project. Heat rejection is approached from several different viewpoints. An overall heat balance is performed, treating the electrolyzer as a control volume, and the heat transfers due to the mass flows are accurately estimated. Modular, scaleable, finite element thermal models are created, using thermal analysis software provided by ElectroChem Inc. Data from real experiments and the actual geometry of the functioning systems are used as the basis of the thermal models. For fuel cells, heat convection to the surrounding air from the fins between consecutive cells, is the main method of heat rejection. The effects of increasing this area, and also the material used to make the fins, are calculated. For electrolyzers, regulation of water flow rate and the temperature differential across the oxygen side are found to be the most convenient methods of controlling heat transport.

2. Objectives:

The overall objectives for the project as a whole are the following:

•Design a small solar electrolyzer/fuel cell system, with H2, O2 storage and recycled water, for remote areas. •Design and build a system for backup energy for telecommunication market (battery replacement) •Design, construct, use a prototype system for proof of concepts and for development •Develop, test new electrolyzer, fuel cell The objectives for this academic year included the following:

1. Examine the mass flows and heat balances for the 400W electrolyzer. Develop a thermal model for this system.

2. Scale up the electrolyzer to 1600W, using the methods developed for the 400W electrolyzer. Recalculate the mass flows and heat balances, and develop a model for this system. 1600 W would be the size of the electrolyzer required to

3. Develop a thermal model for the existing 4kW fuel cell stack. Analyze the heat transfer obtained for different fin materials and exposed areas.

4. Develop a model for the air flow required to ensure adequate heat transfer by forced convection

5. Enable the use of the above calculations and thermal models as design tools for ElectroChem clearly defining the inputs and outputs and underlying assumptions for each model.

6. Differential Pressure Relief Valve Design

Draft, July 16, 2002 5

3. Previous Work:

Fuel cells that run on renewable hydrogen to produce electricity are receiving considerable attention recently due to:

• No emissions besides water from the fuel cells, • Developments in less expensive fuel cells (particularly proton exchange

membrane (PEM) systems), • Hydrogen production from solar sources (such as electrolysis from electricity

generated by photovoltaic or wind systems and solar photochemical processes), • Use on prototype zero-emission vehicles (e.g. buses in Chicago and Vancouver), • Potential use on remote stand-alone solar electrical systems, • Developments in hydrogen storage technologies (such as in metal hydrides,

carbon "Bucky balls," and iron sponges). This project is a part of an ongoing project focused on a promising renewable energy system for use in remote areas, being carried out at University of Massachusetts, Lowell. A subset of that system is intended also for backup power for critical telecommunication systems. The PEM fuel cells along with hydrogen and oxygen storage would replace batteries. Electricity would be produced from solar energy with photovoltaic modules, converted to hydrogen and oxygen in an electrolyzer, and stored in reliable containers. When power would be needed for lights, water pumping, communication equipment, etc., electricity would be produced in the fuel cell with the only byproduct being water, which is recycled back to the electrolyzer. Storage of larger amounts of energy with the hydrogen fuel cell system would be more feasible and probably more reliable than with battery systems. In general, considerable research, development, and commercialization of fuel cells have been undertaken. There are many different types of fuel cells. At present, PEM, phosphoric acid, and alkaline fuel cells appear to be feasible for the application of interest here. There are several hydrogen storage technologies under development. The principal approaches are:

• Compression: Storage of hydrogen at 3000 psi in high-strength composite tanks. • Cryogenics: Storage at a liquid at a temperature of -435 F. • Carbon Adsorption: Hydrogen bound to carbon structures. • Metal Hydrides: Storage as a solid • Methanol, gasoline, and natural gas: Hydrogen reformed from other fuels as

needed but with some emissions resulting. • Iron Sponge: Hydrogen released from iron by oxidation reaction.

The first approach because of its simplicity and cost was chosen for the system here. Reliability and safety issues are addressed in the designs. Factors of safety of at least three are being used in sizing components for yield stresses. The ultimate goal is to have the system be be safer than existing battery systems. Future versions of this system are expected to utilize alternative storage systems.

Draft, July 16, 2002 6

PEM electrolyzers have been developed and operated to pressures of about 200 psi (Ledjeff et al., 1994), but considerable difficulty and expense are encountered with seals and membrane reinforcement. Our approach will allow for much higher storage pressures with electrolyzers built for only about 50 psi. Solar electrolyzer systems have also been developed (e.g., Morimoto, et al., 1986; at Humboldt State U. in CA: Chamberlin et al., 1995; Garcia-Conde et al., 1993). We and ElectroChem have done a literature and patent search, and our system appears to be unique. Some of the relevant patents surveyed are listed below:

• United States Patent # 4,528,251; Yamaguchi et al., 9 July 1985, "Differential-Pressure Control Device For A Fuel Cell"

• United States Patent # 3,839,091; Bloomfield et al., 1 October 1974,

"Regenerative Fuel Cell"

• United States Patent # 3,507,704; NASA, 21 April 1970, "Electrolytically regenerative Hydrogen-Oxygen Fuel Cell"

• United States Patent # 4,769,297; Reiser et al., 6 September 1988, "Solid

Polymer Electrolyte Fuel Cell Stack Water Management System U Mass Lowell has experience in developing solar systems for remote power and design methods for optimizing such systems:

Solar rural electrification by former students Priyantha Wijesooriya in Sri Lanka and Harish Hande in India. Their work is considered a model for other countries. (Hande, 1999; Hande, Martin, Duffy, 1998; Hande and Duffy, 2001).

Development of solar system design methods (Bloom and Duffy, 1988; Duffy and Frye, 1989; Wijesooriya and Duffy, 1992; Reinhardt, 2000).

Development of solar crop dryers for Central and South America with the Mesoamerican Development Institute in Lowell (founded by former students) (Raudales, Villanueva, Munger, and Duffy, 1998)

Development of solar systems for rural medical clinics in the Andes of Peru for transceiver radio communications, vaccine refrigeration, lights, and water pasteurization (Duffy, 1999)

ElectroChem has fifteen years experience in developing, manufacturing, and marketing PEM fuel cells for laboratory use. It's expertise in electro-chemistry and chemical engineering complemented our capabilities in mechanical and solar engineering. The system we developed is summarized below.

Draft, July 16, 2002 7

4. Prior Work On This Project

In the 1999-2000 academic year, UML and ElectroChem developed the original solar electrolyzer/fuel cell system concept into a paper design of a full-scale power system. This system would deliver 4 kW of power over 4 hours, and given sufficient input energy from a photovoltaic array or a utility grid, could recharge itself over 40 hours. ElectroChem intends to commercialize a system of this type in the near future as a grid backup system for premium power applications. A reduced-scale proof-of-concept system was then designed, intended to operate at pressures up to 200 psig. This system would allow exploration of the behavior of the target system in an attainable, cost-effective manner. All the components of this test system were manufactured by UML, purchased (from mostly Massachusetts suppliers), or supplied by ElectroChem. Work by A. Arkin and J. Duffy (2001) complemented this project by considering the task of integrating solar power supply with the energy storage and conversion system. The system design above included an electrolyzer in which, electrical energy is used to dissociate water into hydrogen and oxygen, and hydrogen and oxygen are pressurized for use in a fuel cell. The fuel cell transforms hydrogen and oxygen into electricity and water; and a hydrogen and oxygen storage system capable of pressures of 2000 psi. A prototype system for "proof of concept", consisting of an ElectroChem electrolyzer and a storage subsystem to operate at hydrogen and oxygen system pressures of up to 200 psig was fabricated, refined, tested, and modified. The U Mass Lowell team of Professor J. Duffy and D. Shapiro did the micro-level design and fabrication of the prototype experimental system. Daniel Shapiro’s MS thesis (2002) summarizes nicely the previous work on this project. 1. Heat balance and Thermal Model for existing 400W electrolyzer

The main objective of evaluating the heat balance for the system is to find the magnitude of heat that needs to be rejected from the stack to the surrounding environment. A detailed thermal model of the stack was then used to investigate the temperature distribution on the stack, given varying quantities of heat that needed to be rejected.

H2

H2 to fuel cell...delta-pressure relief

O2 to fuel cell...

Circulating Pump

O2

Electrolyzer

OxygenTank

HydrogenTank

Accumulator

Draft, July 16, 2002 8

Heat Balance Energy balance for the electrolyzer can be written as:

fins from convection gasesby removed by water removed iselectrolys during generated Q Q Q Q ++=

Equation 1

When in operation, the electrolyzer has water circulating on the oxygen side. A part of this water is converted to hydrogen and oxygen, while the rest remains in circulation. The second and third terms above, representing the heat removed by the water and gases, consist of gas and water outflow occurring separately, on the oxygen and hydrogen side outlets. The analysis is complicated also by water transport across the proton exchange membrane due to osmotic drag from the oxygen side to the hydrogen side. This makes a control volume approach to the heat balance necessary. This can be written as:

fins from convection outflow massin contained iselectrolys during generated inflow massin contained Q Q Q Q +=+

Equation 2 _________________________ ________________________ _______________________ ___________________ Term 1 Term 2 Term 3 Term 4

The cell energy balance comes from the fact that the enthalpy flow of the reactants entering the cell will equal the enthalpy flow of the products leaving the cell plus the sum of three terms: (1) the net heat generated by physical and chemical processes within the cell, (2) the dc power input to the cell, and (3) the rate of heat loss from the cell to its surroundings. The calculations for the individual terms of the above equation are shown below:

Term 1: Q contained in mass inflow = Q contained in water inflow + Q contained in gas inflow = inwaterwaterpinwater TCm ___&

Equation 3

The volume flow rate of water to the electrolyzer is controlled by the power input to the water circulation pump. Based on a calibration experiment for the magnetic water pump used, the water flow rate was found to be about 0.5 liters per minute, at about 4volts. Calculations were performed for a range of flow rates around this value: water_inflow .25 .35 .5 .65 .75( ):= Based on the above, the total mass of water coming in was calculated as below:

Rate_water_in_O2_side water_inflowLmin

⋅:=

mdot_water_in Rate_water_in_O2_side ρ_water⋅:=

mdot_water_in 4.167 5.833 8.333 10.833 12.5( )kgs

10 3−=

Term 2:

The heat generated during electrolysis is estimated from the difference between the voltage required for the ideal and the actual performance of the electrolyzer.

Draft, July 16, 2002 9

The energy required for the reaction is determined by equating the Gibbs free energy, ∆G, of the reaction, which is given Nernst Equation, which comes from the change in converted to useful hydrogen at the end of the reaction,

H2O (l) → H2 + ½ O2 Equation 4

The free energy change for the reaction can be calculated as: O(l)HG - OG ½ + HG = G 2

02

02

0r∆

Equation 5

At standard conditions of 25°C (298K) and 1 atmosphere, the chemical energy (∆H = ∆H0) in the hydrogen / oxygen reaction is 285.8 kJ/mol, and the free energy change is 237.1 kJ/mol. [32] Thus, the thermal efficiency of an ideal electrolyzer operating reversibly, on pure water at standard conditions would be:

HG

∆∆=η

Equation 6

83.08.2851.237 ==idealη

Thus, the maximum theoretical efficiency of the electrolyzer can be 83% The ideal performance of a PEM fuel cell or electrolyzer depends on the electrochemical reactions that occur between hydrogen and oxygen in the presence of platinum and other electro-catalysts.

However, in actual performance, losses occur due to activation1, ohmic2 and concentration3 polarization, which appear as an increase in the cell voltage required. As a result, the losses can be estimated by comparing the increase in cell voltage with the ideal performance. The final relation is represented by:

actual

ideal

actual

ideal

inin

useful

VV

IVIV

E

G

EH

⋅=

⋅⋅⋅

=∆

∆

=∆

∆=

83.0)(83.083.0η

Equation 7

From the above relation, the total heat rejected by the stack can be calculated as per the following relationship:

IV

VIVV

VE ideal

actualactualactual

idealin )

83.0()

83.01()1(Q iselectrolys during generated ⋅

−=⋅⋅

−=∆−= η

Equation 8 1 Activation polarization loss is dominant at low current density. At this point, electronic barriers have to be overcome proor to current and ion flow. Activation losses show some increase as current increases. 2 Ohmic polarization losses vary directly with current, increasing over the whole range of current because cell resistance remains essentially constant. 3 Concentration polarization occur over the entire range of current density, but become prominent at high limiting currents where it becomes difficult to provide enough reactant flow to the cell reaction sites.

Draft, July 16, 2002 10

The actual calculation is shown below:

N_cells 4:=

V_ideal 1.23V( ) N_cells⋅:= V_ideal 4.92V=

V_ineff V_eleV_ideal

0.83−:= V_ineff 2.072V=

Q_ineff V_ineff I_ele⋅( ):= Q_ineff 41.446W= Term 3:

Q contained in mass outflow = Q contained in water outflow + Q contained in H2 outflow + Q contained in O2 outflow = )( _2____2_____22__2 sideOoutwaterwaterpsideOoutwateroutwaterpinwateroutHHpinH TCmTCmTCm &&& ++

Equation 9

The production rates of gases were calculated using a coulombic efficiency of 100% for a given level of current. This calculation is abstracted from earlier research conducted as part of this project by Daniel Shapiro [32]. The changes in mass flows inside the electrolyzer due to the consumption of water, and the osmotic drag was then calculated as below:

Rate_water_out_H2_side Rate_water_transported:=

water on H2 side:Rate_water_out_O2_side Rate_water_in_O2_side Rate_water_loss−:=

Rate_water_loss Rate_water_consumed Rate_water_transported+:=

water on O2 side:

DS SysModel WorksheetRate_water_transported 0.008402Lmin

⋅:=

DS SysModel WorksheetRate_water_consumed 0.0014Lmin

⋅:=

DS System Model WorksheetDS System Model Worksheetstd liters/minRate_O2_production 0.2786Lmin

⋅:=

DS System Model Worksheetstd liters/minRate_H2_production 0.55719Lmin

⋅:=

Calculation of Mass Flow Changes inside the electrolyzer, for a current of 20A:

Temperatures used for the heat calculations are given below. T_ambient 32 273+( ) K⋅:=

T_outlet 46.5 273+( ) K⋅:=

T_inlet 45 273+( ) K⋅:=

T_ambient 32 273+( ) K⋅:=

T_outlet 50 273+( ) K⋅:=

T_inlet 45 273+( ) K⋅:=

Draft, July 16, 2002 11

Once the mass flows are obtained, the temperature differential between the inlet and the outlet is obtained based on the actual temperatures 4 between inlet and outlet, as shown below:

Q_H2 mdot_H2_outCp_H2⋅ T_outlet⋅:= Q_H2 3.005W=

Q_O2 mdot_O2_out Cp_O2⋅ T_outlet⋅:= Q_O2 1.534W=

Q_gases Q_H2 Q_O2+:= Q_gases 4.539W=

Q_water_out mdot_water_out_H2_side mdot_water_out_O2_side+( ) T_outlet⋅ Cp_water⋅:=

Q_water_in mdot_water_in( ) T_inlet⋅ Cp_water⋅:=

Q_water_in 5.565 7.791 11.13 14.469 16.695( ) kW=

Q_water_out 5.56 7.796 11.151 14.506 16.742( ) kW=

Q_water Q_water_out Q_water_in−:=

Q_water 5.061− 5.439 21.189 36.939 47.439( ) W=

Term 4: The convection from the fins is calculated by using the mass balance presented in Equation 10.

outflow massin contained iselectrolys during generated inflow massin contained fins from convection Q - Q QQ +=

Equation 10

The final calculations for the thermal model are shown below: Q_convected Q_ineff Q_water− Q_gases−:=

Q_convected 41.968 31.468 15.718 0.032− 10.532−( ) W=

Q_convected_pcQ_convectedQ_ineff

:=

Q_convected_pc 101.259 75.925 37.924 0.078− 25.412−( )%=

Effect of water flow rate: The value of heat removed from the electrolyzer by water is based only on the temperature differential for water between the inlet and the outlet of the electrolyzer. This can be more than the required heat convection of ~42W, as calculated from the electrolyzer efficiency, resulting in two negative values above.

4 Experimental values from October 10, 2001, measured by Daniel Shapiro and Adarsh Das.

Draft, July 16, 2002 12

Different water flow rates, and the resulting heat removal from the electrolyzer are summarized in the table below: Water Flow Rate Liter / min 0.25 0.35 0.5 0.65 0.75Water Heat Removal Watts -5.061 5.439 21.189 36.939 47.439 It is seen that for a flow rate of 0.75 liters per minute, the electrolyzer loses more heat than it gains due to the conversion of electrical input into heat resulting from the inefficiency of the electrochemical reaction. Therefore, the water flow rate is seen to be the main factor affecting the amount of heat rejection required for the electrolyzer.

Draft, July 16, 2002 13

Thermal Model: The heat rejected from the fins by convection, as calculated above, was divided by the total active area5 of the stack to calculate the surface heat load on the active area of the electrolyzer. This surface heat load was used as one of the inputs to a finite model of the stack, containing its three dimensional geometry and material thermal properties. Thermal Analysis System (TAS), was used for creating the model. [For a more detailed discussion, see App. C] A single stainless steel fin on the stack, attached to a graphite interface (representing the flow fields for reactant and product water and gases, and therefore, the active area, where heat is generated,) is a unit that repeats through the stack, and is chosen as the unit to be modeled. Figure 1: XY View of Fin and Active Area of electrolyzer MEA below shows an XY view of this unit. The following screenshots of the thermal model below explain the different model screens, and provide an overview of the modeling and the resulting temperature distribution on the unit.

Figure 1: XY View of Fin and Active Area of electrolyzer MEA

Figure 1: XY View of Fin and Active Area of electrolyzer MEA shows an XY view of a single fin and its active area. The grid in blue represents plate elements 6the stainless

5 The active area consists of the cross sectional area of the stack, where thin graphite electrodes, are attached on both sides of the Solid Polymer Electrolyte (SPE) membrane. This is the area over which current flow and the reaction takes place, and as a result, the heat is generated. It is also the area over which the graphite flow fields are present for the product gases and the reactant (water). 6 Plate elements are two dimensional elements used to represent geometry where the temperature gradient through the thickness of the geometry being represented is not expected to be significant or important to the solution of the problem.

Draft, July 16, 2002 14

steel 316 fin, while the grid in green represents the surface heat load on the active area. The nodes covered by the green nodes are also the active area. The outermost cells, with lines emerging from the center, represent the exposed area of the fin. The lines emanating from the centers of these nodes, connected to one point each on all four sides of the fin, represent convection elements 7 to four boundary nodes 8at 25°C.

Figure 2:XZ View of Fin and Active Area below shows a YZ view of the fin and active area. The blue and green colors once again represent the stainless steel fin and the surface heat load. The short, 'open' vertical lines between the green surface heat nodes represent interface elements9, between the stainless steel fin and the graphite electrode, where the reaction occurs, and which is assumed to be the region where heat is generated.

Figure 2:XZ View of Fin and Active Area

Figure 3: YZ View of Fin and Active Area below shows a YZ view of the area.

Figure 3: YZ View of Fin and Active Area

Figure 4: YZ View, Close-up showing the interface, graphite electrode, and fin below shows a closer view of the stainless steel fin, interface and graphite electrode, represented by red plate elements.

7 Convection elements represent convection from physical surfaces of the model. Convection elements reference either a plate element or a single surface of a brick element or tetrahedron element which defines the heat transfer surface area. Each convection element also references a single node, which is used to define the fluid to which the surface convects. 8 Boundary Nodes are nodes whose temperature is defined by the user. These represent the boundaries of the system with which the system interacts thermally. 9 Interface elements are used to simulate mechanical interfaces between two materials. The heat transferred across the interface is defined by a heat transfer coefficient. The mechanical interface may be the thermal contact between two bolted blocks.

Draft, July 16, 2002 15

Figure 4: YZ View, Close-up showing the interface, graphite electrode, and fin

Parameters required for the thermal model: • Surface heat load: The total surface heat load is found to be 0.653W/cm2 using

the procedure given earlier. • Heat transfer coefficient 'h': The convection heat transfer coefficient is calculated

using the following correlation for forced convection between two parallel plates10

hc

p

h

h

DkNuh

kC

DVHD

⋅=

=

⋅⋅=

=

µµ

ρ

Pr

Re

*2

Where: Nu Nusselt number – dimensionless Re Reynolds number – dimensionless Pr Prandtl number – dimensionless Dh Hydraulic diameter– feet m Fluid Viscosity – lbm/(hour-feet) ρ Fluid density – lbm/ft3 k thermal conductivity – Btu/(hour-feet-F) hc Heat transfer Coefficient - Btu/(hour-feet2-F) For Laminar Flow (Re<2800)

3/2Pr)Re)/((016.01PrRe)/(03.0

54.7⋅⋅⋅+

⋅⋅⋅+=

LDLD

Nuh

h

For Turbulent Flow (Re>2800) For Pr > 0.5 33.08.0 PrRe023.0 ⋅⋅=Nu For Pr < 0.1 PrRe003.08.4 ⋅⋅+=Nu

The geometrical dimensions used above are: V = 46.0 in Calculated as shown in 27 H = 0.1969 in Distance between consecutive MEA's was measured

as 5 mm (0.1969 in) L = 10.0 in Length of fins along which flow takes place

• Physical Dimensions: These numbers are taken from the actual measured

dimensions of the stack. These are given in a labeled diagram in App B.b. 10 D.K. Edward, V.E. Denny, and A.F. Mills, Transfer Processes, An Introduction to

Diffusion, Convection, and Radiation, 2nd Ed. (McGraw Hill, 1979)

V

L

Pi

Draft, July 16, 2002 16

Result: Temperature Distribution The result of the simulation, shown below in Figure 5: Simulated Temperature Distribution, shows the temperature distribution of at steady state of a typical fin-electrode assembly. A maximum expected temperature of 103.97 °C is seen, which is higher than the optimum operating temperature of 80°C 11. This temperature might even be dangerous, being above the boiling point of water at standard temperature and pressure, since it might cause a decrease in hydration of the SPE, leading to drying decrease in the reaction rates, and local heating of the SPE membrane.

Figure 5: Simulated Temperature Distribution

In Figure 5: Simulated Temperature Distribution, we see that the temperature decreases symmetrically from the core outwards, becoming minimum at the corners, which is to be expected. We also see that while the temperature along the shorter edges approaches the ambient temperature (∼ 65°C), it is much higher along the longer edges (∼ 65°C). This appears to suggest a longer fin-length along the Y-direction on both sides for improved heat transfer. Qualification of temperature distribution obtained above: From experimental values at the given current (~20A), the actual temperature near the edge of the fin12, (∼ 65°C) were found to match the results shown above.

Assumptions:

11 Optimum operating temperature for SPE as suggested by ElectroChem, and in available literature 12 Fin Temperature (Tfin) as measured by Daniel Shapiro and Adarsh Das on October 23, 2001, when the system had been working for several hours, and the variation in Tfin over several hours had been a few °C.

Draft, July 16, 2002 17

• In the actual stack, in addition to convection through the exposed area of the fins, conduction to the two fin-electrode assemblies on either side also occurs. This is neglected in the above model.

• Heat loss due to radiation is neglected. • The construction details of the overall stack, the flow-field channels in the

graphite plates, and the temperature variations owing to the actual flow are neglected. The graphite plate flow-field is replaced with a solid graphite plate.

Draft, July 16, 2002 18

2. Heat Balance and Thermal Model for (proposed) 1600W electrolyzer The original 4kW fuel cell-electrolyzer remote power system mentioned earlier has a recharge-time of 40 hours. Thus, about (4000W⋅4 hours) of electrical energy needs to be produced. About twice this amount of energy in the form of the hydrogen fuel is required. Thus, an electrolyzer that needs to recharge the system over 40 hours needs to generate hydrogen at the rate of:

WhhWE ⋅=⋅= 400

4044000

About half of the electrical energy input to the electrolyzer is wasted as heat, and about ¼th is finally utilized for producing hydrogen. Taking these into account, the power rating of the electrolyzer needs to be1600 W. As a result, the 400 Wele electrolyzer is scaled up to 1600 Wele, starting with a current density ~ 1A/cm2. [For the complete calculations, see App A.b]

Thermal Model:

The steady state temperatures required for the system are assumed to be the same as those for the 400 W electrolyzer. The resulting mass flows and heat balances are calculated, and the following main parameters are used for developing the thermal model. • To calculate the dimensions of the new stack, the active area for an existing 4

kW fuel cell stack is scaled down by a factor of 0.533 (=(1.6kW/4kW)* A_Areainitial), taking the design current density (1 A/cm2) to remain the same. The linear dimensions (active area as well as fin dimensions) are therefore scaled down by a factor of √0.533=0.73

• The remaining dimensions of the stack (e.g. the thickness of each MEA and the resulting overall length of the stack) remain the same as the 400W electrolyzer.

J_curr_dens 1.02A

cm2:= Current Density, maximum

expected

A_active30.5

4in2:= Active Area for 4kW stack

A_active 49.193cm2=

I_ele J_curr_dens A_active⋅:= DS System Model Worksheet

I_ele 50.177A=

N_cell 16:=

V_cell 2V:=

V_ele V_cell N_cell⋅:= DS System Model Worksheet

V_ele 32V=

Power_ele I_ele V_ele⋅:= Power to electrolyzer

Power_ele 1.606 103× W=

Draft, July 16, 2002 19

• The total heat load on the active area is found to be 414W. A water flow rate of 2.5 liter /min removes 333W. For the remaining heat to be convected, the generation volume of 2.54 in3 per cel yields a volume heat load of 2.0125 W/cm3.

• The heat transfer coefficient to the air is found to change very little for the change in linear dimensions as above.

The main parameters required for the model are then used to generate the geometry and then completed with the material and thermal properties, as before, to generate a thermal model of the fin and its active area, shown below:

Figure 6:Thermal model for single fin-active area

Draft, July 16, 2002 20

Sixteen such fins are then combined, as shown below, to generate a model of the final stack, as shown below:

Figure 7: Temperature distribution on a 16-cell electrolyzer stack

In the above diagram, it's noticed that the maximum predicted temperature on the stack is 85.86°C. This is close to the optimum, and therefore the stack is expected to perform close to its optimum. It is also seen that temperature increases symmetrically as we proceed outwards from the core of the cell, becoming lowest at the extreme corners of the stack, as expected.

Draft, July 16, 2002 21

3. Thermal model for existing 4kW fuel cell stack For a fuel cell, the problem of heat transfer is compounded, because, unlike an electrolyzer, there is not enough flow of water to act as a natural coolant. A thermal model for a fin-active area is generated as shown below, after doing the necessary calculations for the main parameters, and using geometry from an existing 4kW fuel cell stack under test at ElectroChem. [For calculations see A.c]:

Figure 8: Fin and active area for 4kW stack

Figure 9: Temperature distribution for a fin-active area on a 4kW stack

From the above diagram, it is seen that the maximum temperature at the core is unacceptably high. Since there is not enough water flow occurring in a fuel cell to act as

Draft, July 16, 2002 22

a coolant for which the flow rate can be used as a control mechanism, the heat flow by convection as to account for nearly all the heat being generated, and also keep the maximum temperature to 80°C. The methods to increase the rate of heat rejection were a combination of: (a) To increase the conductivity of the fin, by using a different material (b) To increase the surface area available for cooling by convection, by increasing exposed fin-length. The results of the simulations are described below:

Figure 10: Temperature distribution with TWICE the convection area, SS316 fins

It is seen that the maximum temperature decreases to about 153°C.

Draft, July 16, 2002 23

Figure 11:Temperature distribution with SAME convection area, molded GRAPHITE fins

Figure 12:Temperature distribution with DOUBLE convection area, molded GRAPHITE fins

Draft, July 16, 2002 24

Figure 13: Temperature Distribution with THREE TIMES convection area, GRAPHITE fins

It's seen that the temperature distribution on the fin becomes close to optimum, with the maximum temperature being about 82.58°C. 5. Differential Pressure Relief Valve Design The differential pressure relief valve initial design was reconsidered when it was discovered that there exists no o-rings that can survive the 2000 psi pressure. Daniel Shapiro and Adarsh came up with a new approach, sketched below. The initial design is shown first, then the revised design.

Draft, July 16, 2002 25

` `2.50

1.001.00

0.25

7.00

2.481.612

1.616

0.433

0.433

Cylinder Length

Port

Dia

met

er

CylinderWall

Thickness

PistonLength

PistonDiameter

CylinderOuter

Diameter

CylinderBore

Diameter

1.38

PerforatedLength

1.574

0.175

T

Sid

RollerPilot

Draft, July 16, 2002 26

6. Conclusions • Models have been developed of the heat transfer in the fuel cell and

electrolyzer of the EC Cell system, and the results have been compared to available measurements of the small prototype and of the EC Cell prototype.

• The overheating of the 4 kW fuel cell can be solved by changing the material of the fins to graphite and increasing the fin area by a factor of three, according to the model.

• The potential overheating of the electrolyzer can be avoided by allowing sufficient flow of water through the electrolyzer.

• The initial design work, construction of the prototype, and testing of the system are described in detail in the thesis of Daniel Shapiro (2002).

• A revised differential pressure relief conceptual design was developed. • It appears that the major hurdles to the development of the EC Cell system

have been overcome. • Additional work will completed as part of Adarsh Das’s MS thesis.

Draft, July 16, 2002 27

5. Appendices:

A. Miscellaneous Calculations: a. Detailed Heat Balance: 400 W electrolyzer

"HT - balances - DS electrolyzer.mcd"

b. Detailed Heat Balance: 1600 W electrolyzer

"HT - balances - 1600W electrolyzer -

c. Detailed Heat Balance: 4000 W fuel cell

"HT - balances - 4kW stack.mcd"

d. Air Flow Calculations – for given fan

Ref: 0

Both the amount of heat to be dissipated and the density of the air must be know. The basic heat transfer equation is:

q= Cp x W x D T where q = amount of heat transferred Cp = specific heat of air D T = temperature rise within the cabinet W = mass flow

Mass flow is defined as:

W = CFM x Density

By incorporating conversion factors and specific heat and density for sea level are, the heat dissipation equation is arrived at:

CFM = 3160 x Kilowatts / D T x ° F

Draft, July 16, 2002 28

This yields a rough estimate of the airflow needed to dissipate a given amount of heat at sea level. It should be noted that the mass of air, not its volume, governs the amount of cooling.

.

Determining System Impedance

After the airflow has been determined, the amount of resistance to it must be found. This resistance to flow is referred to as system impedance and is expressed in static pressure as a function of flow in CFM. A typical system impedance curve, in most electronic equipment, follows what is called the "square law," which means that static pressure changes as a square function of changes in the CFM. Figure 1 describes typical impedance curves. For Most forced air cooling application, let n=2; approximating a turbulent system.

Static pressure through complex systems cannot be easily arrived at by calculation. In any system, measurement of the static pressure will provide the most accurate result. Comair Rotron makes this type of testing available. Please contact Application Engineering for more information

System Flow

Once the volume of air and the static pressure of the system to be cooled are known, it is possible to specify a fan. The governing principle in fan selection is that any given fan can only deliver one flow at one pressure in a given system.

The following MathCad worksheet shows the calculation of minimum airflow required:

kJ .1000 joule kilowatts .3.5 kW q kilowatts

Cp .1.0057 kJ.kg K

Specific heat of air

at atmospheric pressure delT .5 K Tout-Tin as

maintained at ElectroChem w q

( ).Cp delT

=w 0.696 kg sec 1

Draft, July 16, 2002 29

Figure 2 shows a typical fan pressure versus flow curve along with what is considered the normal operating range of the fan. The fan, in any given system, can only deliver as much air as the system will pass for a given pressure. Thus, before increasing the number of fans in a systems, or attempting to increase the air volume using a larger fan, the system should be analyzed for possible reduction in the overall resistance to airflow. Other considerations, such as available space and power, noise, reliability and operating environment should also be brought to bear on fan choice.

To demonstrate the impact of system resistance on fan performance, figure 3 shows three typical fans used in the computer industry. A is a 120 CFM fan, B is a 100 CFM fan and C is a 70CFM fan. Line D represents a system impedance within a given designed system. If 50 CFM of air are needed, fan A will meet the need. However, fan A is a high performance, higher noise fan that will likely draw more power and be more costly. If the system impedance could be improved to curve E, then fan B would meet the 50 CFM requirement, with a probable reduction in cost, noise and power draw. And if the system impedance could be optimized to where curve F were representative, then fan C would meet the airflow requirement, at a dramatically lower power, noise and cost level. This would be considered a well-designed system from a forced convection cooling viewpoint. Keeping in mind that a given fan can only deliver a single airflow at a given system impedance, the importance of system design on fan selection becomes obvious. Comair Rotron urges engineers to design fans into their systems, rather than add them as an afterthought, for best performance, noise, power and cost characteristics.

Series and Parallel Operation Combining fans in series of parallel can achieve the desired airflow without greatly increasing the system package size or fan diameter. Parallel operation is defined as having two or more fans blowing together side by side. The performance of two fans in parallel will result in doubling the volume flow, but only at free delivery. As figure 4 shows, when a system curve is overlaid on the parallel performance curves, the higher the system resistance, the less increase in flow results with parallel fan operation. Thus, this type of application should only be used when the fans can operate in a low impedance near free delivery.

Series operation can be defined as using multiple fans in a push-pull arrangement. By staging two fans in series, the static pressure capability at a give airflow can be increased, but again, not to double at every flow point, as Figure 5 displays. In series operation, the best results are achieved in systems with high impedance.

In both series and parallel operation, particularly with multiple fans (5, 6, 7, etc.) certain areas of the combined performance curve will be unstable and should be avoided. This instability is unpredictable and is a function of the fan and motor construction and the operating point. For multiple fan installations, Comair Rotron strongly recommends laboratory testing of the system.

Draft, July 16, 2002 30

e. Air Flow Calculations – required to reject 4000 W

"Air Flow required forHeat Transfer.doc"

B. Physical Dimensions (geometry) used in thermal models

a. Labeled schematic diagram, showing dimensions of 4kW stack The diagram below is a schematic, showing the approximate configuration. The actual number of fins is 68.

b. Labeled dimensions diagram for MEA, as modeled C. Thermal Analysis System:

a. TAS – An overview TAS is a general-purpose tool used to computer-simulate thermal problems. The program provides an integrated, graphical and interactive environment to the user. A single environment provides model generation, execution and post processing of the results. Colors provide feedback of element types and properties and temperature results. Models are generated using a set of elements. Full three-dimensional geometry can be created using two-dimensional plate and three-dimensional brick and tetrahedron elements. Convection, radiation and fluid flow elements are provided. Resistances can be added using resistor elements. Properties can be temperature, temperature

SS 316 Fin

SS Interface, 0.01

Graphite Flow Channel, 0.0625

Nafion Interface, 0.01

Membrane Electrode Assembly (model)

L = 1.4 cm

t = 0.1 cm

33 cm

D=15.3cm

t = 0.5 cm

-z

x

y

x

t = 1.4 cm

t = 30.8 cm

-z

y

Draft, July 16, 2002 31

difference, time and time cyclic dependent. Heat loads can be added on a nodal, surface or volumetric basis. Models generated can be subjected to various environments and thermal loads. The models can be used to determine the feasibility of a design or to determine problem areas. Geometry, thermal properties and parameters of the model can be easily changed to determine their effect. The design can be thermally optimized and characterized before incurring the expense of building and testing a prototype. TAS uses a finite difference solution to solve the model, with which complex models involving many of the nonlinear cases often encountered can be implemented. These include fluid flow, radiation, temperature dependent thermal conductivity and heat transfer coefficients which can be a function of temperature difference. Models represented by thermal resistors can be quickly solved on the PC. The accuracy of the software has been proven over the past years. The results of numerous models agree with classical solutions and results of other programs such as MSC/NASTRAN, ANSYS and SINDA.

b. Abbreviated procedure for creating models in TAS

i. Make geometry: 1. Start with the largest component, in this case, the fin. 2. For the fin, decide the number of subdivisions that is required. 3. Place the first node at the origin, and the next node at the

coordinates of the diagonal. 4. Add a plate with the two points already added, as the

diagonal, and adding the required material as a property. The thickness of the plate should be small.

5. Extrude this plate using brick elements, with the thickness of the bricks being equal to the thickness of the fin, making sure to delete the plate elements.

6. First grouping the required area, and then looking at it in side view, choose the brick faces to be projected. Once the fin is in place, add the interface elements at the appropriate coordinates, by projecting the brick faces. Be sure to add the heat transfer coefficient for the interface (in W/C-in2)

7. Next, create another group for the next layer, i.e. the graphite flow channels, and make it the current group. Project / extrude this layer in an identical manner, using bricks.

8. Using the same procedure, add the next later for the PE membrane.

ii. Add heat load

1. Decide whether the heat load will be a surface load, or a volume load.

2. Calculate the total heat load on the surface. 3. If a surface load, group the surface by making the relevant

group current, in side view, and selecting the required side.

Draft, July 16, 2002 32

4. If a volume load, group the volume by making the relevant group current, and selecting the required volume elements.

5. Calculate the total area or volume over which the heat is getting applied, and calculate the volume / surface heat flux.

6. Add the heat flux as a separate property. iii. Add convection:

1. Add the boundary nodes, (Hint: Use one for each side) 2. Add the boundary temperature as a property, and associate it

with the relevant nodes, turning them into boundary nodes. 3. Define the convection heat transfer property, using either a

known heat transfer coefficient, or calculating it inside TAS. 4. Select and group the required area, which convects to a given

node. 5. Add convection, after selecting the appropriate boundary

node, area, and convection property. iv. Solve steady state / transient

1. Compact and equivalence the elements at various stages, particularly when adding new geometry elements.

2. Before solving the model, 'Check Model' to see that there are no plates with zero thickness, and that the overall numbers for the heat load, temperatures, etc, look all right.

3. If solving for the first time, start with an initial temperature of 25 deg C, otherwise, check 'Use Current Temperatures'. Also, increase the maximum number of iterations to 50000 or more.

4. Finally, solve. v. Changing properties:

1. Define the NEW property, either by changing an existing one in the properties interface, or by creating a new property.

2. Select 'Edit Group' on the toolbar. 3. Define the new group with the properties that need to be

changed, selected. 4. Echo/Select the new group. 5. Select the 'Modify Property' Box, and change the property to

the new property. D. References:

Sysmodel.xls Fuel Cell Handbook (Fifth Edition), October 2000, from Office of Fuel Energy, National Energy Technology Laboratory (US Department of Energy) Hydrogen Storage, by Laura Becker, June 2001, from Cambridge Scientific Abstracts (www.csa.com) Fuel Cells - Green Power, from Los Alamos National Laboratory, (US Department of Energy) (www.lanl.gov)

Draft, July 16, 2002 33

Modeling of PV, electrolyzer and gas storage in a stand-alone solar-fuel cell system, MS thesis by Allison Arkin, 2001, University of Massachusetts, Lowell PEM Electrolyzer and gas storage for a regenerative fuel cell system, MS thesis by Daniel Shapiro, 2002, University of Massachusetts, Lowell Air Flow vs Pressure Characteristics from Comairtron (http://www.comairrotron.com) Arkin, A., and J.J. Duffy, 2001, “Modeling of PV, Electrolyzer and Gas Storage in a Stand-Alone Solar-Fuel Cell System,” Proceedings of the Annual National Solar Conference, American Solar Energy Society Bloom, D., and J.J. Duffy, 1988, "Loss-of-Load Probability for Stand-Alone Photovoltaic Systems with Random Loads," Annual Solar Conference Proceedings, American Solar Energy Society, MIT, June. Chamberlin C.E., Lehman P., Zoellick J., and Pauletto G. (1995) Effects of Mismatch Losses in Photovoltaic Arrays. Solar Energy 54, 165-171. Duffy, J.J., and S. Frye, 1989, "An Expert-System-Based PV Sizing Method for Minimum Cost," Proceedings of the 1989 Annual Conference of the American Solar Energy Society, Denver. Duffy, J.J., 1999, "Peruvian Villages Go Solar," Solar Today, Nov/Dec, American Solar Energy Society, Boulder, CO, p. 30-34 Flavin, C., and M. O'Meara, 1997, "Financing Solar Electricity," World Watch, Vol. 10, No. 3. Garcia-Conde A.G. and Rosa F. (1993) Solar Hydrogen Production: A Spanish Experience. Int. J. Hydrogen Energy 18, 995-1000. Hande, H., and J.J. Duffy, 2001, "A Model for Sustainable Rural Solar Electrification with Photovoltaics," Proceedings of the Annual National Solar Conference, American Solar Energy Society. Ledjeff, K., A. Heinzel, V. Peinecke, and F. Mahlendorf, 1994, "Development of Pressure Electrolyser and Fuel Cell with Polymer Electrolyte," Int. J. Hydrogen Energy, Vol. 19, No. 5, p. 453-455. Morimoto, Y., T. Hayashi, and Y. Maeda, 1986, "Mobile Solar Energy Hydrogen Generating System," Proceedings of the 6th World Hydrogen Energy Conference, International Association for Hydrogen Energy, Pergamon Press, New York, p. 326-332. Raudales, R., D. Villanueva, C. Munger, and J.J. Duffy, 1998, "Solar Coffee Dryers," Annual National Solar Conference Proceedings, American Solar Energy Society.

Draft, July 16, 2002 34

Reinhardt, M., and J.J. Duffy, 2001, “Optimization Methods for Alternative Energy System Design,” Proceedings of the Annual National Solar Conference, American Solar Energy Society US Dept. of Energy, 1999, The Fuel Cell Handbook, Fourth Ed., Office of Fossil Energy, Washington, DC. Wijesooriya, P., and J.J. Duffy, 1992, "Stand-Alone PV System Sizing Based on Both Solar and Reliability Uncertainties," National Solar Conference Proceedings, Amer. Solar Energy Soc., June. Wolk, R., 1999, "Fuel Cells for Homes and Hospitals," IEEE Spectrum, May, IEEE, New York, p. 45-52.