international journal of hydrogen energy 33 (2008) 5931–5940

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/he

Optimization of solar powered hydrogen production

using photovoltaic electrolysis devices

Thomas L. Gibson*, Nelson A. Kelly

General Motors R&D Center, Chemical and Environmental Sciences Laboratory, 30500 Mound Road, Warren, MI 48090-9055, USA

article info abstract

Article history: Hydrogen fuel for fuel cell vehicles can be produced by using solar electric energy from
Received 26 March 2007 photovoltaic (PV) modules for the electrolysis of water without emitting carbon dioxide or
Received in revised form requiring fossil fuels. In the past, this renewable means of hydrogen production has suffered
2 May 2008 from low efficiency (2–6%), which increased the area of the PV array required and therefore,
Accepted 25 May 2008 the cost of generating hydrogen. In this research, the efficiency of the PV-electrolysis system
Available online 26 September 2008 was optimized by matching the voltage and maximum power output of the photovoltaics to
the operating voltage of proton exchange membrane (PEM) electrolyzers. The optimization
Keywords: process increased the hydrogen generation efficiency to 12% for a solar powered PV-PEM
Photovoltaic electrolysis device electrolyzer that could supply enough hydrogen to operate a fuel cell vehicle.
Electrolyzer ª 2008 GM Global Technology Inc. Published by Elsevier Ltd on behalf of International
Hydrogen production Association for Hydrogen Energy. All rights reserved.

1. Introduction inverter to produce AC current which was then fed to the

power grid. Power was taken from the grid through an inde-
A number of solar hydrogen projects have been built to produce pendent circuit and converted to DC current to operate an
hydrogen gas to supply vehicle-fueling stations by using elec- electrolyzer and produce hydrogen (Canadian National Expo-
tricity from photovoltaic panels and commercially available sition, Toronto, and Detroit Edison, Southfield, Michigan).
electrolyzers to split water. In the past, these projects have While these grid-tied systems theoretically have the advan-
proven unsatisfactory due to the low efficiency and high cost of tage of supplying any desired amount of power from the utility
the technology, which only converted 2–6% of the solar energy to meet their needs, the additional cost of the inverters, power
to hydrogen fuel energy. The resulting fuel cost was at least conditioning electronics, connection costs, utility fees, and the
$40/kg of hydrogen, and the land area covered by the photo- losses each time power is converted between DC and AC make
voltaic panels was at least 130 m2 per vehicle based on the them less efficient and more costly than stand-alone systems.
system efficiency and estimated cost of photovoltaic systems Stand-alone PV-electrolysis systems for solar hydrogen
alone [1–5]. In this research, we sought to investigate and fuel production have consisted of an array of photovoltaic
model the factors that affect the efficiency of solar hydrogen modules that supplies DC electricity through storage and
generation and to design a more efficient and cost effective power conditioning systems to an electrolyzer as shown in
solar powered PV-electrolysis system based on our results. major projects in California, Germany, and Saudi Arabia
[1,2,6–12]. In these pilot projects (Fig. 1), PV modules are con-
1.1. Conventional PV-electrolysis systems nected to a charge controller, storage batteries, and a DC–DC
power converter to supply DC current to an electrolyzer con-
Solar hydrogen projects have often consisted of power grid- sisting of a series of catalyzed electrodes to split water and
tied systems, in which PV modules were connected to an produce hydrogen fuel and oxygen.

* Corresponding author. Tel.: þ1 586 986 1615; fax: þ1 586 986 1910.
E-mail address: thomas.l.gibson@gm.com (T.L. Gibson).
0360-3199/$ – see front matter ª 2008 GM Global Technology Inc. Published by Elsevier Ltd on behalf of International Association for Hydrogen Energy. All rights reserved.
doi:10.1016/j.ijhydene.2008.05.106
5932 international journal of hydrogen energy 33 (2008) 5931–5940

PV modules of a conventional, non-optimal system (such as resistances in the system. The PV module was then connected
Fig. 1) are not designed to supply the most efficient voltage for directly to the electrolysis system to maximize the efficiency
operating the electrolyzer. PV panels have not previously been of the system by designing the solar module to give its
designed to efficiently generate hydrogen when connected maximum power at a voltage that would match the fixed
directly to electrolyzer units which require a specific operating voltage required to operate the complete electrolysis system
voltage to split water. Instead, the PV modules usually consist (electrolyzer). Studies were conducted to design each part of
of a number of crystalline silicon (c-Si) solar cells (often 36 cells) the PV-electrolysis system including the solar modules and
connected in series to produce electricity with a potential of electrolysis system (electrodes and the electrolyte) to optimize
about 18 V appropriate for charging 12-volt lead-acid storage the efficiency of the overall system.
batteries when operating at their maximum power point One way to simplify the circuit was to eliminate the power
(Vmpp), where they deliver their highest power under load. If converters (inverters and DC–DC converters) that add both to
there is a mismatch between the Vmpp and the operating circuit resistance and system cost. Since each of these solid
voltage (Voper) of the electrolyzer, the efficiency of the PV state power converters is about 90–95% efficient and costs
system decreases sharply. The PV power circuit (or battery approximately $100/kW, removing them from the design
system) has sometimes been connected to a DC–DC converter significantly improved the efficiency and cost. Instead of
(Fig. 1) to change the PV voltage to the characteristic operating relying on a DC–DC converter (as in Fig. 1) to raise or lower the
voltage of the electrolyzer [1], but the additional resistance PV output voltage to the voltage level needed to operate the
imposed by the voltage converters and batteries in these electrolyzer, the PV system was designed with the correct
systems also leads to significantly reduced efficiency. Small number of solar cells in series to generate the optimum elec-
photoelectrochemical (PEC) devices, consisting of electrodes trolysis voltage. This system design constitutes direct
with integrated PV and electrolytic functions, have been tested connection PV-electrolysis. An optimized PV-electrolysis
in the laboratory, but they are less practical than PV-electro- system was designed and built using PV modules directly
lyzer systems for collection and storage of hydrogen for vehicle connected to a multicell electrolyzer system (Fig. 2). Tests of
fueling, because they operate at approximately one atmo- systems incorporating a variety of PV modules were used to
sphere pressure with a low energy density [13]. optimize the system efficiency.

1.2. Strategies for optimizing PV-electrolysis

2. Experimental
Major improvements in solar powered electrolysis are needed
2.1. Current and voltage instruments
to increase the efficiency and reduce the cost of the system
and make it more practical for hydrogen generation. Since PV
Fluke digital multimeters from the 175 series calibrated by the
electricity generation and electrolysis are coupled to produce
manufacturer with standards traceable to the United States
solar hydrogen, the electrical efficiency of the PV system must
National Institute of Standards and Technology (NIST) using
be multiplied by the efficiency of the electrolysis system for
ISO-9001 controlled procedures were used for the current and
converting electricity to hydrogen fuel energy to find the
voltage measurements needed to calculate electrolysis effi-
overall efficiency of solar hydrogen generation (Eq. (1)). Both
ciency. The accuracy of measurements with these instru-
efficiencies need to be optimized for maximum system effi-
ments over the ranges used in the study was expected to be no
ciency.
more than 0.35% for potentials (DC volts) and 1.3% for
Solar H2 efficiency ¼ PV efficiency  electrolysis efficiency current (DC amps).
(1)

Two strategies for improvements were chosen for testing in 2.2. Hydrogen flow rate measurements
this study. Simplification of the PV-electrolysis system by
removing the charge controllers, storage batteries, and DC–DC Hydrogen flow rates were measured with a calibrated
converters was done to reduce cost as well as electrical mechanical (piston type) instrument: the Bios DryCal primary

H2

O2
Electrolyzer
Photovoltaic Charge
Modules Controller
DC/DC
Storage
Converter
Batteries

Fig. 1 – A conventional photovoltaic-electrolyzer hydrogen generation system made from a non-optimized photovoltaic
power supply, electrolyzer, and other hardware.
 international journal of hydrogen energy 33 (2008) 5931–5940 5933

Multicell PEM
Electrolyzer

H2

O2

Optimized solar modules

~ 33 volts each

Fig. 2 – Optimized design for a photovoltaic electrolysis system using solar cells with the optimum electrolysis potential
connected in parallel to a proton exchange membrane (PEM) multicell electrolyzer (a device with 20 electrolysis cells in
series is shown).

flow meter with digital readout, model DCL-MH for rates up to 8-48). Comparison of the UDT and Eppley sensors showed
20 L/min with an accuracy of 1% (Bios International Corp., that there was a linear relationship between the two
Butler, NJ). Oxygen was measured with the Bios Drycal model instruments on both sunny days, with direct beam
DCL-L for flows up to 700 mL/min with an accuracy of 1%. sunshine, and cloudy days when virtually all of the
sunshine was from diffuse (scattered) solar radiation. The
Eppley pyranometer is widely used to measure global solar
2.3. Solar irradiance measurements
irradiance [15]. It responds to radiation over a hemisphere,
has built-in temperature compensation, and outputs a mV
All solar intensity levels were determined using a calibrated
signal. According to the manufacturer, it is linear to within
United Detector Technology (UDT) crystalline silicon photo-
1% from 0 to 1400 W/m2 and has a cosine response to
diode sensor, model 10 DP/SV, by measuring the DC current
within 2% for incident radiation with zenith angles from
output of the sensor with a Fluke Model 179 multimeter. The
0 to 70 .
UDT sensor was calibrated at the National Renewable
Energy Laboratory, Golden, CO using their Spectrolab X-25
solar simulator, a filtered xenon arc that produces a stan- 2.4. Solar hydrogen efficiency measurements
dard air mass 1.5 global (AM 1.5 G) [14] spectrum with an
irradiance of 1000 W/m2, commonly referred to as one sun, 2.4.1. Calculated from operating current
the maximum solar energy on a clear cloud free summer Experimentally, the efficiency of hydrogen generation from
day at mid-latitudes in the northern hemisphere. On sunny PV-electrolysis is determined by connecting the PV power
days, the sensor was pointed directly toward the sun while source, irradiated with sunlight at a known intensity (usually
the current was monitored to measure the solar irradiance. one sun), to a high efficiency electrolysis cell or electrolyzer
A current of 21.0 mA corresponded to an irradiance of (where N is the number of electrolysis cells in series) and
1000 W/m2 with an accuracy of approximately 2% based on measuring the electrolysis current Ioper. Then, the solar
the accuracy of the Fluke multimeter. hydrogen production efficiency is calculated using Eq. (2).

N  Ioper ðmAÞ  1:23 ðVÞ

Solar to H2 Efficiency ¼ 100% (2)
PV area ðm2 Þ  Solar irradiance ðW=m2 Þ

The current output by the UDT sensor was a linear 2.4.2. Calculated from hydrogen flow
function of the solar irradiance. This was determined by The solar hydrogen production efficiency can also be calculated
comparing the UDT sensor over a range of sunlight condi- from the hydrogen flow rate measured with the digital flow meter
tions with an Eppley Black and White pyranometer (model (Eq. (3)).
5934 international journal of hydrogen energy 33 (2008) 5931–5940

H2 flow  H2 LHV  H2 density

Solar to H2 Efficiency ¼ 100% (3)
PV area ðm2 Þ  Solar irradiance ðW=m2 Þ

In Eq. (2), hydrogen flow is in L/h at 1 atm and 298 K, modules which contain a c-Si layer sandwiched between
hydrogen LHV (lower heating value) is equal to 33.35 kWh/kg, two thin layers of a-Si.
the hydrogen density factor is 0.002 kg/24.45 L at 298 K and Table 1 shows the solar modules that we tested, and the
1 atm, and solar irradiance is in W/m2. manufacturer stated area, efficiency, Voc, Isc, Vmmp, and Immp,
and Pmax. This set of solar modules provides a range of input
voltages (Vmmp ¼ 17–54 V) which we used individually and in
2.5. Photovoltaic materials combinations to generate a series of voltages (17–105 V) that
were used to measure the response of the electrolyzer
2.5.1. PV-PEM electrolyzer system hydrogen output to changes in the PV voltage driving elec-
The relationships of voltage, current, and power from typical trolysis. Modules 3 and 12 used monocrystalline cells.
solar cells used in this study are shown in Fig. 3. An important Modules 8, 9, and 16 used multicrystalline silicon. Modules 10
concept in the efficient use of solar cells and PV modules made and 11 used layered amorphous silicon and crystalline silicon.
up of solar cells in series and parallel arrangements is the Module 13 also used single-crystal silicon cells. The cells had
maximum power point (mpp). When a PV device is connected solar to electricity efficiencies ranging from 11.5% to 17.5%,
to any load including an electrolysis unit, motor, etc., the while the efficiencies of the modules ranged from 10.6% to
voltage drops below the open circuit potential (Voc) of the PV 15.2%. Cell area and efficiency refer to the actual silicon PV
system while the current and power increase. The mpp is the material that can generate electrical current. Module area and
point on the current–voltage curve and on the power curve at module efficiency refer to the entire surface of the PV module
which the photovoltaic material produces its maximum including ‘‘dead areas’’ between and around the solar cells as
power when operating under load [16]. The mpp is found by far as the outer edge of the module. The module efficiency is
plotting the current (I ) versus potential of the PV device while always lower than the cell efficiency because it includes this
it is irradiated by sunlight of a known intensity [17]. These inactive area in addition to the active surface of the actual
measurements are usually carried out under standard test solar cells.
conditions (STC) using 1000 W/m2 (one sun) with an AM 1.5 G Voper is a characteristic voltage at which the electrolyzer
spectrum. operates. It is influenced by the electrode and membrane
A variety of PV modules were used to optimize solar materials, their catalyst coatings, and the electrolyte (in a PEM
hydrogen production from a PV-PEM electrolyzer system electrolyzer the membrane between the electrodes acts as the
by conducting electrolysis tests over a range of voltages electrolyte). Voper of the electrolyzer is the sum of the standard
(Table 1). The modules were manufactured by Sharp Elec- water splitting voltage plus the overvoltage of the electrolyzer
tronics Corp., Huntington Beach, CA; Shell Solar, Camarillo, [18,19].
CA; and Sanyo Electric Co. Ltd., Osaka, Japan. All of the PV
Voper ¼ N  ð1:23 ðVÞ=cell þ overvoltage=cellÞ (4)
modules were constructed with c-Si except the Sanyo
The lowest reported overvoltage is approximately 0.4 V,
meaning that the most efficient electrolysis cells have Voper of
7 200 1.65 V/cell [18]. For a multicell, series electrolyzer stack, water
Power 180 is split between each pair of electrodes, and the Voper is
6
MPP 160 multiplied by N, the number of electrolysis cells in series. For
Current
5 example, Voper of a 20-cell electrolyzer stack would be about
Current (Amps)

140
Power (Watts)

120
33 V if optimum electrode materials and catalyst coatings are
4 used.
100
3 80
H2
2 60
40 Deionized
1 Water
VMPP
20
VOC
0 0 Pump Gas/Liquid
0 5 10 15 20 25 30 35 40 45 50 Separator
Voltage
O2
Fig. 3 – Typical current versus potential (voltage) curve for
a photovoltaic module (Table 1, no. 13) showing the PEM Electrolyzer

maximum power point (mpp) under standard test

Optimum PV System
conditions. The corresponding power curve
(P [ voltage 3 current) is also shown. Fig. 4 – Simplified schematic of a PV-electrolyzer system
used to test direct connection optimization (DCO).
 international journal of hydrogen energy 33 (2008) 5931–5940 5935

H2 2.6.2. Operation of the PEM electrolyzer and gas separation

system
Pump The PEM electrolyzer stack in the HOGEN 40 series hydrogen
Deionized
Water generator has two electrical connections and four gas
connections (two for hydrogen and two for oxygen). At high
Gas/Liquid current inputs (above about 12 A), the electrolyzer also needs
Separator to have cooling water circulated to the electrolyzer to prevent
O2
the stack from overheating. We found that pumps could be
used to supply a constant circulation of water to the hydrogen
PEM Electrolyzer
DC-DC and oxygen electrodes and to remove hydrogen and oxygen
Converter
from the stack. A pumped circulation system and gas/liquid
Non-Optimum PV System separator was connected to both the hydrogen and oxygen
Fig. 5 – Simplified schematic of a PV-electrolyzer system sides of the stack (Fig. 5) although we found that efficiency
with DC–DC converter (DDC) optimization. was not increased by the operation of the pumps (only one of
the pumps is shown in the figure). The overvoltage in the 20-
cell PEM electrolyzer used in our optimization tests was
0.40 V/cell so that the Voper was approximately 33 V [from Eq.
2.6. PV-proton exchange membrane (PEM) electrolyzer (4): 20  (1.23 þ 0.40) ¼ 33].
system
2.6.3. Test equipment for electrolyzer experiments
Two types of PV powered PEM electrolyzer systems are shown The test unit for optimizing PV-electrolyzer systems (Fig. 6)
schematically in Figs. 4 and 5. In the first system, the PV consisted of the 20-cell electrolyzer stack mounted on a cart. It
modules are directly connected to the electrolyzer to generate was connected by stainless steel and Teflon tubing to gas–
hydrogen, and in the second system, the PV output is routed liquid separators for hydrogen and oxygen that also served to
through a DC–DC converter to modify the voltage and current supply deionized water to the electrolyzer, keeping the water
input to the electrolyzer. level covering the membrane electrode assemblies (MEA).
Each gas exited the electrolyzer stack from a separate port and
2.6.1. PEM electrolyzer stacks was then bubbled through water in a gas separator bottle
The proton exchange membrane (PEM) electrolyzer stack used before exiting from the top of the separator through a collec-
in the experiments was made by Proton Energy Systems, Inc., tion port. After the gases passed through filters to trap water
Wallingford, CT. The stack (serial number XEM0134701) con- droplets, the flow rate of each gas was measured using the
sisted of 20 MEA electrolysis cells in series with a total height Bios flowmeter. The flow rates required a correction coeffi-
of 10.3 cm, with each MEA consisting of two metallic electrode cient of 2% (at 20  C) for the volume of water vapor when
surfaces separated by a Nafion type membrane to act as a solid measuring the moist gases.
electrolyte. The active area of each MEA was about 100 cm2.
The proprietary catalysts coated on the MEA plates and 2.6.4. DC–DC converter devices (linear current boosters and
membrane are believed to contain platinum on carbon at the charge controllers)
cathode (hydrogen producing electrode) and ruthenium oxide A DC to DC converter (DDC) is a device that can change an
or iridium oxide at the anode. This PEM stack was designed for input voltage and current, V1 and I1 to a different output
use in the Proton Energy Systems, Inc. HOGEN 40 electro- voltage and current, V2 and I2. Its usefulness is in optimizing
lyzer system but has been replaced more recently by a larger the output voltage of a PV system by matching it to the input
stack with 25 cells in series. Test stand operations were voltage required by the load. As noted earlier, solar panels
monitored by a laboratory safety system (hydrogen sensors have a maximum power point, Pmpp, where the product V1  I1
and LEL alarm) while the cabinet of the intact HOGEN 40 was is maximized. This may not correspond to the voltage and
equipped with similar sensors and safety devices. current draw of the load (in our case, an electrolyzer). When

Table 1 – Characteristics of photovoltaic modules used in PV-PEM electrolyzer experiments

Module Module Model Voc (V) Isc (A) Vmmp (V) Immp (A) Rated Module Module Cell Cell
number manufacturer number power (W) effic. (%) area (m2) effic. (%) area (m2)

3 Shell SQ-75 21.7 4.8 17.0 4.4 75 11.9 0.632 13.3 0.566
8, 9 Sharp ND-NOECU 24.9 7.8 20.0 7.0 140 12.1 1.153 13.4 1.042
10 Sanyo HIP-J54BA2 66.4 3.7 54.0 3.3 180 15.2 1.181 17.3 1.040
11 Sanyo HIP-G751BA2 64.9 3.6 51.4 3.3 167 14.1 1.181 16.1 1.040
12 Shell SP-140-PC 42.8 4.7 33.0 4.3 140 10.6 1.320 11.5 1.22
13 Sharp NT-185U1 44.9 5.6 36.2 5.1 185 14.2 1.301 17.5 1.057
14, 15 Sharp ND-070ELU 12.4 7.8 10.0 7.0 70 12.1 0.577 13.4 0.521
16 Sharp NE-165U1 43.1 5.5 34.6 4.8 165 12.7 1.301 14.6 1.134
5936 international journal of hydrogen energy 33 (2008) 5931–5940

voltage and current. DC–DC converters, however, stall and

cannot send the PV current to a charged system when irra-
diance is reduced by low light conditions.
A third DC–DC converter, the model PPT 48-20 GM charge
controller (input voltage 65–90 VDC) from Solar Converters,
Inc., Toronto, ON, was used in the same way as the PPT 48-10
DC–DC converter to modify PV voltage to reach a higher PV-
electrolyzer efficiency. The charge controller, however, has
the ability to operate when connected to a charged battery or
PEM electrolyzer without stalling due to additional diodes in
the circuit. This ability to control voltage and current flow to
a charged system is the chief advantage of the charge
controller compared to the DC–DC converters described
above.
A small amount of current is lost due to resistance in
the DC–DC converters, which have an efficiency of 95% (at
low current) to 90% (at high current flow), but this loss may
Fig. 6 – The 20-cell PEM electrolyzer and water circulating
be offset by the improvement in efficiency if the DDC is
pumps installed on the lower level of the test system.
necessary to match the PV voltage to the electrolyzer
operating voltage. This gain is especially important when
the PV system has a low intrinsic Vmpp, which would
the PV outputs and load are mismatched, efficiency is prevent the start of electrolysis since the 20-cell PEM elec-
reduced. Furthermore, as the intensity of sunlight changes, trolyzer we tested requires at least 31 V to begin electro-
the current output of a solar module will change, and to lyzing water.
a lesser degree, so will the voltage output. A DC–DC converter When the PV-electrolyzer system is operated in the DDC
can be designed to track the maximum power point of the optimization mode, pairs of PV modules can best be con-
solar panels to optimize the efficiency of the PV-electrolyzer nected in series to multiply the voltage of the PV system
system. (for example to Vmpp ¼ 72 or 108 V), which will decrease so-
Two DC–DC converters (linear current boosters) made to called ‘‘copper losses’’ of DC current caused by resistance in
our specifications were used to improve the efficiency of the wiring. The higher voltage is then converted to the
mismatched PV-electrolyzer combinations: (1) a model for operating voltage for the electrolyzer when passing through
low-voltage PV modules that gave a higher output voltage (but the DDC.
lower current), to match the operating voltage of the electro-
lyzer, and (2) a model for high-voltage PV modules that
decreased the output voltage to match the electrolyzer oper-
ating voltage and give an increased current. Both converters 3. Results and discussion
were provided by Solar Converters Inc. (Guelph, Ontario) and
consisted of semiconductor electronics with primary and The efficiency of the electrolysis system or electrolyzer is
secondary coils wound to give an output voltage of 36 V. The given by Eq. (5) [19].
low-voltage model was a modified Model CV 20/33-20 special N  1:23 ðVÞ
ordered with a potentiometer to manually adjust the output Electrolysis Efficiency ¼  100% (5)
Voper
voltage, and the high-voltage model (PPT 48-10) was modified
to have potentiometers for manually adjusting both output where N is the number of electrolysis cells in series and Voper
is the operating voltage drop across the electrolysis system.

14
3.1. Hydrogen generation by PV powered electrolysis
12
Solar to Hydrogen
Efficiency (%)

10 The rate of water splitting by electrolysis using the optimized

8 anode and cathode materials was first measured as a function
of the potential applied to the electrodes (operating potential,
6
Voper) using adjustable direct current sources. It was found
4
that the current in the electrolysis experiments, which is
2 proportional to the rate of electrolysis, the production of
0 hydrogen gas, and hydrogen fuel energy, increased with
0 20 40 60 80 100 120 increasing operating potential according to Faraday’s rela-
Vmpp of PV Systems tionship (Eq. (6)) where F is Faraday’s constant (w96,500
Fig. 7 – Average solar hydrogen production efficiencies of coulomb/equivalent weight).
photovoltaic systems with a range of Vmpp directly equivalent weight  Ioper  time
Mass of H2 ¼ (6)
connected to a PEM electrolyzer. F
 international journal of hydrogen energy 33 (2008) 5931–5940 5937

3.2. Optimization of a PV-PEM electrolyzer methods was compared. A series of tests described below
were conducted to optimize electrolysis driven by solar PV
Background: We first optimized PV-electrolysis systems on panels with different PV output voltages and current levels
a small scale in the laboratory using a single alkaline elec- operating a PEM electrolyzer with and without the use of DC–
trolysis cell connected to various numbers of solar cells con- DC converters. It should be noted that such PEM electrolyzers
nected in series to the electrodes in KOH solution and have the potential advantage of compressing the hydrogen
irradiated with a calibrated light source. After establishing without the need for an external mechanical compressor.
that the efficiency reached a peak when the electrolysis
voltage matched the maximum power point of the combined 3.3. Direct connection optimization
solar cells, we looked for a means of scaling up this experi-
ment to study the phenomenon more exactly. To scale up The results of direct connection optimization of a 20-cell PEM
solar hydrogen production, we chose to use a multicell proton electrolyzer stack are shown in Table 2. The overall efficiency
exchange membrane (PEM) electrolyzer, because it was more of the PV-electrolyzer system was determined in this study by
efficient than an alkaline electrolysis tank and provides measuring the solar irradiance and the area of the PV solar
a compact system with higher energy density and uses cells to obtain the input energy and by measuring the current
neutral-pH deionized water, eliminating the need for flowing through the electrolyzer using a low resistance
hazardous KOH. In addition, future advances in efficiency and ammeter in the circuit. The current is then multiplied by the
cost reduction for PEM fuel cells may result in similar standard voltage for water electrolysis (1.23 V) to determine
improvements to PEM electrolyzers. Both direct connection the energy of the hydrogen generated (Eq. (2)). The hydrogen
optimization (DCO), in which the PV system is wired directly energy production was also calculated from the hydrogen
in series to the electrolyzer system and DC–DC converter volume measured with a calibrated flow meter as a check on
optimization (DDO) in which a DC–DC converter is wired in the results (Eq. (3)). The efficiency calculated from the current
the circuit were used and the effectiveness of the two was the same as that calculated from the hydrogen produced.

Table 2 – Hydrogen production efficiency results in PV-PEM electrolyzer optimization tests

PV module Vmpp (V) Irradiance Isc (A) Voper Ioper (A) H2 flow PV area % System efficiency % System efficiency
number(s) (W/m2) (VDC) rate (L/min) (m2) (from Ioper) (from H2)

3 17.0 970 4.99 17.0 0 0 0.57 0 0

8 20.0 950 7.3 23.6 0 0 1.04 0 0
8 þ 9 Parallel 20.0 950 15.3 23.6 0 0 2.1 0 0
8 þ 15 Series 30.0 770 5.1 31.2 3.3 0.51 1.56 6.8 6.9
8 þ 15 Series 30.0 825 5.1 31.3 3.5 0.55 1.56 6.8 7.0
12 33.0 920 4.4 31.5 3.8 0.56 1.22 8.3 8.1
12 33.0 971 4.5 31.6 3.9 0.62 1.22 8.3 8.8
16 34.6 970 5.5 32.0 4.5 0.70 1.13 10.1 10.5
16 34.6 970 5.5 32.0 4.6 0.72 1.13 10.3 10.7
16 34.6 970 5.5 32.1 4.6 0.72 1.13 10.3 10.7
13 36.2 970 5.6 32.1 5.2 0.80 1.06 12.4 12.8
13 36.2 970 5.6 32.2 5.1 0.80 1.06 12.3 12.8
13 36.2 970 5.6 31.7 4.7 0.81 1.06 12.4 12.9
8 þ 9 Series 40.0 905 7.2 32.1 6.3 1.02 2.08 8.3 8.9
8 þ 9 Series 40.0 971 7.7 32.5 7.1 1.12 2.08 8.7 9.0
13 þ 15 Series 46.2 820 6.5 31.4 4.4 0.69 1.58 8.4 8.7
13 þ 15 Series 46.2 870 6.5 31.6 4.9 0.76 1.58 8.8 9.1
11 51.4 880 3.4 31.3 3.3 0.53 1.04 8.9 9.4
11 51.4 950 3.6 31.5 3.7 0.59 1.04 9.2 9.7
10 þ 11 Parallel 52.7 905 7.0 32.2 6.6 1.05 2.08 8.7 9.1
10 þ 11 Parallel 52.7 925 7.2 32.2 7.0 1.11 2.08 9.0 9.4
10 54.0 880 3.2 31.3 3.2 0.49 1.04 8.5 8.8
10 54.0 940 3.4 31.5 3.4 0.53 1.04 8.6 8.9
10 54.0 945 3.4 31.5 3.6 0.53 1.04 9.0 8.9
12 þ 16 Series 67.6 850 4.3 31.3 3.6 0.59 2.35 4.5 4.8
12 þ 16 Series 67.6 860 4.3 31.3 3.8 0.59 2.35 4.6 4.8
13 þ 16 Series 70.8 860 4.8 31.4 4.4 0.71 2.19 5.8 6.1
13 þ 16 Series 70.8 880 4.8 31.5 4.5 0.74 2.19 5.7 6.2
10 þ 12 Series 87.0 870 3.8 31.2 3.3 0.53 2.26 4.1 4.4
10 þ 12 Series 87.0 870 3.8 31.2 3.3 0.53 2.26 4.1 4.5
10 þ 13 Series 90.2 890 4.9 31.2 3.9 0.61 2.10 5.2 5.4
10 þ 13 Series 90.2 860 4.9 31.3 3.8 0.62 2.10 5.2 5.6
10 þ 11 Series 105.4 925 3.1 31.5 3.5 0.55 2.08 4.4 4.7
10 þ 11 Series 105.4 870 3.1 31.4 3.1 0.54 2.08 4.2 4.9
5938 international journal of hydrogen energy 33 (2008) 5931–5940

20 (32–33 V). Thus, the highest efficiency (12.4%) was achieved

18 when the Vmpp range of the PV module (32–38 V) matched or
PV Electrical Efficiency (%)

16 overlapped the Voper of the PEM electrolyzer system (32–33 V),

as expected. The PEM electrolyzer stack had an operating
14 17% Efficiency Range
(32 - 38 Volts) temperature of 25  C and current density of 30–70 mA/cm2
12
during the optimization tests. The Voper of 32–33 V corre-
10
sponds to an electrolyzer efficiency of 75–77% (see Eq. (5)).
8 When a larger, scaled-up version of the PV-electrolyzer
6 system is built and operated in the direct connection mode,
4 the PV modules in the expanded PV system can be connected
2 in parallel to maintain the optimum Vmpp to operate the
0 electrolyzer most efficiently and supply enough current to
0 5 10 15 20 25 30 35 40 45 50 produce the hydrogen fuel requirements of one fuel cell
Operating Voltage (VDC) vehicle (for example 30 PV modules with Vmpp ¼ 36 V could be
used for the 20-cell PEM electrolyzer we tested to operate it
Fig. 8 – PV electrical efficiency of module #13.
with 100 amp and produce ~0.5 kg H2 per day). This number of
PV modules would be enough to both generate and compress
the hydrogen for practical onboard use.
Using direct connection optimization (DCO), the PV
Two other PV modules, #12 and #16, also had their Vmpp
modules to be tested were connected directly to the PEM
(33 V and 34.6 V, respectively) and their maximum efficiencies
electrolyzer stack to produce hydrogen by splitting deionized
near the Voper of the PEM electrolyzer. Thus, they could also
water (Fig. 7). An efficiency of up to 12.4% for the conversion of
operate near their highest efficiency when powering the
solar energy to hydrogen fuel energy was achieved by
electrolyzer. However, these modules had lower inherent
choosing PV modules that closely matched the characteristic
electrical efficiencies (11.5% and 14.6%, respectively) than
operating voltage (Voper) requirements of the 20-cell PEM
module #13 (which had 17.5%) and so gave lower solar
electrolyzer. These optimum PV modules (module #13) were
hydrogen efficiencies (8.3% and 10.3%) when connected to the
manufactured with 72 solar cells connected in series with the
electrolyzer (Table 2). The other PV-electrolyzer systems
correct type of semiconductor material to give a Vmpp of
tested had Vmpp levels too far from the Voper of the electrolyzer
0.50 V/cell and generate their maximum power when oper-
for efficient PV electricity and hydrogen generation (Fig. 8).
ating at 36 V, the maximum power point.
This comparison illustrates the fact that the PV cells selected
The relationship between Voper of the load (the PV-elec-
for a PV-electrolyzer system must both match the electrolyzer
trolyzer system) and the electrical efficiency of the PV module
operating voltage and also have high electrical efficiency.
#13, taken from the product literature, is graphed in Fig. 8.
Efficiency calculations were based upon the total PV solar cell
Since the PEM electrolyzer had a measured Voper of 32–33 V in
area.
all the PV-PEM electrolyzer experiments, the Vmpp of the PV
system must match the 32–33 V operating range to achieve the
highest efficiency. This Voper range is the characteristic oper- 3.4. DC–DC converter optimization
ating condition of the electrolyzer system (the sum of the
water splitting voltage and overvoltage of the electrodes – Eq. When a PV system is not designed to produce electricity effi-
(4)). Indeed, PV module #13 has a broad maximum in its ciently at the electrolysis operating voltage while directly
operating efficiency curve (Fig. 8) at a Vmpp range of 32–38 V in connected to the PEM electrolyzer, it is still possible to
which the module generates power with 17% electrical effi- increase the system efficiency by using a DC–DC converter to
ciency (which is 97% of its maximum electrical efficiency). convert the PV output voltage to the electrolyzer operating
This Vmpp range (17%) overlaps the Voper of the electrolyzer voltage as shown in Table 3 and Fig. 9. In this process, which

Table 3 – Optimization of solar hydrogen production using DC-DC converter devices

PV Vmpp (V) Solar irradiance DCO solar DDC solar DC-DC
module of PV System (W/m2) H2 efficiency (%) H2 efficiency (%) converter used

#3 17.0 960 0 7.6 CV 20/33-20

#8 20.0 650 0 7.9 CV 20/33-20

#8 20.0 960 0 8.4 CV 20/33-20

#10 54.0 770 9.0 10.4 PPT 48-10

#10 54.0 860 9.0 10.5 PPT 48-10

#12 33.0 950 8.3 5.7 PPT 48-20

#12 33.0 960 8.3 5.8 PPT 48-20

#13 36.2 990 12.4 9.4 PPT 48-20

#13 36.2 990 12.4 9.0 PPT 48-20
 international journal of hydrogen energy 33 (2008) 5931–5940 5939

14 converter or charge controller can also be used with PV

12 modules to match the output voltage to the operating voltage
Solar Hydrogen

Direct Connection
Efficiency (%)

10 of the electrolyzer. With a system using a DC–DC converter,

8 the maximum efficiency was reduced to 10.6%.
DC-DC Converter Based on the increased efficiency of the optimized PV-PEM
6
4 electrolyzer system, solar powered fueling systems can be
2 developed for fuel cell vehicles for home and fleet use at
0 a more affordable cost than previously [1,2,6–12]. The costs for
10 20 30 40 50 60 PV-electrolysis production of hydrogen fuel have been esti-
Vmpp of PV System mated using equipment cost and fuel cost analysis
approaches [3,4]. Based on the increase in efficiency and the
Fig. 9 – Average solar hydrogen production efficiency using
decreasing future PV system costs (a doubling of cumulative
PV-electrolyzer systems with DC–DC converters for
production and 20% price decrease has occurred every 2.5
optimization versus direct connection.
years) that have been forecast based on trends in photovoltaic
technology [20], the cost to generate a given amount of
hydrogen could decrease by a factor of 5 or more within the
we have called DC–DC converter (DDC) optimization, the PV next 15 years. Costs for solar hydrogen generation can also be
output voltage at the maximum power point (Vmpp) is raised or reduced by eliminating charge controllers, batteries, DC
lowered until it matches the operating voltage demanded by converters, and mechanical compressors included in previous
the PEM electrolyzer (Fig. 9). In the case of PV module #3, the PV-electrolyzer systems. The economic case for solar
Vmpp was too low to electrolyze water in the PEM electrolyzer hydrogen is expected to improve substantially as prices for PV
with direct connection optimization (DCO), but DDC raised the modules and electrolyzers continue to decline, and PV effi-
Vmpp from 17.0 V to 36 V and allowed 7.6% solar hydrogen ciency continues to improve as predicted from past experi-
efficiency (Table 3). Similarly, PV module #8, which could not ence [20–22]. PV modules, the chief cost for the systems, have
produce hydrogen by DCO, could produce hydrogen by DDC increased more than 60% in efficiency and decreased 50% in
optimization with about 8% efficiency. PV module #10 with cost per kilowatt in the past decade [21]. Commercial high
a Vmpp of 54 V (far above the operating voltage of the PEM pressure electrolyzers have been developed that produce
electrolyzer) operated with 9% efficiency by direct connection, hydrogen at 35–70 Mpa (5000–10,000 psi) to replace inefficient
but gave 10.5% efficiency when the voltage was stepped down and burdensome mechanical compressors [21]. The estimated
to 32 V by DDC optimization. cost of hydrogen fuel generated by PV-electrolysis, could,
When a PV module (#13), which was designed to give the therefore, be reduced from $14/kg for earlier, conventional PV-
optimum voltage to operate the electrolyzer, was tested with electrolyzer systems [3] to $4/kg for an optimized system. The
DCO and DDC methods, the DCO method gave 12.4% effi- hydrogen fuel cost would then compare favorably to equiva-
ciency, and the DDC device gave 9% efficiency (Table 3). In this lent driving costs using gasoline engines (with gas prices at
case, a percentage of the power and, thus, a significant part of $3–4/gallon) while a kg of hydrogen used in a fuel cell vehicle
the potential efficiency was lost due to the resistance of the (FCV) can give the same mileage as 2 gallons of gasoline in an
DC–DC converter. These results indicated that better solar ICE vehicle [21]. The estimated fuel requirements given here
hydrogen efficiency could be achieved by using DCO with a PV are based upon a FCV that is driven an average of 53 km/day
system that produces the optimum Vmpp for operating the (19,300 km/year) with a fuel economy of 97 km/kg of
PEM electrolyzer and is directly wired to the electrolyzer hydrogen.
compared to a system achieving optimization by going The expected decrease in costs for non-polluting solar
through a DDC. hydrogen from using an optimized PV-electrolyzer system can
Thus, DDC optimization improved solar hydrogen effi- assist in building a future hydrogen economy in which
ciency in three cases where the PV system Vmpp was suffi- hydrogen would replace fossil fuel consumption. Switching to
ciently different from the electrolyzer Voper, but in two cases renewable hydrogen fuel would, of course, eliminate the
where the Vmpp was close to Voper choosing direct DCO was associated green house gas emissions from fossil fuels and the
preferable to DDC. resulting potential for global warming. The compact size and
simplicity of the optimized PV-electrolyzer system described
in this report could supply hydrogen for both home FCV
4. Conclusions and future implications refueling and home power generation systems using
hydrogen fuel cells. In sunny parts of the U.S., small (~6 kW)
We found that the solar to hydrogen efficiency of PV-electro- electrolyzers powered by approximately 37 m2 (6 m  6 m) of
lyzer systems is maximized when the Vmmp of the PV system PV modules could usually produce enough hydrogen (0.6 kg/
matches the Voper of the electrolyzer. The best and simplest day) to operate a FCV for 36 miles/day. This estimate is based
way to obtain this match is with a PV system that is directly upon an area such as Los Angeles that can average 5.6 kWh/
connected to a PEM electrolyzer and has a Vmmp at or slightly m2 of sunlight per day on PV modules with a fixed tilt. The area
above Voper. By directly connecting a solar module with the of PV modules and power of the electrolyzer system could be
proper Vmpp to a high efficiency PEM electrolyzer we were able multiplied to supply a larger fleet of vehicles or operate
to achieve a solar to hydrogen efficiency of 12.4%, compared to a commercial filling station. The PV system can be stationary
past systems that only reached 2–6% efficiency. A DC–DC and mounted facing to the south tilted at an angle
5940 international journal of hydrogen energy 33 (2008) 5931–5940

approximating the site latitude, or a solar tracking system installation for a residential home in Switzerland. Int J
could be used to follow the sun and increase energy collection. Hydrogen Energy 2000;25:97–109.
Using a solar tracking system instead of a fixed orientation [9] Clapper WL. SunLine Transit Agency: hydrogen
commercialization for the 21st century, http://www.eere.
(latitude tilt) PV system could decrease the area of PV modules
energy.gov/hydrogenandfuelcells/pdfs/30535ah.pdf; 2001.
required by up to 25% depending on the location. [10] Ghosh PC, Emonts B, Janssen H, Mergel J, Stolten D. Ten years
of operational experience with a hydrogen-based renewable
energy supply system. Sol Energy 2003;75:469–78.
[11] Vidueira JM, Contreras A, Veziroglu TN. PV autonomous
Acknowledgement installation to produce hydrogen via electrolysis and its use
in FC buses. Int J Hydrogen Energy 2003;28:927–37.
Our warmest thanks go to Andy Galant for his help in devising [12] Neil D. A week without dinosaurs Los Angeles Times
the power supplies, data systems, sensors, general elec- Magazine, August 29; 2004, p. 9–11.
tronics, and instrumentation used in our system optimization [13] Kelly NA, Gibson TL. Design and characterization of a robust
photoelectrochemical device to generate hydrogen using
experiments.
solar water splitting. Int J Hydrogen Energy 2006:1658–73.
[14] NREL solar radiation data manual for flat-plate and
concentrating collectors, http://rredc.nrel.gov/solar/pubs/
references redbook/; 2004.
[15] Iqbal M. Introduction to solar radiation. New York: Academic
Press; 1983. p. 354–62 [chapter 8].
[1] Schucan T. International energy agency hydrogen [16] Gray JL. Physics of the solar cell. In: Luque A, Hegedus S,
implementing agreement task 11: in systems: final report of editors. Handbook of photovoltaic science and engineering.
subtask A: case studies of integrated hydrogen energy New York: Wiley; 2003. p. 87–106 [chapter 2].
systems. IEA/H2/TII/FR1; 2000. [17] Emery K. Measurement and characterization of solar cells
[2] Lehman PA, Chamberlin CE, Pauletto G, Rocheleau MA. and modules. In: Luque A, Hegedus S, editors. Handbook of
Operating experience with a photovoltaic hydrogen energy photovoltaic science and engineering. New York: Wiley;
system. Int J Hydrogen Energy 1997;22:465–70. 2003. p. 701–47 [chapter 16].
[3] National Research Council/National Academy of Engineering. [18] Licht S, Wang B, Mukerji S, Soga T, Umeno M, Tributsch H.
Hydrogen economy: opportunity, costs, barriers, and R&D Efficient solar water splitting, exemplified by RuO2-
needs. Washington: National Academy Press; 2004. catalyzed AlGaAs/Si photoelectrolysis. J Phys Chem B 2000;
[4] Simbeck DR, Chang E. Hydrogen supply: cost estimate for 104:8920–4.
hydrogen pathways – scoping analysis. NREL/SR-540-32525. [19] Gileadi E. Electrode kinetics for chemists, chemical
US Department of Energy National Renewable Energy engineers, and materials scientists. New York: Wiley; 1993. p.
Laboratory; 2002. 502–05 [chapter 29].
[5] Turner JA. A realizable renewable energy future. Science [20] Surek T. Crystal growth and material research in
1999;285:687–9. photovoltaics: progress and challenges. J Cryst Growth 2005;
[6] Brinner A, Bussmann H, Hug W, Seeger W. Test results of the 275:292–304.
hysolar 10 kW PV-electrolysis facility. Int J Hydrogen Energy [21] Myers DB. Cost performance of stationary hydrogen fueling
1992;17:187–97. appliances. DOE hydrogen program review, NREL/CP-610-
[7] Winter C-J, Fuchs M. HYSOLAR and solar-Wasserstoff- 32405; 2002.
Bayern. Int J Hydrogen Energy 1991;16:723–34. [22] Schaeffer GJ, de Moor H. Learning in PV trends and future
[8] Hollmuller P, Joubert JM, Lachal B, Yvon K. Evaluation of prospects. In: 19th Euro PV Solar Energy Conference, June 7–
a 5 kWp photovoltaic hydrogen production and storage 11, Paris; 2004.