i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0

Available online at www.sciencedirect.com

ScienceDirect

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

On the ability of pem water electrolysers to provide

power grid services

L. Allidi
eres a, A. Brisse b, P. Millet c,*, S. Valentin a, M. Zeller b

a
Air Liquide, 1 Chemin de La Porte des Loges, 78350 Les Loges-en-Josas, France
b
EIFER, Emmy-Noether-Str. 11, 76131 Karlsruhe, Germany
c
Paris-Sud University, ICMMO, Ba^ t. 410, 15 Rue G. Clemenceau, 91405 Orsay, France

article info abstract

Article history: Water electrolysis is considered as a cornerstone technology for the large scale storage of
Received 23 September 2018 energy and for carbon abatements in the frame of the energy transition. The purpose of
Received in revised form this research work was to analyze power grid operational constraints, to design specific
3 November 2018 load profiles of interest for power grid management and then to use these protocols for the
Accepted 22 November 2018 characterization and qualification of polymer electrolyte membrane (PEM) water electro-
Available online 19 December 2018 lysers in view of grid balancing services. In the first section, management constraints of
European power grids are described and analyzed. Using a typical regulation mechanism as
Keywords: an illustrative example, power specifications for primary and secondary reserve manage-
Hydrogen production ment are specified. The economics of such management procedures is also analyzed. In the
PEM water electrolysis second section, some key technical characteristics of PEM water electrolysis stacks are
Power grids described. Test specifications designed for the qualification of water electrolysers to both
Grid services primary and secondary reserve markets are defined. In the last section, selected test results
are reported and the ability of PEM water electrolysis stacks to provide the services of in-
terest is analyzed. In particular, a set of key performance indicators, designed for the
characterization of PEM water electrolysers operating in transient power conditions of
interest for grid services, are defined. Test results show the ability of PEM water electrolysis
stacks to satisfy the most stringent grid constraints, but remaining limitations are iden-
tified. The main innovative contributions of this research work were to design test pro-
tocols for both primary and secondary power reserves management, and to demonstrate
that PEM water electrolysers can be used for such applications.
© 2018 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

on a two-step reversible process (the water dissociation for

Introduction energy storage and hydrogen electrochemical combustion in
oxygen or air for re-electrification), and establishes water
Large scale energy storage is at the heart of the energy tran- electrolysis as a cornerstone technology for future carbon
sition [1,2]. Since the mid-1970s and up to now, hydrogen has abatements [4]. Among various technical options, water
been considered as the main chemical energy vector for that electro-dissociation based on polymer electrolyte membrane
purpose. The so-called concept of hydrogen economy [3] relies (PEM) cells [5,6], operating in near-ambient temperature

* Corresponding author.
E-mail address: pierre.millet@u-psud.fr (P. Millet).
https://doi.org/10.1016/j.ijhydene.2018.11.186
0360-3199/© 2018 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0 9691

conditions, is currently considered as very promising for grid is obviously a complex and vital task. A large variety of
operation with transient power sources. Potentially, electro- stakeholders are contributing to the regulation process, in
lytic hydrogen could find applications in the domestic, energy order to ensure the best possible power service, a pre-requisite
and transport sectors. However, large (>100 MW) water elec- to any economic activity. Transmission system operators
trolysis plants will be needed to satisfy future market re- (TSO) are entities entrusted to transport energy through fixed
quirements. Such plants will have to support severe infrastructures at local, regional, national or international
operational constraints, both on the upstream (specified by levels. The TSOs operating the UCTE grid are now part of the
grid operators) and downstream (specified by hydrogen end- European Network of Transmission System Operators for
users) sides of the process, including those related to storage Electricity (ENTSO-E), an association of various European
and distribution options. The cost of electricity is a key driver TSOs [14]. The national TSO in continental France is Re  seau de
for the large scale implementation of such technologies 
Transport d’Electricit  (RTE). The job of TSOs is therefore to
e
because on large (>MW-scale) water electrolysis plants, the guarantee the equilibrium between the supply of, and the
electricity cost amounts up to ~80% of the hydrogen cost [7]. In demand for, electrical power. For both safety and economical
this context, the participation of water electrolysis plants to reasons, equilibrium and reliability are critical issues. The lack
grid regulation procedures [8,9] can potentially provide the of balance between power generation and consumption can
necessary additional revenues required to reduce the be the source of various critical problems such as frequency
hydrogen production cost down to suitable levels (~5 V/kgH2), fluctuations and interruptions of supply. Market actors incur
in order to become cost-competitive with steam methane their financial liability whenever then fail to maintain a set of
reforming. operational indicators within a range of legal specifications.
Whereas most recent papers related to power grid man- Besides the uptime of the supply on the grid, the instan-
agement discuss issues related to the use of PV [10], wind taneous voltage and frequency available any time and in any
turbine [11], fuel cells [12], or combined technologies [13], the place on the grid, are the two main parameters that tell the
interest of using large power stand-alone water electrolysers quality of delivered power. In particular, the 50 Hz frequency
to provide grid balancing services is a more recent topic which set point of AC power on the UCTE grid is the main indicator
is currently under investigation among different groups. The used to assess that the grid is operating in a proper way. When
purpose of this research paper was to analyze power grid a lack of balance occurs between power generation and con-
operational constraints, to design specific load profiles of in- sumption (the balance of power imports and exports among
terest for power grid management, and then to use these interrelated networks is also taken into account), there is a
protocols for the characterization and qualification of labo- change of frequency on the grid: the frequency tends to in-
ratory and commercial PEM water electrolysis technologies. In crease when more power than necessary is supplied to the
the first section, management constraints of European power grid and conversely, the frequency tends to decrease when the
grids are described and analyzed. Using a typical regulation demand overcome the supply. At any time, the frequency
mechanism as an illustrative example, power specifications should remain within the 49.5e50.5 Hz range. The situation
for primary and secondary reserves management are speci- can be illustrated by the mechanical analogy of Eq. (1) where
fied. The economics of such management procedures is also the rotating speed u is analog to the grid frequency, the motor
analyzed. In the second section, the technical characteristics torque Cm is analog to the generation load, and the resistant
of PEM water electrolysis plants are summarized. Some test torque Cr is analog to the consumption load, J being the inertia
specifications required for the qualification of water electro- of the electrical system:
lysers to both primary and secondary power reserves are
du
defined. A list of key performance indicators used to charac- J, ¼ Cm  Cr (1)
dt
terize PEM water electrolysis plants operating under transient
power conditions of interest, are provided. Some key findings To satisfy such operating requirements, the TSO has
showing the ability of PEM water electrolysis stacks to satisfy several options. In particular, he has at his disposal a primary
the most stringent grid constraints and to identify remaining (PR) and a secondary (SR) power reserves, a list of pre-specified
technology limitations, are summarized and discussed. and pre-qualified volume of power that can be supplied to/
withdrawn from the grid within a strict dynamic timeframe.
The set of dynamic specifications applicable to both PR and SR
Description of grid requirements in Germany are publicly available [15]. Despite some differ-
ences (that should disappear in the near future due to the
Overview of grid management constraints deeper interconnection of European national grids), a similar
set of specification is available in continental France. From a
The synchronous grid of Continental Europe or Continental practical viewpoint, both primary and secondary reserves are
Synchronous Area (formerly known as the UCTE grid i.e. the put at the disposal of the TSO by producers and consumers via
grid of the Union for the Coordination of Transmission of a prescription/market mechanism:
Electricity) is the largest synchronous power grid in operation
in the world. This is the interconnection of single phase-  The mechanism used to release the primary reserve is
locked 50 Hz (main frequency) electricity grids that supply controlled by the instantaneous frequency on the grid; it is
over 400 million customers in 24 different countries, including activated by local speed regulators and made available
most of the European Union. The reliable operation of such within 30 s; its main role is to stop any frequency deviation;
9692 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0

when fully activated, the frequency drift is stopped but the From a practical viewpoint, the grid frequency is measured
difference between actual and nominal frequencies re- locally, close to the place where the equipment used for
mains unchanged and still needs correction. regulation is connected to the grid, with an accuracy of
 The purpose of the secondary reserve, which is activated 10 mHz. The frequency shift threshold for the activation of the
after the primary reserve, is to set the grid frequency back primary reserve is 10 mHz. The instantaneous power response
to the nominal value; it is made available within 5 min; in of the equipment should not exceed the expected theoretical
France, it is activated using a specific command signal response and should not oscillate around the power set point
designed by RTE and updated every 5 s. once the target value is reached. It should be outlined here
that grid power regulation is a dynamic process: there is a
The activation of the primary reserve is a continuous pro- continuous variation of the set point. A template is used to
cess and the activation of the secondary reserve is a semi- control the dynamic response of the equipment. This tem-
continuous process. The technologies which are selected to plate contains two model curves that bracket the expected
contribute to this regulation process must be able to contin- power response of the equipment. The first one is the theo-
uously adapt their power. The same constraints apply to those retical and instantaneous set point curve provided by the TSO;
who inject power onto the grid (energy producers) and to the second one is deduced from the theoretical one by
those who take power from the grid and contribute to the applying a first order filter with a time constant of 10 s. This is
regulation (regulators). In the equations used to define the the worst expected response of the qualified equipment.
specifications, there is only a difference of sign (a plus sign for During operation, instantaneous power provided by the
producers and a minus sign for regulators). This difference equipment should remain at least 95% of the time somewhere
has no impact on the dynamics of the power flows. In the case between these two limit curves. Let DPðtÞ be the instantaneous
of numerical regulation, the sampling rate must be suffi- power difference between (i) P (t), the instantaneous set point
ciently high to stick to the specifications. In France and Ger- power specified by the TSO (the Heaviside step in red on Fig. 1),
many, primary and secondary contributing technologies must and (ii) P0, the constant baseline power of the equipment
demonstrate their ability to operate in a symmetrical way operated by the entity. DPðtÞ is a function of PR and of the
(power injection/withdrawal, at the same rate). There are instantaneous grid frequency f(t):
some differences between the two countries in terms of
  PR
technical specifications for the secondary reserve. The most DPðtÞ ¼ PðtÞ  P0 ¼ ½fðtÞ  50 : (3)
0:2
stringent conditions imposed in France were used to design
the power profiles of interest for the qualification of PEM water A “real-life” example is provided on Fig. 2, where the situ-
electrolysers. ation that prevails for a 1 MW baseline system qualified for a
200 kW PR (primary reserve) power is highlighted. The plot
Power specifications for the primary reserve extends over 30 min of time only to better show the fluctua-
tions. The black dotted line (Y-axis on the right side of the plot)
The entity planning to become an operator on the primary shows the instantaneous grid frequency that fluctuates (in
reserve market and seeking the qualification of its equipment this example between ± 50 mHz) due to grid activity. The two-
(for example a water electrolysis plant operator) for a power curve template (Y-axis on the left side of the plot) is defined by
amount of PR kW, should demonstrate that its technology can the green (instantaneous theoretical set point) and red
change its baseline power P0 over the (P0-PR) to (P0þPR) power (instantaneous degraded set point still acceptable by the TSO)
range within a maximum of 30 s. The driver used to release curves. In order to contribute to the regulation of the grid
the PR is the grid frequency. The PR should be released pro- frequency, the PEM water electrolysis plant must follow the
portionally to the frequency shift measured on the grid, and template and spend at least 95% of the time somewhere be-
the total PR amount should be released for a frequency shift of tween the green and red lines otherwise penalties may apply.
200 mHz. The amount of power released per frequency unit is Controls are made on a periodical basis, either remotely or
given by Eq. (2): onsite.

PR Power specifications for the secondary reserve

K MW:Hz1 ¼ (2)
0:2
Whenever there is a regulation PR power demand on the The entity who plans to become an operator on the market of
grid, the power should be released according to a set of the secondary reserve and seek the qualification of its equip-
specifications (the situation is summarized in Fig. 1 for a ment (for example a water electrolysis plant) for a power
ramp-up regulation process): amount of SR kW, should demonstrate that its technology can
change its baseline power P0 according to Eq. (4):
 50% of the power must be released in less than 15 s.  
 100% of the power must be released in less than 30 s. DPðtÞ ¼ PðtÞ  P0 ¼ SR,NðtÞ (4)
 The instantaneous fraction of the PR used for regulation where DP(t) has the same meaning as for primary reserve
should be maintained at least 15 min to leave enough time regulation; NðtÞ is the instantaneous value of the command
to the TSO to engage the secondary and tertiary reserves in signal provided by the TSO; SR is the total amount of power
order to bring the frequency value back to the nominal qualified by the entity for secondary reserve regulation.
50 Hz set point. Depending on whether power is injected onto (or taken from)
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0 9693

Fig. 1 e Schematic diagram showing the technical response of an equipment operating for the primary reserve in France
(source RTE): (d) power set point; (----) minimum expected response.

Fig. 2 e Plots showing the main power requirements as a function of grid frequency.

the grid, NðtÞ ¼ ±1. Its sign also depends on whether this is a lost, it might be reinitialized and dN/dt might take very large
ramp-up or a ramp-down of power. There are two main cat- values. To avoid problems, the maximum value of dN/dt (met
egories of ramps: soft ramps (SR available within 800 s) and during emergency power ramps) is set to ±2/133 s. For SR
emergency ramps (SR available within 133 s). The total applications, the equipment should be designed in such a way
amount of power used for SR activities should become avail- that the power qualified for regulation should be maintained
able for N ¼ 1 (respectively 1). From the dynamic viewpoint, indefinitely (in practice, until a new set point is provided by
the expected power response of the equipment should satisfy the TSO). As for the primary reserve, a two-curve template is
Eq. (5) at any time: used as a guideline to control the dynamic response of the
  equipment. P*(t) should remain at least 95% of the time
dN
PðtÞ  P ðtÞ  ,SR, Teq (5) somewhere between these two curves. Fig. 3 shows two ex-
dt
amples of power regulation constraints for the secondary
where dN/dt is the first time derivative of the command signal reserve (power ramps up and down). The two power profiles
N(t) and Teq ¼ 20s is a time constant that expresses the apply to soft (within 800 s) and emergency (within 133 s) power
maximum time delay acceptable by the TSO. At any time requests. For the ramp-up, the power set point (black line) is
during the regulation process, the transient power response given by Pc ¼ (Pmax-SR)þ(N x SR). The test starts with N ¼ 1
P ðtÞ of the equipment should remains within 20 s of the (Pc ¼ Pmax-2. SR) and ends with N ¼ þ1 (Pc ¼ Pmax). Conversely,
power set point PðtÞ and there should be no oscillations for the ramp-down, the power set point (black line) is given by
around the set point value. Whenever the command signal is Pc ¼ (Pmin þ SR)þ(N x SR). The test starts with N ¼ þ1
9694 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0

Fig. 3 e Schematic diagram showing the power requirements for operation on the secondary reserve in France.

(Pc ¼ Pmaxþ2. SR ¼ Pmax) and ends with N ¼ 1 (Pc ¼ Pmin). The well as errors made while forecasting expected energy con-
blue lines show the power limits between which the power of sumption or generation, especially when renewable energy
the equipment contributing to the regulation process should sources such as PV, wind or hydro are used, make the entire
remain at any time. The first limits (Pc þ εv for ramps-up and Pc system unpredictable. At the same time, TSOs, as grid oper-
- εv for ramps-down) are defined by εv which is the measure- ators, have the responsibility to ensure anytime the physical
ment uncertainty of the active power (the maximum value of balance of energy flows between production (þimport) and
εv is set to 5% of SR). The second limits (Ptol - εv for ramps-up consumption (þexport). In order to achieve this critical
and Ptol þ εv for ramps-down) are defined by Ptol ¼ Pc/(1þ objective, TSOs can mobilize several types of downward and
Teq. p). This is a tolerated set point obtained by the Laplace upward reserves ready to activate anytime, at different rates,
transformation (p is the Laplace operator) of Pc, using the time as discussed before. The procurement of reserves over vari-
constant Teq ¼ 20 s. These two limits are the worst expected able periods of time occurs typically through call of tenders or
responses of the equipment qualified for the secondary auction with different periodicity and contractual duration
reserve regulation. In Fig. 3, tb is the time at which the sec- (TSOs are not allowed to operate production means by
ondary reserve is made fully available on the grid and T is the themselves, despite some exceptions found in some Eastern
ramp duration þ100 s. Europe countries). Both producers and consumers (assuming
they can offer a critical power capacity) can put capacity at the
The economics of power grid services disposal of the concerned TSO, and get remunerated for this
when awarded. Two different kinds of remuneration exist
The existing regulation framework was set by successive En- (sometime simultaneously) for the same reserve product. One
ergy packages issued in the European legislation. A distinction (in V/MW/h) is based on the availability of the capacity offered
is made between the operational and the financial re- (this is called capacity payment). The other (in V/MWh) is based
sponsibility of the different actors, in order to ensure a on the amount of energy effectively activated upwards or even
continuously balanced operation of the European electrical downwards (this is called energy payment) during regulation. It
system (“load frequency control”). Grid actors, both electricity should be outlined here that the corresponding energy
producers and commercial companies selling electricity to a amounts involved in these regulation processes are small
portfolio of end-consumers (as well as electro-intensive compared to those considered on the European Energy Ex-
companies purchasing electricity by themselves through change (EEX) market. Producers can sale their production
bilateral contracts or orders passed on the gross market, i.e. above market prices and consumers can access cheaper
the European Energy Exchange) are financially responsible for electricity. The main issue is that the operation rate is low:
the balance of their portfolio, at anytime. This means that the operational expenditures (Opex) of electro-intensive equip-
energy amount consumed by their end-clients (or sold to third ments are reduced but the return on investment is much
parties) has to be equal, at anytime, to the amount of energy longer than in conventional business.
there are generating or buying from third parties. If this is not When a consumer participates to these reserves (demand
the case, TSOs are charging the concerned misbalances with side management), these two types of remuneration can both
significant fees. This is a situation illustrating the financial or separately contribute to reduce significantly the overall
responsibility of each grid user for balancing their opera- costs related to electricity purchase. Thus, for electro-
tional/commercial activities. intensive processes such as water electrolysis, these reserve
From a technical viewpoint, a perfect balance between markets represent a great opportunity for reducing the OPEX
energy consumption and supply is scarcely reached since the part of the hydrogen generation costs. Some examples of the
electrical system and the processes that rule the commercial current remuneration observed on the German auction plat-
activities are constantly affected by various perturbations. form for reserve products can be found in Ref. [16]. Fig. 4 il-
Indeed, plant outages, the behavior of transmission lines as lustrates the capacity payments (expressed in V/MW/h)
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0 9695

module during operation. Power consumption required for

water pumping is counted in the energy balance and usually
amounts to only a few percent of the entire plant consump-
tion. The liquid water/gaseous biphasic mixtures are gravi-
metrically separated in separators. On the cathodic circuit
(right side of the stack on Fig. 5), gaseous hydrogen is
collected; traces of oxygen are removed and the gas is dried.
Electro-osmotic water transferred from the anode compart-
ment down to the cathode compartment (as a result of the
transport of hydrated protons) is usually recycled to the water
reservoir or to the anodic loop for cost reasons.

Test specifications for the qualification of water electrolysers

Power profiles available in the literature usually consider

power changes over the entire (0%e100%) power scale. The
Fig. 4 e Capacity remuneration for primary, secondary and
main innovative contribution of this research work was to
tertiary reserve in France (year 2017).
specify test protocols closer to real operating conditions (those
shown on Fig. 2), that would translate into electrical load
observed for French reserve products during fiscal year 2017. profiles the main grid constraints described in the previous
The remuneration for the SR is constant (flat rate) because this sections, and would take into account the necessity to operate
service is not yet open to market due to different technical the unit with a power base-load for economical reasons. This
requirements and practices among European countries. was done for both primary and secondary reserves (Figs. 6 and
7 respectively). Both power profiles contain a succession of
ramp-up and down and intermediate stationary power pla-
The PEM water electrolysis plant teaus of fixed duration, in accordance to the specifications
discussed above. The basic idea behind such design is that the
Overview of PEM water electrolysis plants PEM water electrolysis plant as a whole should operate around
a pivotal nominal power (the power base-load), which is
Depending on its size and hydrogen production capacity, a approximately 50% of the maximal power. In addition to this
commercial PEM water electrolysis system is either placed in a power base-load, the system should also be able to provide
transportable container, or directly installed on the hall of a grid services in response to grid requirements, using the set of
plant, usually on industrial sites. Most commercial PEM water ramp-up/ramp-down and stationary power plateaus of in-
electrolysers are designed for operation under stationary terest. The time constants of these different steps are dictated
power conditions. The objective of this research work was to by the grid requirements and the type of reserve of interest, as
determine the potential of the technology as a whole (elec- discussed above.
trolysis stack and balance of plant or BoP) to address markets
in transient power operating conditions, and whether it is
appropriate or not to provide grid services. Fig. 5 provides a Selected test results
schematic overview of a conventional PEM water electrolysis
plant [16]. The core component is the electrochemical reactor, Efficiency
the electrolysis stack of cells itself which, depending on the
production capacity of the module, can contain up to several The efficiency of the PEM water electrolysis plant as a whole is
hundred individual cells connected in series in a filter press mainly dictated by the efficiency of the PEM water electrolysis
configuration. Each cell has usually a surface area in the stack. The efficiency of the PEM water electrolysis stack is
1000e1500 cm2 range, but efforts are currently made to mainly dictated by operating current density and operating
develop much larger cells, at the square meter scale. During temperature. At the laboratory scale, most setups are using a
operation, each cell requires an operating voltage in the temperature regulation unit that can set the test temperature
1.8e2.0 V range, corresponding to a current density of to any value of interest, independently of the operating cur-
1e3 A cm2. The necessary operating DC power source is rent density. On larger commercial systems, electrolysis
provided by a power conditioning unit containing an AC/AC stacks are usually operated in exothermic mode and there is
transformer and an AC/DC rectifier. Water purification (milli- only a need to extract the excess heat produced at nominal
Q® grade water, corresponding to a residual conductivity of operating conditions: the maximum temperature is set to the
0.1e1.0 mS cm1, is required) is achieved onsite using several nominal value and there is no temperature regulation unit as
water treatment units including UV treatment, total organic such, only a heat extraction unit used to prevent overheating.
carbon abatement and osmosis. Purified water is pumped This is because most commercial PEM water electrolysis
through the hydraulic circuit placed on the anode side of the stacks are operated at constant current density and temper-
reactor (left side of the stack on Fig. 5) and water conductivity ature. But grid services require continuous power changes.
is continuously monitored since various deleterious corrosion Hence, the stack temperature is not constant and this situa-
processes tend to degrade the efficiency of the electrolysis tion impacts the energy consumption (and efficiency) of the
9696 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0

Fig. 5 e Schematic diagram showing the main modules of a PEM water electrolysis plant.

Fig. 7 e Specific power profiles designed for secondary

Fig. 6 e Specific power profiles designed for primary reserve qualification.
reserve qualification.

5 kW h.kg1
H2 (curve a) and a loss of efficiency of approximately
5% (curve b). Such behavior is directly related to the changing
water splitting reaction. A typical example is provided in temperature during the test, a situation affecting both mem-
Fig. 8. The PEM stack is operated galvanostatically and non- brane conductivity and charge transfer kinetics. Power regu-
isothermally. The driving parameter used during the experi- lation instead of current regulation and strategies to buffer
ment is the operating current density supplied to the stack (up heat (and favor isothermal operation) still need to be imple-
to z 2 A cm2). The initial temperature of water is sub- mented on large commercial PEM water electrolysis stacks in
nominal (z48  C). When current is turned on to the set view of power grid services to avoid unnecessary efficiency
point within seconds, the stack temperature gradually in- losses.
creases due to internal irreversible dissipations (curve d).
Operation in galvanostatic mode leads to a stack voltage Flexibility and power stability
overshoot and, as a result, to a power overshoot (curve c) that
take z10 min to stabilize in that case (overshoots gradually Flexibility is a term making reference to the possibility to
decrease as the stack temperature reaches a stationary tem- operate the PEM water electrolysis plant at different power
perature value of z57  C). As a result, there is an over- values. Again, flexibility capabilities of the entire system are
consumption of specific energy of approximately mainly dictated by stack capabilities, and to a lesser extent by
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0 9697

Fig. 8 e Results of a galvanostatic on/off test using a 20-cells PEM water electrolysis stack (a) energy consumption; (b) water
splitting efficiency; (c) power density; (d) water temperature.

balance-of-plant (BoP) capabilities (at least when the BoP is the cell design but also by the temperature of the environment
adequately designed). Fig. 9 shows the result of a typical for cooling purposes. When the water electrolysis unit is
flexibility test. The data were collected during operation under operated in such a way, remuneration can be obtain from both
current-controlled (galvanostatic) conditions, a choice usually hydrogen production and grid services.
dictated by down-stream hydrogen-consumption re- Power fluctuation on power plateaus is another issue of
quirements. The observed power overshoot can be corrected interest for grid services. Specifications may vary from one
by using power-controlled DC power sources. Results show grid to another but qualification for grid services usually re-
that during operation, the electrolysis unit can stand idle as quires percent-range power stability. The electrolysis stack is
long as necessary and then jump to either nominal or the main power dissipation contributor of the PEM water
maximal current (or power) density and back within seconds. electrolysis plant, and the main contributor to the power
These results show that PEM water electrolysis stacks have consumption signature. The problem is that during operation,
the necessary flexibility required to provide electrical services the stack impedance (which is a function of operating pa-
to power grids. PEM stacks and plants can be designed for rameters, mainly current density and temperature), can fluc-
operation over the entire 0e100% power range. From the tuate. As a result, there are power fluctuations that need
economical viewpoint however, there is no real interest to careful optimization to remain within specifications. Fig. 10
maintain the system in idle state. As discussed above, the shows the situation measured on a 20-cell stack at both
current trend in the industry is to develop systems that can nominal and maximum operating current density. There is a
combine a power base-load operation (at nominal power z ± 2% power fluctuation over the mean power value on the
condition) to a grid service capability, on top of the base-load,
up to a maximum power value. Such upper limit is dictated by

Fig. 10 e DC power fluctuations measured on a 20-cells

Fig. 9 e Current density versus time plot measured on a 20- PEM water electrolysis stack operated at (a) nominal and (b)
cells PEM water electrolysis stack showing flexibility. max power conditions.
9698 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0

Reactivity

Reactivity is a term making reference to the time it takes to the

PEM water electrolysis stack/plant to change its power set
point. Again, the reactivity is mainly dictated by the water
electrolysis stack, and to a lesser extent by the balance-of-
plant. Fig. 11 shows the situation for both primary (Fig. 11a)
and secondary (Fig. 11b) reserves. In both cases, data were
obtained during a ‘hot start’ experiment, i.e. when the stack
temperature is kept constant at nominal operating value
before turning power on/off. For the primary reserve test
(Fig. 11a), the constraint imposed during the test was to switch
power density from 100% down to 75% of nominal power
density within less than 30 s, applying a linear power tran-
sient. For the secondary reserve test (Fig. 11b), the constraint
Fig. 11 e Power density versus time plots showing the was to switch power density on from initial (0%) to final (100%)
reactivity of a 20-cells PEM water electrolysis stack for (a) set points in less than 2 min and backwards (soft and emer-
primary and (b) secondary reserve operation. gency ramps), again using a linear power sweep. The two
curves of Fig. 10 show that both constraints are adequately
satisfied. This is a general characteristic of PEM water elec-
trolysis technology (the same situation prevails on large, MW-
scale, PEM water electrolysis plants), showing that the tech-
power plateau. There is a number of factors that contribute to nology is suitable for operation on both power reserves. Such
such situation (e.g. turbulences due to gas production within characteristics are due to the use of a thin proton-exchange
the cells, pressure changes and variable heat transfer char- membrane as polymer electrolyte and cell separator instead
acteristics) but it is beyond the scope of this paper to discuss of a liquid electrolyte, a unique feature of PEM water elec-
these issues in detail. trolysis cells.

Fig. 12 e KPIs of PEM water electrolysis technology. (--) reference case (stationary, 2015); (¡) test results from this work
(transient PEM, 2017).
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0 9699

KPI of PEM water electrolysers for grid services for such applications. KPI values of PEM water electrolysis
technology were found mostly above average, except for the
A SWOT analysis of PEM water electrolysis technology has temperature of operation (a limitation due to the thermal
been performed. A set of key performance indicators (KPIs) stability of the polymer electrolyte) and the safety of operation
has been designed to measure the level of development of PEM (gas cross permeation issues across the polymer electrolyte at
water electrolysers, to measure the impact of operation with low current density).
transient power sources (for grid services) on the perfor-
mances, and to facilitate the comparison of commercial
products available on the market. This is an extension of the Conclusions and recommendations
list of indicators provided by the European Commission in its
2014e2020 annual work program [17]. Main results are sum- Hydrogen of electrolytic grade obtained by water electrolysis
marized on the polar plot of Fig. 12. The reference case used is potentially a game-changing energy vector. In the future, it
for comparison is intended to measure mean water electrol- is expected that water electrolysis plants will contribute to the
ysis performances (KPIs are arbitrarily set to 5 on a 0e10 scale, large-scale storage of transient renewable energy sources, but
corresponding values are compiled in Table 1). This is a also to power grid regulation. In this research paper, power
measure of the state-of-art in 2015, for systems (mainly grid operational constraints have been analyzed, specific
alkaline ones) operating under stationary power conditions. A power profiles of interest for power grid management have
first group of KPIs is used to describe operating conditions. been designed and these profiles have been used to test and
This is mainly the T, P, j range (T: temperature, P: pressure, j: qualify polymer electrolyte membrane (PEM) water electro-
current density) of operation. The KPIs of the second group are lysers. In the first section, management constraints of Euro-
used to measure the level of advancement of the technologies, pean power grids have been described and analyzed, with a
in terms of production capacity, engineering, and special focus on the power profiles of interest for primary and
manufacturing capabilities. This is mainly the active area of secondary reserves management. The economics of power
individual cells (that shows the ability of the manufacturers to grid management has also been discussed. Some test pro-
design large cells) and the stack power (that shows the capa- tocols required for the qualification of water electrolysers to
bility of the manufacturers to design large systems needed by both primary and secondary reserves have been designed and
the market). The KPIs of the third group are specifically those are presented. They have been used to test PEM water elec-
used to assess the ability of the technology to provide grid trolysis stacks and plants of different sizes. Main test results
services. In the frame of the energy transition, power grids are have been summarized. In particular, the impact of grid ser-
calling for electrical systems that are flexible (i.e., which can vice operation on the efficiency, flexibility and reactivity of
change their power set point between minimum and PEM water electrolysis stacks is reported and discussed. A list
maximum values), and that are reactive (i.e., which can of key performance indicators has been used to assess the
quickly change power set point). The KPIs of the last group are ability of PEM water electrolysis plants to operate in transient
used to measure various performance levels and to visualize power conditions of interest. Key findings were that PEM
the impact of transient operating conditions. water electrolysis stacks/plants show the necessary flexibility
The power profiles of Figs. 6 and 7 have been used as test and reactivity to address the markets of primary and sec-
specifications for the qualification of different laboratory and ondary power reserves. However, there are two main critical
commercial PEM water electrolysis stacks and plants (i.e. issues that are limiting the performances of existing tech-
electrolysis stacks þ balance of plant or BoP) in view of grid nology and will require further developments: (i) the need for
services. All test results are summarized in Fig. 12. Despite the operation in isothermal conditions in order to reduce effi-
difficulty associated with the construction and use of quan- ciency losses and degraded power responses; (ii) the need to
titative scales that could be unanimously recognised by the reduce power fluctuations down to the appropriate level on
water electrolysis community, the polar plot provides a quick power plateaus. Based on these results, it is recommended to
overview of main strengths and limitations of PEM technology technology manufacturers to take these aspects into consid-
eration in order to further customize PEM water electrolysers
and make the technology ready for grid balancing applica-
Table 1 e Reference case of Fig. 12. tions. The financial support of funding agencies is required to
j range (A.cme2) 0e1
facilitate such developments. Limitations of commercial
T range ( C) 20e80 water electrolysers to comply with grid requirements should
P range (bar) 0e35 be taken into account by these agencies and existing road-
Cell active area (cm2) 1,000 maps should be updated to facilitate the large scale deploy-
Power range (MW) 0e1 ment of the technology.
Flexibility stationary at different j values
Reactivity (0 ® nominal power) < minutes
Efficiency (kWh-1kgH2) < 60
Capex (V/kW @ 1 MW-scale) 1,500 Acknowledgement
Durability (% performance loss/ < 5%
year)
The support of the French Research Institute on Energy
H2 purity at delivery (xN). 4
 Energe
Transition, Paris-Saclay Efficacite tique (PS2E), in the
Safety (% H2 in O2) < 25 % of the inferior limit of
explosivity (ILE) frame of the FlexiPEM project, is acknowledged. Financial
9700 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 4 ( 2 0 1 9 ) 9 6 9 0 e9 7 0 0

support from the French Agence Nationale de la Recherche [8] Guinot B, Montignac F, Champel B, Vannucci D. Profitability
(ANR 11 EMMA 048-01, ANR-RF-2015-01) and the French of an electrolysis based hydrogen production plant providing
ADEME (Ademe 1694C006) is also acknowledged. grid balancing services. Int J Hydrogen Energy
2015;40(29):8778e87.
[9] Bennoua S, Le Duigou A, Que  me
 re
 MeM, Dautremont S. Role
references of hydrogen in resolving grid issues. Int J Hydrogen Energy
2015;40(23):7231e45.
€
[10] Ozbay €
H, Oncü S, Kesler M. SMC-DPC based active and
reactive power control of gried-tied three phase inverter for
[1] Michalski J, Bünger U, Crotogino F, Donadei S, Schneider GS, PV systems. Int J Hydrogen Energy 2017;42(28):17713e22.
Pregger T, et al. Hydrogen generation by electrolysis and [11] Fang R, Shang R, Wang Y, Guo X. Identification of vulnerable
storage in salt caverns: potentials, economics and systems lines in power grids with wind power integration based on a
aspects with regard to the German energy transition. Int J weighted entropy analysis method. Int J Hydrogen Energy
Hydrogen Energy 2017;42(19):13427e43. 2017;42(31):20269e76.
[2] Akizu O, Urkidi L, Bueno G, Lago R. Tracing the emerging, [12] Mosaad MI, Ramadan HS. Power quality enhancement of
energy transition in the global north and the global south. Int grid-connected fuel cell using evolutionary computing
J Hydrogen Energy 2017;42(28):18046e63. techniques. Int J Hydrogen Energy 2018;43(25):11568e82.
[3] Scipioni A, Manzardo A, Ren J. Hydrogen economy, supply [13] Cervantes I, Hernandez-Nochebuena M, Cano-Castillo U,
chain, life cycle analysis and energy transition for Araujo-Vargas I. A graphical approach to optimal power
sustainability. Academic Press; 2017. management for uncertain OFF-Grid PV-FC-electrolyzer-
[4] Weidner S, Faltenbacher M, François I, Thomas D. Feasibility battery hybrid systems. Int J Hydrogen Energy
study of large scale hydrogen power-to-gas applications and 2018;43(42):19336e51.
cost of the systems evolving with scaling up in Germany, [14] https://www.entsoe.eu/. Accessed on September 18, 2018.
Belgium and Iceland. Int J Hydrogen Energy [15] https://www.regelleistung.net/ext/. Accessed on September
2018;43(33):15625e38. 18, 2018.
[5] Godula-Jopek A, Stolten D, editors. Hydrogen production by [16] Smolinka T. In: Bessarabov D, Wand H, Li H, Zhao N, editors.
water electrolysis. Wiley-VCH; 2015. PEM water Electrolysis for hydrogen production: principles
[6] Carmo M, Fritz DL, Mergel J, Stolten D. A comprehensive and Applications. Boca Raton: CRC Press; 2015.
review on PEM water electrolysis. Int J Hydrogen Energy [17] Multi-annual work program 2014-2020, ID 4221004, fuel cells
2016;38:4901e34. and hydrogen joint undertaking. European Commission;
[7] Saba SM, Müller M, Robinius M, Stolten D. The investment 2013.
cost of electrolysis- A comparison of cost studies over the
past 30 years. Int J Hydrogen Energy 2018;38:1209e23.