PreprintPDF Available

Grid-Forming Services From Hydrogen Electrolyzers

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Hydrogen electrloyzers are power-to-gas storage devices that can facilitate large-scale integration of intermittent renewable sources into the future power systems. Due to their fast response and capability to operate in different loading conditions, they can be used as responsive loads providing support to AC grid during transients. This paper suggests taking one step further and using hydrogen electrolyzers to provide grid-forming services to the grid. As a result, the electrolyzer's role is elevated from supporting the grid (responsive load) to actively participating in forming voltage and frequency of the grid. The grid-forming capability of electrolyzer is linked to its hydrogen production constraints, which can potentially pose limitations on the grid-forming services. Besides the grid-forming mode, two additional operating modes, i.e., DC voltage mode and constant power mode, are proposed to ensure a safe operation of the electrolyzer in case of adversary interaction between grid-forming operation and hydrogen production constraints. This paper also studies the impacts of grid-forming services on the electrolyzer's physical features such as hydrogen stack temperature and efficiency. Comprehensive simulations are conducted on a low-inertia test network whose topology is inspired by a portion of the transmission grid in South Australia to confirm the effectiveness of the proposed concept under various operational conditions of the electrolyzer and upstream AC grid. Moreover, the practical feasibility of the proposed control system is experimentally validated by conducting hardware-in-the-loop tests.
Content may be subject to copyright.
1
Grid-Forming Services From Hydrogen
Electrolyzers
Saman Dadjo Tavakoli, Mehdi Ghazavi Dozein, Senior Member, IEEE, Vin´
ıcius A. Lacerda, Marc Cheah
Ma˜
ne, Member, IEEE, Eduardo Prieto-Araujo, Senior Member, IEEE, Pierluigi Mancarella, Fellow, IEEE, and
Oriol Gomis-Bellmunt, Fellow, IEEE
Abstract—Hydrogen electrloyzers are power-to-gas storage
devices that can facilitate large-scale integration of intermittent
renewable sources into the future power systems. Due to their fast
response and capability to operate in different loading conditions,
they can be used as responsive loads providing support to AC grid
during transients. This paper suggests taking one step further
and using hydrogen electrolyzers to provide grid-forming services
to the grid. As a result, the electrolyzer’s role is elevated from
supporting the grid (responsive load) to actively participating
in forming voltage and frequency of the grid. The grid-forming
capability of electrolyzer is linked to its hydrogen production
constraints, which can potentially pose limitations on the grid-
forming services. Besides the grid-forming mode, two additional
operating modes, i.e., DC voltage mode and constant power mode,
are proposed to ensure a safe operation of the electrolyzer in
case of adversary interaction between grid-forming operation
and hydrogen production constraints. This paper also studies the
impacts of grid-forming services on the electrolyzer’s physical
features such as hydrogen stack temperature and efficiency.
Comprehensive simulations are conducted on a low-inertia test
network whose topology is inspired by a portion of the trans-
mission grid in South Australia to confirm the effectiveness of
the proposed concept under various operational conditions of
the electrolyzer and upstream AC grid. Moreover, the practical
feasibility of the proposed control system is experimentally
validated by conducting hardware-in-the-loop tests.
Index Terms—Hydrogen electrolyzer, Storage, Responsive load,
Grid-forming load.
I. INTRODUCTION
Hydrogen electrolyzers (HE) are electrochemical devices
used to split water molecules into hydrogen (H2) and oxygen
(O2) by passage of a DC current. They provide power-to-gas
energy storage solution which creates opportunities to leverage
intermittent renewable generations [1]. Major HE technologies
are polymer electrolyte membrane (PEM), alkaline with liquid
S. D. Tavakoli, V. A. Lacerda, M. Cheah, E. Prieto, and O. Gomis are with
the Centre d’Innovaci´
o Tecnol`
ogica en Convertidors Est`
atics i Accionaments,
Departament d’Enginyeria El`
ectrica, Universitat Polit`
ecnica de Catalunya,
Barcelona 08028, Spain. E-mail: saman.dadjo@upc.edu; O. Gomis, E. Prieto,
and M. Cheah are also with eROOTS Analytics.
M. Dozein is with the School of Electrical and Electronic Engineering,
The University of Melbourne (UoM), Victoria 3010, Australia (e-mail: mg-
hazavi@student.unimelb.edu.au).
P. Mancarella is with the School of Electrical and Electronic Engineering,
The University of Melbourne (UoM), Victoria 3010, Australia, and also
with University of Manchester, M13 9PL Manchester, U.K. (e-mail: pier-
luigi.mancarella@unimelb.edu.au).
This work was supported by Ministerio de Ciencia e Innovaci´
on
(MCIN/AEI/10.13039/501100011033) and by the European Union’s Next
Generation PRTR program (FAIR project), with reference TED2021-
129796B-C22. This project has also received funding from FEDER / Ministe-
rio de Ciencia, Innovaci´
on y Universidades - Agencia Estatal de Investigaci´
on,
Project EQUIRED PID2021-124292OB-I00. The work of O. Gomis was also
supported by the ICREA Academia program. E. Prieto and M. Cheah are
under Serra H´
unter Programme.
or solid electrolyte, and recently solid oxide electrolyzer.
PEM electrolyzers are manufactured from small sizes (several
kilowatts) up to large, utility-scale systems (multi-megawatt
plants) [2]. They can respond to changes in their power set-
points in sub-second time frame and operate at a wide range
of loading conditions [3]. In [2], it is experimentally identified
that a PEM electrolyzer takes only few hundred milliseconds
to change its hydrogen production rate in response to step
changes in the power set-point.
Hydrogen production in an electrolyzer is directly propor-
tional to its power consumption. The produced hydrogen is
typically stored in a large H2tank, from which the hydrogen
is consumed by a downstream process [4]. Such H2tank acts
as a buffer, so that hydrogen can be produced in a different
rate than it is consumed as long as the hydrogen content in the
tank is enough to supply the downstream hydrogen demand
continuously. This creates an opportunity for HE to operate
in different power set-points (different loading conditions)
without affecting the downstream hydrogen consumption [5],
[6]. Such flexibility in power consumption, along with the
ability to respond to the changes in sub-second time frame,
makes an electrolyzer as a potential candidate to be used as a
responsive load which supports AC grid dynamics during fast
transients.
Although several studies have already investigated the
techno-economic aspects of integrating HE as a power-to-gas
storage in power systems for maximizing the benefits from
the intermittent renewable energy sources [7]–[10], only few
studies have been conducted to identify the full potential of
HE as a responsive load. In [2], [11]–[13], it has been shown
that a HE can effectively operate as a responsive load to reduce
frequency deviations during grid disturbances. In these studies,
the HE plant is interconnected to the AC grid via voltage-
source converter (VSC) that operates in grid-following (GFL)
mode. In such operation mode, the active power set-point of
the VSC defines the power consumption of the HE. To make
HE operate as a responsive load , an active power-frequency
characteristic is added to the power control loop of the HE
VSC so that the HE power consumption varies as a response
to frequency variations. However, these studies either fully or
partially disregard the dynamics of the HE plant; in particular,
the operational constraints related to the hydrogen production
process are not linked to the control system of HE VSC.
In [14], the control system of HE VSC is designed based on
GFL operation and the hydrogen production constraints are
considered. It is discussed that the hydrogen content in the
tank may pose limitations on the active power set-point of the
HE VSC, and eventually its capability in frequency response
provision.
A version of this article has been accepted for publication by IEEE trans. on sustainable energy.
2
Recently, it has been shown that dispatchable loads with
demand flexibility can provide grid-forming services to the AC
grid [15]. In this concept, the role of loads in power system
is elevated from providing only frequency support to actually
participating in forming the voltage and frequency of the grid.
In fact, the power balance in the grid can be mostly achieved
through load demand variations rather than changes on the
generation side. To that aim, a grid-forming load regulates
frequency and AC voltage of the grid by, respectively, regu-
lating its power consumption and reactive power exchange. It
should be noted that such concept is only applicable to the
loads that are connected to AC grid via a VSC, so active
and reactive power can be regulated independently at the
point of connection. Several studies have considered grid-
forming functionality for hydrogen electrolyzer, most often
with the combination of a fuel cell unit [16], [17]. The hybrid
electrolyzer-fuel cell unit can operate in grid-forming mode
and depending on the loading condition in the AC grid, the
hybrid unit can either inject power to or absorb power from
the AC grid.
This paper explores the feasibility of implementing the grid-
forming (GFM) functionality in the HE VSC with the consid-
eration of hydrogen production constraints. The advantages
are twofold: (i) the function of HE in the power system will
change from responsive load (GFL control) to grid-forming
load (GFM control); hence, the HE actively participates in
forming voltage and frequency of the AC grid instead of
only supporting it, (ii) the tandem operation of HE with a
nearby renewable source enables islanded operation in case of
contingency in the AC grid. The HE operation in GFM mode
may interact with the constraints in hydrogen production,
which can negatively affect HE grid-forming capability. Under
this condition, two additional operating modes are proposed:
DC voltage mode which is activated if the hydrogen plant
is shut down due to an internal failure in the electrolyzer or
the hydrogen tank is full, and constant power mode which is
brought to operation in case the hydrogen content in the tank
is critically low and the downstream hydrogen demand may
not be fulfilled. A PEM HE with the proposed control system
is integrated into a low-inertia test network whose topology
is inspired by a portion of the transmission grid in South
Australia. Matlab simulations are conducted on such system to
compare the dynamic performances of these operation modes
during disturbances in the grid, confirming the ability of HE to
safely switch among different operation modes under abnormal
conditions in the hydrogen production process and AC grid.
We also investigate the impact of grid-forming services from
hydrogen electrolyzer on its physical features such as tempera-
ture and efficiency. Finally, the HE is compared with a battery
storage system in terms of their capabilities in providing
grid-forming services. The contributions of this study can be
summarized as follows:
Proposing a dynamic modelling framework for hydro-
gen electrolyzers equipped with three operation modes
(i.e., grid-forming mode, DC voltage mode, and constant
power mode),
Highlighting possible interactions between HE grid-
forming services and the operational constraints in hydro-
gen production, as well as proposing the corresponding
solutions,
Studying the impacts of HE grid-forming operation on
the physical features of HE such as temperature and
efficiency.
II. FUNDAMENTALS OF HE DYNAMIC MODELING
In this section, the basic operation of a PEM HE is outlined,
and its dynamical model is developed. The constraints related
to the hydrogen production process are also discussed through
its fundamental equations.
A. Basic HE Operation Principles
Referring to [1], [4], a simplified process flow diagram of
a PEM HE is shown at the bottom of Fig. 1. The deionized
water is supplied to a water tank (marked by 1 ) from which
it is fed into the PEM stack (anode side) via a water pump.
The applied DC voltage at the stack, Vdc
h, causes water split
to hydrogen and oxygen. This process draws a DC current,
Idc
h. The oxygen is then passed through a gas-water separator
( 3 ) in which oxygen is separated from water. The remained
water is removed at the bottom and returned to the PEM stack
via a circulation pump, while the oxygen is released from the
top. On the cathode side, the produced hydrogen is passed
through a gas separator ( 6 ) to eliminate the remaining water
in the gas. After cooling down in a heat exchanger, hydrogen
is passed through the gas condensate and de-oxygenation ( 7
and 8 , respectively). Then, it is fed into a gas dryer ( 10 )
through a compressor. Finally, the produced hydrogen is stored
in the tank ( 11 ) which acts as a buffer between hydrogen
production and consumption.
B. Efficiency and Temperature of PEM stack
The voltage-current characteristic of a single PEM cell is
shown in Fig. 2, which illustrates a direct relationship between
cell’s voltage and current: a higher current passing through the
cell causes a slight increase in the cell’s voltage. With regard
to the cell’s efficiency, a higher current density leads to the
greater losses inside the cell, and thus the efficiency is reduced.
On the contrary, if a PEM cell is allowed to operate in a higher
temperature, it will have higher efficiency for the same current
density. However, higher operating temperature has a negative
impact on ageing and longevity of the cell. Hence, the optimal
temperature is commonly decided to find a balance between
efficiency and the service life [18].
C. HE Dynamical Model
There are two major methods for building the dynamical
model of an electrolyzer: (i) using the exact mathematics that
represent the chemical reaction in the cells and the dynamics
of major ancillary equipment, such as pumps, compressors,
heaters, etc., to derive the HE model [19]–[21]; and (ii)
experimentally identifying the electrolyzer model based on its
DC terminal characteristics [2], [22]. The former method, of
course, can obtain model with a very high accuracy, provided
3
ω
n
Pc
*
-
+
kp
+
+
Pc
PIv
-+
PIp
-+
Δω
Pm
*
Pc
Mode 2
Mode 1
Mode 3
θ
c
Pc
ω
c
Qc
vc
qd
ic
qd
power
calculation
*
rate limiter
Qc
Qc
*
-+
kq
+
+
vn
Δcurrent
control
vm
qd0
AC voltage
control
is
qd*
is
qd
qd0
abc
vc
qd
vm
abc
load
management limits
& trips
secondary
control
qd limits
modul.modul.
Lf
Cf
is
abc ic
abc
+
-
AC
Grid
Vt
dc
It
dc
+
-
Vh
dc Ih
dc
+
-
DC voltage
control
DC current
control
dm
θc
vc
qd
is
qd
ic
qd
abc
qd0
Pref
Qref
selector
water
inlet
vented O2
+
-
Pc
Qc
vc
abc
ic
abc
vc
abc
is
abc
H2 outlet
mode 1,2,3
1
2
3
4
4
5
6
6
7
8
89
10
11
water tank
water pump
heat exchanger
electrolyzer stack
H2 separator
condensate
de-oxygenation
compressor
hydrogen
Anode
Cathode
water
H2 dryer
O2 separator
H2 tank
1
2
3
5
7
9
10
11
local
control
vc
qd*
0
v
CB
Vt
dc
Vt
dc*
Vt
dc
Vt
dc*
It
dc
It
dc*
mH
mH
'
Fig. 1. Overall system structure of a PEM hydrogen electrolyzer
1.85
1.80
1.48
Current density (A.cm-2)
Voltage (Volt/cell )
full-load
(1~3 A.cm-2 )
partial-load
(0.2~0.8 A.cm-2 )
Efficiency (%)
100
82
80
60 °C
60 °C
80 °C
80 °C
Fig. 2. Voltage-current and efficiency of a PEM cell [18]
that all details and parameters of the HE system is available.
Then, the equations related to the chemical reactions in the
cathode, anode, and the membrane are derived and coupled
with those of water pumps, cooling fans, storage tanks, etc.,
to arrive at the final model. However, a disadvantage of
such complex approach, particularly in a multi-vendor HE
project where several companies deal with various parts of
the project, is that the design parameters of the electrolyzer
are often proprietary information which companies wish to
keep confidential.
In the latter modeling method, on the other hand, the HE’s
dynamic behaviour is analysed only based on its DC voltage-
current characteristics at the connection point (Vdc
h-Idc
hshown
in Fig. 1), which is more suitable for power system studies. In
this method, the power reference of the HE is varied between
a very low power to the rated power and the result DC current
and voltage are plotted [2]. From these plots, a general transfer
function which relates DC current to the DC voltage at the
terminal of the HE is estimated. Such modelling technique is
developed in [23] and adopted in this study.
Referring to [23], the dynamics of the PEM HE can be
formulated into a second-order transfer function:
H(s) = Idc
h
Vdc
h
=p1p2(sz)
(sp1)(sp2)(1)
where Vdc
hand Idc
hare DC voltage and current at the HE
terminal, respectively. The poles, p1and p2, are located on the
left half plane (LHP) and calculated based on the rise time, tr,
and settling time, ts, measured from the HE transient response,
ln ( p2
p1p2
+p1
p2p1
e
p2
p1)p1ts= ln (yss )(2)
ln ( p2
p1p2
+p1
p2p1
e
p2
p1)p1tr= ln (yr)(3)
where yss=0.05, corresponding to the response’s steady-state
error at the settling time, commonly 5%, and yr=0.5, which
is related to 50% of response at the rise time. Once p1and p2
are calculated, the zero, z, is selected to be within the LHP
region given by
p1+p2< z p1(4)
In fact, only trand tsfrom the experiment are needed to
calculate H(s). Such transfer function represents an equivalent
electrical circuit in which a DC voltage source and RC
elements are connected in series as illustrated in Fig. 3. The
DC voltage source, Vrev, represents the reverse voltage, i.e.,
the minimum voltage that allows the current flow through the
PEM stack,
Vrev =Vrev0+R T
2Fln ( P
P0
)(5)
where, Pand Tare stack pressure and temperature, respec-
tively; Vrev0and P0are reference voltage and pressure; Ris
the ideal gas constant expressed in 1 atm K1mol1, and F
is the Faraday constant. The passive electrical components are
calculated based on the poles and zero of H(s)as
Cop =p1p2
z(z2z(p1+p2) + p1p2)(6)
Rm=z2z(p1+p2) + p1p2
p1p2
(7)
Ri=z(p1+p2z)
p1p2
(8)
4
EDL branch
reverse
voltage
eq. (5)
+
-
Rm
Cop
+
-
Ri
Vrev
Ih
dc
Vh
dc
TP
gas concentration
resistor eq. (9)
stack
efficiency
eq. (11)
H2
production rate
eq. (10)
++
+
Rco
Rms
H2
tank dynamics
eq. (13)
F
mH
mH
'
Fig. 3. Equivalent model of a PEM hydrogen electrolyzer
The parallel connection of Cop and Rmaccounts for the
electrical double layer (EDL) phenomenon, which is related to
the non-uniform distribution of negative and positive charges
at the electrode-electrolyte interface. The EDL phenomenon
creates an energy barrier against the flow of electrons when
there is a need for sudden current change, which leads to a
response delay. Moreover, this phenomenon causes a transport
loss as modelled by Rm.
The internal resistance, Ri, is the sum of two resistances:
(i) resistance of the stack material, Rms, and (ii) the resistance
related to the concentration of hydrogen and oxygen on the
electrode surface, Rco, which acts as a barrier to the flow of
electron to reach the electrodes. Rco is a non-linear function
of the HE’s current,
Rco =β1ln ( β2
AIdc
h+ 1) (9)
where the constant coefficients, β1and β2, are obtained from
measurements, and Ais the electrode surface. Please see [23]
for further details.
D. HE Operational Constraints
The PEM HE is allowed to operate within a power con-
sumption range, resulting in variations in its input DC current,
Idc
h. Such demand flexibility comes from the ability of HE to
produce hydrogen in varying rates. Hydrogen production rate,
mHin Moles/s, is a function of Idc
h, as
mH=ηF
Ns
2FIdc
h(10)
where Fis the Faraday constant and Nsis the number of cells
per stack. The Faraday efficiency, ηF, is the ratio between
the actual and theoretical maximum amount of hydrogen as
produced by the HE. It is also a non-linear function of Idc
h,
ηF=α1exp ( α2
Idc
h
+α3
(Idc
h)2)(11)
where α1,α2, and α3are constant coefficients obtained
from experiments [23], [24]. As given by (11), the hydrogen
production efficiency is reversely related to the current, i.e.,
efficiency drops for higher currents (also see Fig. 2).
As illustrated in Fig. 1, the produced hydrogen, with the rate
given by mH, is transferred to a storage (H2tank 11 ). The
outlet of the storage supplies the hydrogen demand which is
given as hydrogen consumption rate, m
H. Such configuration
provides a buffer zone for hydrogen; hence, depending on
the size of the storage, there can be a mismatch between
hydrogen production and consumption rates, which provides
a great source of flexibility for the power system. In fact, the
PEM HE can vary its power consumption to produce hydrogen
in a varying rate as long as the hydrogen content in the tank
(buffer) is within the range given by
Vmin
HVH(t)Vmax
H(12)
where Vmin
His the minimum hydrogen content in the H2tank
that guarantees a steady hydrogen supply for the downstream
process (hydrogen consumption), and Vmax
His the maximum
capacity of the H2tank. So long as VHis within the limits,
the HE can be regarded as a fully flexible load capable of pro-
viding GFM services to the grid, i.e., varying its active power
consumption depending on the grid’s frequency deviation.
The capacity constraint on the H2storage given by (12)
can be expressed in terms of the stack’s current as follows:
assuming a disturbance occurs in the grid at t=t0, the HE
adjusts its power consumption automatically (based on the
GFM control) to respond to such event. The change in the
power consumption leads to a change in hydrogen production
rate (mH), affecting the hydrogen content in the tank over this
period of time,
VH(t) = V0
H+Zt
t0
mH(t)dt Zt
t0
m
H(t)dt (13)
where V0
His the initial content of hydrogen in the tank
at t=t0. A time horizon of this considered. If it is
small enough, the hydrogen consumption rate, m
H(t), can
be assumed to be constant due to the slow dynamics of the
downstream hydrogen consumption. With such assumption,
(13) can be simplified and combined with (12) to arrive at:
Vmin
HV0
H+m
H(tht0)Zth
t0
mH(t)dt
Vmax
HV0
H+m
H(tht0)
(14)
Referring to (10), and calculating the Taylor approximation
of (11), the integration of mH(t)in (14) can be expressed as:
Zth
t0
mH(t)dt
Nsα1
2FZth
t0Idc
h(t) + α3
Idc
h(t)dt +α2(tht0)
(15)
In large-scale HE applications, DC current Idc
his significantly
larger than α3in (15). Hence, the second term (α3/Idc
h) under
integral is much smaller than the first term (Idc
h), which can
be ignored. Replacing (15) into (14), a constraint on hydrogen
production with respect to stack’s current is defined as:
Smin Zth
t0
Idc
h(t)dt Smax (16)
where the maximum and minimum are calculated as,
Smin =2F
Nsα1
(Vmin
HV0
H+m
H(tht0)) α2(tht0)
Smax =2F
Nsα1
(Vmax
HV0
H+m
H(tht0)) α2(tht0)
During the normal operation of HE, when there is no internal
5
fault, Smin and Smax can be regarded as the upper and lower
limits for the flexible operation of the HE, i.e., if the constraint
(16) holds, the electrolyzer can operate based on the GFM law
and flexibly varies its active power depending on the frequency
conditions in the grid. Variations in the active power lead to the
changes in the DC current of HE, Idc
h, which directly impacts
the efficiency (ηFas given by (11)) and temperature of the
PEM stack. As explained earlier (please also see Fig. 2), higher
Idc
hcauses higher losses in the stack (lower efficiency). The
losses appear as heat in the stack which is calculated as
qh=Idc
hVdc
h(1 ηF)(17)
where qhis the generated heat (in Watts) in the stack. Depend-
ing on the stack’s material and dimension, such generated heat
can change the operation temperature of the stack [19].
III. PROP OS ED HE CONTROL: OP ER ATION MODE S AN D
LIMITATIONS
In this section, a control system for the PEM HE is sug-
gested. It has three control modes, each activated depending
on the HE operational constraints in hydrogen production.
A. Control Principles
The proposed control structure for the PEM HE is shown
in Fig. 1. It consists of the HE VSC’s cascaded control with
the proposed three operation modes and the DC-link voltage
control of a DC-DC converter. The following major control
blocks are used in this control scheme:
Current control: it regulates the HE VSC’s output current,
iqd
s, at the references provided by the outer AC voltage
control. The output of the current control loops, vqd0
m, is
transformed to the abc-frame via angle θcto build the
modulation signal. The time response of the current control
is commonly a few milliseconds.
AC voltage control: the outer AC voltage control regulates
HE VSC’s output voltage, vqd
c, at the references which are
generated by the reactive power droop. The time response
of the voltage control is at least ten times slower than that
of the current control.
Reactive power droop: a droop gain, kq, is used to generate
voltage reference in q-axis, vq
c, from the reactive power.
The voltage reference in d-axis is set to zero. The peak of
the base AC voltage is given by vn, which comes from a
secondary controller. The reactive power-voltage droop is
expressed by
vq
c=kq(Q
cGq(s)Qc) + vn(18)
where Gq(s)=1/(τqs+ 1) is a low-pass filter with a time
constant of τqused to regulates the speed of voltage support.
DC voltage control: depending on the control mode of the
HE, the DC-link voltage, Vdc
t, is tightly regulated at the
nominal value via either (i) DC voltage control loops of the
DC-DC converter (operation Modes 1 and 3) or (ii) the DC
voltage control of the HE VSC (operation Mode 2). At any
given time, only one of these two controls is activated. The
mentioned operating modes are discussed in details in the
next section.
Secondary control: it defines the steady-state operation of
the electrolyzer as given by the active and reactive power
references, Pref and Qref , respectively. The secondary
controller uses a very slow PI controller [25], commonly
with a response time of few minutes, so it does not affect
the primary response from the electrolyzer. If the secondary
controller is activated, the power consumption of the elec-
trolyzer always goes back to Pref with a very slow time
constant. The load management control division provides
active power reference for the secondary control.
Load management: located between HE local control and the
HE VSC’s secondary control, it ensures that the operation
of HE VSC does not violate the limitations related to the
hydrogen production process. More specifically, it monitors
whether the constraint given by (16) holds or not. Based
on the outcome, one of the HE VSC’s operation modes
(Modes 1, 2, 3) are selected, which are explained in the
next section.
HE local control: this is the manufacturer’s controller that
locally regulates HE plant. It receives local signals, such
as H2tank level, tank pressure, stack temperature, H2
production and consumption flow rates, etc., and generates
valve’s positions, pumps and fans’ speeds, etc. In case of
any failures during HE operation, an alarm/trip signal is sent
to the load management control block.
B. Modes of Operation
Referring to (16), for every time horizon of th, the upper and
lower limits of hydrogen production process, Smax and Smin
given by (16), respectively, are calculated. If the integration
of Idc
h(t)violates either of the limits during this time horizon,
the HE VSC’s operation mode will change to ensure a safe
operation of the electrolyzer. Three operation modes are pro-
posed for the HE VSC, and the load management control block
is in charge of choosing the appropriate one. As illustrated
in Fig. 1, all three modes are related to the specific ways of
generating frequency deviation needed to determine the power
consumption of the HE.
1) Mode 1: this is the grid-forming operation mode of the
HE VSC, and the HE plant can be regarded as a grid-forming
load. The integration of Idc
h(t)is within the constraint given
by (16), and no trip/alarm signals are sent from HE local
control to the load management block. The DC-link voltage
is regulated by the control loops of the DC-DC converter,
and the HE VSC is allowed to operate fully according to
the GFM law, i.e., the power consumption of HE is flexibly
varied according to the variations in the grid’s frequency. A
droop gain, kp, is used to link HE VSC’s active power to the
frequency deviation. The output frequency deviation is added
to the based frequency, ωn, to build the operating frequency of
the converter, ω
c. A low-pass filter is placed on the feedback
from the measured active power to emulate virtual inertia in
the control system. The droop equation is given as
ω
c=kp(P
cGp(s)Pc) + ωn(19)
where Gp(s)=1/(τps+1) is a low-pass filter with time constant
of τp.
6
2) Mode 2: in this mode, the hydrogen production process
is experiencing an abnormal condition. A possibility is that
the H2tank is nearly full and it cannot store any further
hydrogen. Hence, the integration of Idc
h(t)violates Smax in
(16), and therefore the operation Mode 2 is activated. Another
possibility is that a trip related to a component failure in the
hydrogen production process, such as a pump, compressor,
dryer, etc., or abnormal stack temperature and pressure, is
triggered. In such conditions, the HE needs to be shut down
and remained offline for certain duration until it is ready for
normal operation again. During this period, the HE and the
DC-DC converter are disconnected by the circuit breaker (CB)
shown in Fig. 1. As the DC-DC converter is no longer available
for DC-link voltage control, the HE VSC’s control fulfils such
control objective by the activation of Mode 2. In this mode, a
dedicated PI controller, P Iv, ensures a tight DC-link voltage
regulation. Although the HE VSC could be disconnected
during this period too, it is suggested keeping it connected to
the grid to provide AC voltage regulation services. A limited
amount of active power is absorbed from the grid only to
charge the DC-link capacitor and cover converter’s losses. The
operation Mode 2 cannot offer grid-forming services as in
Mode 1 due to its limited ability in varying its active power
consumption.
3) Mode 3: when the hydrogen content in the tank is
critically low, and the downstream hydrogen demand may
not be fulfilled given the current hydrogen production rate,
the HE VSC can switch to Mode 3 to operate in constant
power mode, and steadily absorbs high power from the grid.
In this operation mode, the HE plant does not flexibly adjust
its power consumption to provide frequency support to the
grid. The operation Mode 3 is activated automatically in case
the integration of Idc
h(t)violates Smin in (16). In this operation
mode, a PI controller, P Ip, is used to track the active power
reference given by P
m.
In summary, if the internal operation of the HE plant is
normal and no constraints are violated, the HE is allowed
to operate in Mode 1 (grid-forming mode). However, if the
load management block detects any abnormal condition in
the HE plant, it automatically switches to either operation
Mode 2 or 3 (non grid-forming modes) depending on the
type of the abnormality. However, such mode changing can
have a negative impact on the grid’s frequency as it causes a
change in the power consumption of the electrolyzer. Hence,
if it is requested by the grid operator, a frequency condition
criterion can be added to the mode selection logic in the load
management block to only switch modes during normal grid’s
frequency conditions (i.e., the grid frequency is experiencing
no major over or under frequency events).
C. Control Tuning
The PI gains of the inner current control are determined
based on the filter impedance and desired time constant of
the loop. The proportional (kpc) and integral (kic ) gains are
defined as
kpc =Lf
τs
, kic =Rf
τs
(20)
where Lfand Rfare the filter inductance and parasitic
resistance, and τsis the desired time constant, typically in
order of a few milliseconds. The PI controller of the outer AC
voltage control loops can be defined as
kpv = 2CfDvωv, kiv =ω2
vCf(21)
where Dvis the damping, commonly 0.707, and ωvis the
bandwidth of the voltage control loop, which is defined to be
at least ten times slower than the inner current loops.
The tunings of the droop gains in the active and reactive
power loops, kpand kq, respectively, are based on the maxi-
mum permissible deviations in the frequency and voltage,
kp= 0.02 ωn/Sn, kq= 0.1Vn/Sn(22)
where ωn,Vn, and Snare the nominal frequency, voltage,
and power of the VSC, respectively. Here, it is assumed that a
maximum of 2% of frequency and 10% of voltage deviations
are allowed during transients.
The response time of the power-frequency droop in Mode 1
(grid-forming mode) is determined by the time constant of the
low-pass filter, τp, and the droop gain, kp. As discussed in
[26], such response time is equivalent to the mechanical time
constant that represents rotor inertia and calculated as
Tvi =τp
1
kp
(23)
where Tvi is the equivalent time constant, and kpis expressed
in per unit value (0.02 in this study). The response time of the
grid-forming control (Tvi) should be slower than the response
time of the electrolyzer (ts) to make sure that the electrolyzer
has enough time to respond to the power changes. Hence, the
following constraint is assumed:
Tvi > ts(24)
Based on (23) and (24), the cut-off frequency of the low-
pass filter Gp(s), which is given by (19), should satisfy the
following requirement:
τp> kpts(25)
For the tunings of P Ivin Mode 2 and P Ipin Mode 3,
a heuristic method is used. Since the outputs of all three
operation modes are the frequency of HE VSC, they can be
tuned to have a similar response time. Therefore, P Ivand P Ip
are tuned to have a similar response time as in Mode 1.
D. Fault Operation
The HE VSC has to be protected during faults in the AC
grid, as they may cause frequency deviation and overcurrent
condition. To this aim, a rate limiter is placed on the frequency
reference to limit the rate-of-change-of-frequency (ROCOF),
and a qd-saturation block is placed on the current reference to
protect the HE VSC from overcurrent conditions.
During fault condition in the AC grid, the PCC voltage (vabc
c
in Fig. 1) dips to a low value. The PI controller of the AC
voltage loop tries to bring the PCC voltage back to its nominal
value by increasing the current reference. Note that the HE
VSC does not inject any active current to the grid since it is
a load by nature and it can only contribute reactive current.
7
q-axis
*
c
is
*
In
q
In
d
d-axis
In
-axis
-axis
c
is
d*
is
q*
is
q*
qd-saturation
AC voltage control
In
d
In
q
AC current control
is
d*
Fig. 4. Current saturation in qd-frame
TABLE I
SYS TEM PA RAM ET ER S
Parameter Value
—Upstream AC Grid—
Base RMS voltage, Un132 kV
Base power, Sn50 MVA
Base frequency, fn50 Hz
—Power Transformer—
short-circuit impedance 10%
HV/MV rated voltage 132/22 kV
—Voltage Source Converter
output L filter, Lf5.09 mH
output C filter, Cf67.66 µF
maximum current rating, In1.2 pu
current loop PI controller 5.09+80/s
AC voltage loop PI controller 0.27+0.94/s
DC voltage loop PI controller 3.14e-04+1.57e-03/s
Active Power PI controller 1.25e-06+1.13e-05/s
power-frequency droop, kp0.02 pu
reactive power droop, kq0.1 pu
active power low-pass filter, Gp(s)1/(1+0.04s)
reactive power low-pass filter, Gq(s)1/(1+0.04s)
—DC/DC Converter—
DC voltage loop PI controller 7.5e-05+7.5e-04/s
DC current loop PI controller 9.3e-04+2.4e-03/s
—PEM Electrolyzer—
hydrogen tank capacity, V
H9000 Moles
hydrogen production rate, mH10 Moles/MW/s
reverse reference voltage, Vrev0400 V
double layer phenomenon, Cop 0.14 µF
double layer phenomenon, Rm0.01
Faraday efficiency constants, α1,α2,α396.5, 0.09, 75.5
However, such reactive current contribution may exceed the
converter’s current rating. Hence, the qd-current saturation
block is used to limit the magnitude of the VSC’s current
reference, iqd
s, while keeping its angle constant in qd-axis.
The upper and lower saturation limits are set dynamically.
Assuming the rated current of the VSC is In, and referring to
Fig. 4,
Iq
n=In|sin ϕ|,Id
n=In|cos ϕ|(26)
where ϕis the angle of the current with respect to d-axis and
simply calculated as ϕ= arctan iq
s
id
sThe positive values of
(26) define the upper limits of current, and the negative values
set the lower limits. An anti-windup scheme is necessary
to avoid integrator windup during current saturation. In this
study, a conventional back-calculation anti-windup method is
adopted [27].
IV. SIMULATION RESULTS
The performance of the PEM HE is tested in the power
system shown in Fig. 5. Such test system is inspired from
the Dalrymple substation located in Yorke Peninsula, Aus-
tralia [28]; however, the original battery energy storage system
at Dalrymple substation is replaced by the PEM HE and the
total grid inertia and short-circuit ratio (SCR) are reduced to,
respectively, 2 s and 3 in order to represent a low-inertia,
weak network. The details related to the upstream AC grid
are extracted from [29], [30].
The main parameters of the test system are given in Ta-
ble (I). The VSC and DC-DC converter are simulated by their
average models, and the DC-DC converter topology is based
on the full-bridge structure. It is considered that the wind
farm operates in GFL mode and tracks its maximum power
point reference. The PEM HE is represented by its equivalent
electrical circuit as shown in Fig. 3. It is assumed that the
response time of HE is 600 ms (please see Appendix) based
on which the electrical elements of the equivalent model is
calculated and given in Table (I). The rated capacity of the
H2tank storage, V
H, is 9000 Moles. The hydrogen content in
the tank is critically low if it hits the bottom 10% capacity
(Vmin
H=0.1V
H), and the tank is full if it reaches to its 100%
capacity (Vmax
H). The downstream dynamics of the hydrogen
consumption is commonly very slow; hence, it is assumed that
m
His constant in the next case studies. The following cases
are considered:
Case study 1: HE ability to flexibly adjust its power
consumption, and therefore hydrogen production rate, in
response to the variations in the wind power is studied.
Also, the impact of power variations on the temperature and
efficiency of HE stack is demonstrated.
Case study 2: potential and limitations of three operation
modes of HE (as discussed in Section III-B) in response
to an under-frequency event in the upstream AC grid are
investigated.
Case study 3: HE ability to automatically switch among dif-
ferent operation modes, depending on its internal limitations
and grid conditions is studied.
Case study 4: performances of HE and battery system while
both operate in grid-forming mode are compared.
A. Case Study 1
In this case study, the power generated from the wind, Pw,
varies with a ramp rate of 0.25 pu/s. It is first increased from
0.5 pu to 0.75 pu, then reduced again to 0.5 pu. After 2 s,
it is further reduced to 0.25 pu. Finally, it is brought back to
its initial operation point (0.5 pu). The constant PQ load has
a fixed power consumption of Pl=0.23 pu, and the hydrogen
consumption rate is fixed at m
H=30 Moles/s for the entire
period. First, it is considered that the HE is allowed to operate
in Mode 1 (grid-forming mode) for the entire period, i.e., HE
operates as a GFM load. It is assumed that the hydrogen
production constraint as given by (16) is not violated and
the hydrogen content in the tank is within the limits. As
illustrated in Fig. 6(a), the power consumption of the HE,
Pc, flexibly follows the changes in the wind power based on
8
25 km 66 km
CB1
CB2
CB4
132kV
22kV
fixed PQ
load
hydrogen
electrolyzer
CB3
Pw
PlPc
Pg
Wattle Point
Wind Farm Dalrymple
Substation
Ardrossan West
Substation
Hummocks
Substation
Kadina East
Substation
Snowtown
Substation
Bungama
Substation
To the rest
of network 50 km
34 km
77 km
42 km
22.6 MW
1.3 MVAr
15 MW
4 MVAr
7 MW
1.4 MVAr
Upstream
AC Grid
Fig. 5. Single line diagram of the test system
the GFM law: it absorbs more power when there is higher
power availability from the wind. Hence, the locally generated
renewable power is consumed locally by the HE and fixed
load, and the upstream AC grid supplies the additional power
(Pg). The changes in Pcreflects on the H2production rate,
mH, which is shown in Fig. 6(d). mHvaries from about 170
to nearly 250 Moles/s when the wind power reaches to its
maximum value.
Opposite to this operation mode, the HE can operate as
a constant power load, which is the operation Mode 3. This
operation mode is activated in case that the lower limit on
hydrogen content in the tank given by (16) is violated, so
the HE operates in fixed power to fill the tank faster. Note
that Mode 3 is activated manually in this case study only
to compare performances of this operation mode with that
of Mode 1. As shown in Fig. 6(b), regardless of the power
availability from the wind or grid, the HE keeps consuming a
constant power of 0.35 pu (operation Mode 3), and therefore,
mHis also constant at 170 Moles/s.
Comparing the frequency responses from the operation
Mode 1 and 3, the power consumption flexibility provided
by Mode 1 results in a smaller frequency deviation as shown
in Fig. 6(c). This is resulted from the ability of HE to change
its hydrogen consumption rate (mH) quickly. However, such
advantage is only secured if the capacity of H2storage is
within the range [Vmin
H,Vmax
H], i.e., the storage is neither crit-
ically low, nor full. During this period, the level of hydrogen
in storage reaches to about 60%, which is shown in Fig. 6(e).
The impacts of operation Mode 1 and 3 on the PEM stack’s
efficiency are compared in Fig. 6(f). As explained by (11),
the stack’s efficiency is a function of the stack’s current.
In operation Mode 1 (GFM mode), the stack’s power varies
based on the GFM control, leading to the variations in the
stack current, which in turn changes the stack’s efficiency
(please also see Fig. 2). On the contrary, if the HE operates
in Mode 3 (fixed power), the stack’s current, and therefore
stack efficiency, remains constant during operation as long as
the power set-point is not changed. Similar situation exists
(a)
(b)
(c)
(d)
(e)
(g)
(f)
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
55
60
65
70
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
80
85
90
95
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
30
40
50
60
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
100
150
200
250
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
49.4
49.6
49.8
50
50.2
50.4
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
-0.2
0
0.2
0.4
0.6
0.8
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
-0.2
0
0.2
0.4
0.6
0.8
Mode 1 Mode 3
Mode 1 Mode 3
Mode 1 Mode 3
Mode 1 Mode 3
Mode 1 Mode 3
Fig. 6. HE’s response to the wind power variations, (a) active powers for
HE operating in Mode 1, (b) active powers for HE operating in Mode 3, (c)
system frequency, (d) hydrogen production rate, (e) hydrogen content in the
tank, (f) stack efficiency, and (g) stack temperature
for the stack’s temperature. For the operation Mode 1, the
stack’s temperature varies according to the changes in the
stack’s power. Higher power consumption increases stack’s
temperature since part of this power (depending on the stack’s
efficiency) is dissipated as heat inside the stack. The stack’s
temperature is constant in operation Mode 3 as illustrated
9
4 5 6 7 8
49.2
49.4
49.6
49.8
50
50.2
45678
-0.2
0
0.2
0.4
0.6
45678
0
0.2
0.4
0.6
45678
0
100
200
300
4 5 6 7 8
-0.2
0
0.2
0.4
0.6
45678
-0.06
-0.04
-0.02
0
0.02
4 5 6 7 8
1
1.01
1.02
1.03
1.04
45678
0.9
0.95
1
45678
0
0.2
0.4
0.6
0.8
45678
50
60
70
Mode 1
Mode 2
Mode 3
VH
mH
Fig. 7. HE’s response to an under frequency event in the upstream AC grid,
comparing performances of Mode 1, Mode 2, and Mode 3
in Fig. 6(g).
This case study thus indicates that a HE which operates
in GFM mode can provide frequency regulation and flexibly
adjust its power consumption according to the availability of
power generation in the network. However, such HE operation
may require further control and arrangement to keep HE’s
temperature and efficiency within an acceptable range.
B. Case Study 2
In this case study, an under-frequency event occurs in the
upstream AC grid at t=4 s. The dynamic behaviour of HE in
the proposed operation modes is discussed. If the HE operates
in Mode 1 (grid-forming control) prior to the under-frequency
event, it reduces its power consumption to avoid large fre-
quency deviation as shown in Fig. 7. Pcis reduced (based on
droop fucntion) from 0.5 pu to 0.24 pu, so more power from
the wind is injected to the grid, helping with the frequency
recovery. Accordingly, mHreduces from 250 to 120 Moles/s.
Operation Mode 3 (constant power control) is indifferent to the
under-frequency event, and keeps consuming Pc=0.5 pu for
the production of mH=250 Moles/s. The frequency deviation
is larger in this mode.
Different from Mode 1 and 3, in the operation Mode 2
(DC voltage control), the HE is shut down (mH=0 Moles/s)
and disconnected (due to internal fault) prior to the event.
Only the HE VSC remains connected to the grid for the AC
voltage regulations. Such operation mode cannot help with the
frequency response as the HE power consumption is zero.
Regardless of the operation modes, DC-link voltage, Vdc
t,
and AC voltage, Vc, remained regulated within the permissible
range as it can be seen from Fig. 7. The current flowing
into the HE, Idc
h, is zero during operation Mode 2 since the
HE is shut down and disconnected, and Vdc
hremains at the
open-circuit voltage level. The initial hydrogen content in the
tank is V0
H=50% at t=0 s. While it remains unchanged during
operation Mode 2, it increases during operation Mode 1 and
3 with different rates.
C. Case Study 3
In this case study, the HE ability to transiently switch among
different operation modes, depending on its internal limitations
and grid operation, is explored. The operation modes (1,2,
and 3) are automatically selected by calculating the hydrogen
production constraint given by (16) in real time. The timeline
of the events is shown on the top of Fig. 8. Initially, the HE
operates in Mode 1 (GFM mode), and the initial hydrogen
content in the tank is only 13%. The hydrogen consumption
rate (m
H=250 Moles/s) is higher than its production rate
(mH=230 Moles/s), so the hydrogen in the tank is reduced
over time. The condition that the HE is allowed to operate in
Mode 1 is explained by (16), which is plotted in Fig. 8(c).
The green curve is the integral of Idc
hfor the time horizon
of 0.5 s. After each time horizon, the integrator is reset. At
t=6 s, Smin (red line) hits the peak of the green line, meaning
that the operation Mode 1 is no longer feasible since the
storage is critically low. Hence, the operation mode of HE
is automatically switched to Mode 3 in order to increase the
power consumption of the HE, and therefore, store hydrogen
at a higher rate (mH=400 Moles/s) to fill the storage faster
(see Fig. 8(e)).
After few seconds, an internal fault occurs inside the HE at
t=10 s, causing the plant to shut down. However, the VSC
remains at the service by switching its operation mode to
Mode 2. Since the plant is shut down, mH=m
H=0. The HE
VSC only regulates Vdc
tand Vcduring this period.
The internal fault is cleared and the HE is back to operation
after 2 seconds. Once the HE is back online, and Smin and Smax
are not violated, it switches to operation Mode 1 again. Subse-
quently, a three-phase fault occurs in the Dalrymple substation
at t=18 s and cleared after 800 ms. During fault period, voltage
dips to about 0.2 pu causing the power consumption of HE
to drop to zero. However, the HE VSC remains connected to
the faulty substation to provide voltage support by injecting
reactive power, Qc. As it can be seen from Fig. 8(h), the
HE VSC’s full current capacity (1.2 pu) is used for reactive
power injection. The fault performance shown in this case
study is related to operation Mode 1. It is to note that the
fault performance of HE depends on its operation mode. In
particular, the post-fault performance of operation Mode 2 and
Mode 3 differs from that of Mode 1 since they have different
outer control structures. Finally, a system separation occurs
10
Fig. 8. Transient responses of HE while switching among different modes
during various internal and external events, (a) frequency, (b) active power,
(c) hydrogen constraint, (d) hydrogen content, (e) hydrogen production rate,
(f) AC voltage, (g) reactive power, and (h) stack current
(a)
(b)
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
0
0.2
0.4
0.6
0.8
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
49
49.5
50
Pc
Fig. 9. Impact of secondary controller on the HE GFM operation, (a) active
power, and (b) frequency
at t=24 s, and the upstream grid is disconnected (CB1 is
open). The islanded system consisting of the HE, wind energy,
and the fixed load, continues stable operation since the HE,
operating in GFM mode, is forming the voltage and frequency
of the islanded system.
The GFM HE (operation Mode 1) adjusts its power con-
sumption to reduce the frequency deviation in the network.
Thus, the power consumption of the HE depends on the power
imbalance in the upstream grid. The secondary controller
shown in Fig. 1 can ensure that the HE power consumption
goes slowly back to a default value (Pref ) after the primary
frequency response. It is to note that P
cis the power reference
for droop control, while Pref is a power reference to a
slow PI controller which is located in the secondary control.
Considering the case study 3 and assuming that the wind
power reduces from 0.75 pu at t=7 s to 0.25 pu with a
ramp rate of 0.25 pu/s, the impact of the slow secondary
controller on the HE GFM operation is shown in Fig. 9. The
GFM HE primary response (first few seconds) depends on
the droop control, while the steady-state operation (with a
settling time of few minutes) relies on the secondary controller.
Here it is assumed that Pref and P
care equal; however, they
can be different. The activation of secondary controller can
ensure a default steady-state power consumption (Pref ) can
be reached after disturbances, but at the cost of a larger steady-
state frequency deviation (see Fig. 9(b)). If the secondary
controller is deactivated, the steady-state power consumption
of HE differs from Pref and depends on the power imbalance
in the grid, which results in a smaller frequency deviation. The
deactivation of the secondary controller is more consistent with
the grid-forming functionality as it results in smaller frequency
deviation; however, both options are available in the suggested
control scheme for selection.
This case study illustrates a great HE potential in providing
grid-forming services to the network even during islanded
operation. Therefore, PEM HE can be considered as a potential
rival for other technologies capable of offering grid-forming
provisions (e.g., synchronous machine).
11
(a)
(b)
4 6 8 10 12 14
-0.2
0
0.2
0.4
0.6
0.8
1
4 6 8 10 12 14
-0.2
0
0.2
0.4
0.6
0.8
1
Pc
Pbat
Fig. 10. Comparison between performances of electrolyzer and battery
system, (a) system with electrolyzer, and (b) system with battery system
D. Case Study 4
The Dalrymple substation originally has a battery storage
system which has been replaced by the HE in the previous case
studies. In this final case study, a comparison between PEM
HE and a battery storage system in terms of their capability
in providing grid-forming services is presented.
Two scenarios are briefly studied. In the first scenario, the
previous test system with the electrolyzer is considered where
constant PQ load consumes Pl=0.35 pu and the wind power
gradually reduces from Pw=0.8 pu to 0.4 pu with a ramp rate
of 0.2 pu/s. At t=6 s, the upstream grid is disconnected, and
electrolyzer continues to form the grid (operation Mode 1).
During the period of 8 to 14 s, the generated power from the
wind is not sufficient to meet the load demand (constant PQ
load and electrolyzer); hence, the electrolyzer automatically
reduces its power consumption to obtain the required power
balance in the system as shown in Fig. 10(a). Therefore, the
required grid-forming services are supplied from the load side
(electrolyzer).
In the second scenario, the electrolyzer is replaced by a
battery system of similar size. The constant load consumes
Pl=0.92 pu and the wind power follows similar pattern as
the previous scenario. As it can be seen from Fig. 10(b),
the wind power is not sufficient to supply the constant PQ
load; therefore, the power generation of battery system, Pbat,
gradually increases to compensate the reduction in the wind
power. Hence, the required grid-forming capability is on the
source side (battery system).
From the simulation results, it can be concluded that both
electrolyzer and battery system can provide grid-forming ser-
vices to the grid, even after system separation. Hence, in terms
of grid-forming functionality, both technologies have fast
response and are capable of providing such services. However,
hydrogen has many uses in transportation and industry sectors
in addition to be a great power-to-gas energy storage solution.
V. CO NT ROL HAR DWARE-IN-T HE -LO OP TE ST S
In order to verify the robustness and effectiveness of the
proposed concept, the HE control system is embedded and
vm
abc
dm
Lf
Cf
is
abc ic
abc
+
-
AC Grid
Typhoon HIL 602+
vc
abc
Vt
dc
It
dc
+
-
HE
model
+
g
*
-
+
-
SCR=3
g
kdroop Tr(s)
Pm
*
+
-
Pe
2Hs
1
vg
abc*
g
control system
signal processing 11 AI
MCU F28335 measurment unit
4 DI
Fig. 11. Test system simulated in real-time
tested in hardware running in real-time. The algorithm is
embedded in a Texas Instruments Microcontroller Unit (MCU)
F28335 evaluation board and the power grid is simulated in
real-time in a Typhoon HIL 602+ device. The digital control
algorithm samples signals at the rate of 20 kHz. The system
simulated in real-time is shown in Fig. 11. It is composed
of a PEM hydrogen electrolyzer with the proposed control
system and an equivalent representation of the upstream AC
grid with dynamics given by a simplified turbine and governor
[31]. The governor model is based on a droop gain as given
by kdroop = 15.7, and the turbine dynamic is simulated by
the transfer function Tr(s)=khp +klp
τrhs+1 whose parameters
are khp = 0.3,klp = 0.7,τrh = 0.1s. The equivalent
inertia and SC R of the grid are 2 s, and 3, respectively. The
converter nominal voltage and power are 22 kV and 50 MVA,
respectively.
A. Experimental setup
The experimental setup is shown in Fig. 12. Eleven analog
signals (AI), including vabc
c,iabc
c,iabc
s,Vdc
tand Idc
tas indicated
in Fig. 1, and four digital signals (DI) related to the operating
mode selection (OP1,OP2,OP3and Enable) are sent by
the Typhoon HIL to the MCU. The analog signals are scaled
between 0–3 V before sending to the MCU. The first three dig-
ital signals indicate the three operation modes as explained in
Section III-B, and the last digital signal enables the controller.
The MCU outputs are the three-phase voltage of converter,
vabc
m, and dm, as highlighted in Fig. 1. The output three-phase
signals are modulated by a PWM at 10 kHz and filtered by
the passive first-order anti-aliasing filters before sending them
back to Typhoon HIL. The filter parameters are: R= 1000 ,
C= 0.1µF with a cut-off frequency of 1591 Hz.
B. Experimental results
Two test cases are discussed. The purpose of the first test
case is to confirm the ability of the proposed HE controller
to dynamically switch among three operation modes in real-
time. As it is presented in Fig. 13, the HE operates in Mode 1
in the period of 0 to 1.5 s. Subsequently, HE switches to
operation Mode 3 and 2 at 1.5 s and 3 s, respectively. The
operation modes are communicated between MCU and real-
time simulator via OP digital signals. Referring to Fig. 13, it
can be observed that the transition among different operation
modes are preformed smoothly and without a major transient
12
Fig. 12. Control hardware-in-the-loop setup
Fig. 13. Test case 1: (a) vabc
moutput from the MCU, (b) operation modes
sent to the MCU, (c) HE active and reactive power, (d) system frequency, (e)
HE current, and (f) HE DC voltage
or unwanted dynamics, which is consistent with the results
obtained from simulation.
The second test case focuses on the HE performance in
operation Mode 1 (GFM mode) during frequency events in
the upstream grid. An under-frequency event occurs at t=1.5 s
which is followed by an over-frequency event at t=4 s. As it is
shown in Fig. 14, the HE reacts to those frequency events by
decreasing and increasing its power consumption, respectively.
The over-frequency event causes an increase of 0.75 Hz in the
system frequency. The HE immediately increases its power
consumption in an attempt to reduce frequency deviation.
However, the HE power consumption reaches to about 1 pu
that is the maximum frequency support that the HE can
Fig. 14. Test case 2: (a) vabc
moutput from the MCU, (b) HE active and
reactive power, (c) system frequency, (d) HE current, (e) HE efficiency, (f)
HE temperature, and (g) hydrogen content in the tank
provide for the grid. During this frequency excursion, the
stack temperature and efficiency also vary as shown in Fig. 14.
Although different systems have been analyzed in simulation
and in real-time, the results in real-time confirm that the HE
can successfully operate in GFM mode and switch stably
among all operation modes. Such stable operation and smooth
transition among modes can be observed from Fig. 13 and
Fig. 14.
VI. CONCLUSION
This study proposed a control system for a PEM hydro-
gen electrolyzer, which consists of three operation modes:
grid-forming mode (Mode 1), DC voltage mode (Mode 2),
and constant power mode (Mode 3). The activation of each
operation mode was carried out automatically based on the
hydrogen production constraint. The operation Mode 1 (grid-
forming) was selected if the upper and lower limits given
by the hydrogen production constraint were not violated, i.e.,
there is sufficient headroom/footroom in hydrogen storage
tank. So, the electrolyzer could fully participate in forming
the voltage and frequency of the grid. If the upper limit of the
constraint was violated (i.e., the hydrogen tank was nearly full)
or an internal failure has occurred in the electrolyzer, operation
Mode 2 was activated. In this mode, the electrolyzer has to be
shut down for safety, but its VSC could remain connected
to the AC grid to provide voltage regulation services. In
case the lower limit of the constraint was violated (i.e., the
hydrogen content in the tank was critically low), the operation
Mode 3 was activated. In this mode, the VSC was operating
13
step 1
recording
HE response
to power changes
step 2
measuring
ts and tr
from response
step 3
calculating
H(s) given by (1)
step 4
mapping H(s)
to equivalent
electrical circuit
shown in Fig. 3
step 5
evaluating
equivalent model
response
step 6
scaling up
the model
Fig. 15. Steps for HE equivalent model validation
in constant power mode and extracting maximum power from
the grid to quickly fill the hydrogen tank in order to avoid
interruption in hydrogen supply for the downstream process.
The PEM electrolyzer with the proposed control system has
been tested on a low-inertia grid whose topology is inspired
by a portion of the transmission grid in South Australia.
The following points can be concluded from the simulation
results: (i) an electrolyzer has a great potential to provide grid-
forming services (frequency and voltage regulations) to the
network, (ii) with the tandem operation with a nearby wind
generation system, a grid-forming electrolyzer can form the
voltage and frequency of an islanded network, (iii) the grid-
forming operation of an electrolyzer is only allowed if the
hydrogen production constraints are not violated; in case of
violation, the operation mode of electrolyzer should switch to
either DC voltage mode or constant power mode depending
on the hydrogen production conditions, and (vi) electrolyzer
can have the same grid-forming capability as a battery storage
system. The difference is that such services can come from the
load side (electrolyzer), or source side (battery system). The
results from the tests in hardware confirm that the proposed
concept can be successfully implemented in practice.
APPENDIX
ELE CT ROLYZE R MOD EL VALIDATION
In order to evaluate the accuracy of the HE equivalent
electrical model which is shown in Fig. 3, its response is
compared with the experimental results obtained in [2]. The
steps that are followed for the HE equivalent model validation
are shown in Fig. 15. In the step 1, the power consumption
of HE is varied and the DC current is recorded [2]. The tests
are conducted on a 40 kW PEM electrolyzer, so the equivalent
model is validated for this rated power but later it is scaled up
for a MW-scale PEM HE that is used in the case studies. As
discussed in Section II.C, only the rise time (tr) and settling
time (ts) are needed to identify the model of a HE. Such
information is extracted from the responses in the step 2.
Please note that the HE has a first-order response to power
variations [2], [23]. Having trand ts, the transfer function
H(s), given by (1), is calculated in the step 3. Next, H(s)is
mapped to an equivalent electrical circuit (shown in Fig. 3) in
the step 4. In the step 5, the similar tests as in [2] are carried
out on the HE equivalent circuit to validate the accuracy of the
model. Once the equivalent model is validated for a 40 kW
0.95 1 1.05 1.1 1.15 1.2 1.25
0
0.2
0.4
0.6
0.8
1
0.2
Fig. 16. Response evaluation of the HE equivalent model
electrolyzer, the model is scaled up for the high-power HE
used in this study.
The experimental results on a 40 kW PEM HE are compared
with the results obtained from the equivalent electrical circuit
in Fig. 16. The trigger signal is used to change the DC power
consumption of HE. The power consumption is changed (i)
from 25% to 100%, (ii) from 50% to 100%, and (iii) from 75%
to 100%, and the DC current of HE is plotted for every cases.
As it can be seen, the responses of the equivalent electrical
circuit and the experimental results have a good matching. The
DC current reaches to its reference value after about 150 ms
in all cases. However, this evaluation is valid for a 40 kW HE
while we use MW-scale HE in our study. Hence, we need to
scale-up the equivalent circuit to have realistic response for a
MW-scale electrolyzer.
Here we need to make some assumptions. Referring to [32],
a MW-scale HE consists of arrays of HE modules. Each HE
module can have the nominal power of 150-250 kW. Then,
such modules are connected in a form of an array to scale
up HE’s power. From the test results on a 40 kW HE, we
can conclude that the HE can vary its power consumption
from 0 to 40 kW within 150 ms. So, we may assume that
each HE module of 150-250 kW is able to change its power
consumption from 0 to the rated value within the range of
400-700 ms. In this study, we use 600 ms as the settling time
of each HE module.
Since a MW-scale HE is built by an array of HE modules,
we assume here that the HE modules have similar DC rated
voltage and they are connected in parallel to the main DC bus.
Hence, the settling time of each module (600 ms) is in fact the
settling time of the entire HE arrays. By such assumptions, we
re-calculate the HE transfer function, H(s), based on the new
settling time (600 ms) and map it to the equivalent electrical
circuit. Please note that the response time of the HE is an
input parameter to the equivalent model calculation. So, any
response time, which could be faster or slower than 600 ms,
can be used to derive the equivalent model.
REFERENCES
[1] IRENA, “Green hydrogen cost reduction,
2020. [Online]. Available: https://www.irena.org/-
/media/Files/IRENA/Agency/Publication/2020/Dec/IRENA Green -
hydrogen cost 2020.pdf
[2] J. Eichman, K. Harrison, and M. Peters, “Novel electrolyzer
applications: Providing more than just hydrogen,” NREL, Tech. Rep.,
9 2014. [Online]. Available: https://www.osti.gov/biblio/1159377
[3] Gigastack, “Bulk supply of renewable hydrogen, Ele-
ment Energy, Tech. Rep., February 2020. [Online]. Avail-
able: https://gigastack.co.uk/content/uploads/2020/06/Gigastack-Phase-
1-Public-Summary.pdf
14
[4] J. J. Caparr´
os Mancera, F. Segura Manzano, J. M. And´
ujar, F. J. Vivas,
and A. J. Calder´
on, “An optimized balance of plant for a medium-
size pem electrolyzer: Design, control and physical implementation,”
Electronics, vol. 9, no. 5, 2020.
[5] T. H. Ruggles, J. A. Dowling, N. S. Lewis, and K. Caldeira, “Oppor-
tunities for flexible electricity loads such as hydrogen production from
curtailed generation,” Advances in Applied Energy, vol. 3, p. 100051,
2021.
[6] N. Gyawali and Y. Ohsawa, “Integrating fuel
cell/electrolyzer/ultracapacitor system into a stand-alone microhydro
plant,” IEEE Transactions on Energy Conversion, vol. 25, no. 4, pp.
1092–1101, 2010.
[7] M. Korpas and A. Holen, “Operation planning of hydrogen storage con-
nected to wind power operating in a power market, IEEE Transactions
on Energy Conversion, vol. 21, no. 3, pp. 742–749, 2006.
[8] A. M. O. Haruni, M. Negnevitsky, M. E. Haque, and A. Gargoom, A
novel operation and control strategy for a standalone hybrid renewable
power system, IEEE Transactions on Sustainable Energy, vol. 4, no. 2,
pp. 402–413, 2013.
[9] J. Li, J. Lin, H. Zhang, Y. Song, G. Chen, L. Ding, and D. Liang,
“Optimal investment of electrolyzers and seasonal storages in hydrogen
supply chains incorporated with renewable electric networks, IEEE
Transactions on Sustainable Energy, vol. 11, no. 3, pp. 1773–1784,
2020.
[10] A. M. Abomazid, N. El-Taweel, and H. E. Farag, “Optimal energy
management of hydrogen energy facility using integrated battery energy
storage and solar photovoltaic systems, IEEE Transactions on Sustain-
able Energy, pp. 1–1, 2022.
[11] A. E. Samani, A. D’Amicis, J. D. De Kooning, D. Bozalakov, P. Silva,
and L. Vandevelde, “Grid balancing with a large-scale electrolyser
providing primary reserve, IET Renewable Power Generation, vol. 14,
no. 16, pp. 3070–3078, 2020.
[12] N. Veerakumar, Z. Ahmad, M. E. Adabi, J. R. Torres, P. Palensky,
M. van der Meijden, and F. Gonzalez-Longatt, “Fast active power-
frequency support methods by large scale electrolyzers for multi-energy
systems,” in 2020 IEEE PES Innovative Smart Grid Technologies Europe
(ISGT-Europe), 2020, pp. 151–155.
[13] M. Ghazavi Dozein, A. Maria De Corato, and P. Mancarella, “Fast
frequency response provision from large-scale hydrogen electrolyzers
considering stack voltage-current nonlinearity, in 2021 IEEE Madrid
PowerTech, 2021, pp. 1–6.
[14] M. G. Dozein, A. Jalali, and P. Mancarella, “Fast frequency response
from utility-scale hydrogen electrolyzers,” IEEE Transactions on Sus-
tainable Energy, vol. 12, no. 3, pp. 1707–1717, 2021.
[15] O. Gomis-Bellmunt, S. D. Tavakoli, V. A. Lacerda, and E. Prieto-Araujo,
“Grid-forming loads: Can the loads be in charge of forming the grid in
modern power systems?” IEEE Transactions on Smart Grid, vol. 14,
no. 2, pp. 1042–1055, 2023.
[16] X. Quan, Q. Hu, X. Dou, Z. Wu, L. Zhu, and
W. Li, “Control of grid-forming application for fuel
cell/electrolyser system,” IET Renewable Power Generation,
vol. 14, no. 17, pp. 3368–3374, 2020. [Online]. Avail-
able: https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/iet-
rpg.2020.0508
[17] K. Prabakar, Y. Nag Velaga, R. Flores, J. Brouwer, J. Chase, and
P. Sen, “Enhancing distribution system resiliency using grid-forming
fuel cell inverter,” NREL, Tech. Rep., 2022. [Online]. Available:
https://www.nrel.gov/docs/fy22osti/82111.pdf
[18] P. Lettenmeier, “Efficiency electrolysis, Siemens, Tech.
Rep., 2020. [Online]. Available: https://assets.siemens-
energy.com/siemens/assets/api/uuid:5342163d-2333-4c8d-ae85-
2a0e8d45db56/white-paper-efficiency-en.pdf
[19] T. Yigit and O. F. Selamet, “Mathematical modeling and dynamic
simulink simulation of high-pressure pem electrolyzer system,” Interna-
tional Journal of Hydrogen Energy, vol. 41, no. 32, pp. 13 901–13 914,
2016.
[20] N. Cooper, C. Horend, F. R¨
oben, A. Bardow, and N. Shah, A
framework for the design & operation of a large-scale wind-
powered hydrogen electrolyzer hub, International Journal of Hydrogen
Energy, vol. 47, no. 14, pp. 8671–8686, 2022. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S0360319921050278
[21] B. Han, S. M. Steen, J. Mo, and F.-Y. Zhang, “Electrochemical
performance modeling of a proton exchange membrane electrolyzer
cell for hydrogen energy, International Journal of Hydrogen
Energy, vol. 40, no. 22, pp. 7006–7016, 2015. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S036031991500837X
[22] B. Flamm, C. Peter, F. N. B¨
uchi, and J. Lygeros, “Electrolyzer modeling
and real-time control for optimized production of hydrogen gas,” Applied
Energy, vol. 281, p. 116031, 2021.
[23] M. G. Dozein, A. M. De Corato, and P. Mancarella, “Virtual inertia
response and frequency control ancillary services from hydrogen elec-
trolyzers,” IEEE Transactions on Power Systems, pp. 1–12, 2022.
[24] M. Casta˜
neda, A. Cano, F. Jurado, H. S´
anchez, and L. M. Fern´
andez,
“Sizing optimization, dynamic modeling and energy management strate-
gies of a stand-alone pv/hydrogen/battery-based hybrid system,” Inter-
national Journal of Hydrogen Energy, vol. 38, no. 10, pp. 3830–3845,
2013.
[25] Y. Khayat, Q. Shafiee, R. Heydari, M. Naderi, T. Dragiˇ
cevi´
c, J. W.
Simpson-Porco, F. D¨
orfler, M. Fathi, F. Blaabjerg, J. M. Guerrero, and
H. Bevrani, “On the secondary control architectures of ac microgrids:
An overview,” IEEE Transactions on Power Electronics, vol. 35, no. 6,
pp. 6482–6500, 2020.
[26] S. D’Arco and J. A. Suul, “Equivalence of virtual synchronous machines
and frequency-droops for converter-based microgrids, IEEE Trans.
Smart Grid, vol. 5, no. 1, pp. 394–395, 2014.
[27] S. Dadjo Tavakoli, E. Prieto-Araujo, O. Gomis-Bellmunt, and
S. Galceran-Arellano, “Fault ride-through control based on
voltage prioritization for grid-forming converters, IET Renewable
Power Generation, vol. n/a, no. n/a, 2023. [Online]. Available:
https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/rpg2.12682
[28] Electranet, “Dalrymple escri-sa battery project,” 2022. [Online]. Avail-
able: https://www.electranet.com.au/electranets-battery-storage-project
[29] AEMO, “Australian electricity network map, 2022. [Online]. Available:
https://www.aemo.com.au/aemo/apps/visualisations/map.html
[30] SA Power Network, “Zone substation data,” 2020. [Online]. Avail-
able: https://www.sapowernetworks.com.au/data/308954/2019-2020-
zone-substation-data
[31] P. Kundur, Power System Stability and Control. CRC Press, may 2007.
[32] K. Harrison, “Mw-scale pem-based electrolyzers for res
applications,” NREL, Tech. Rep., 2021. [Online]. Available:
https://www.nrel.gov/docs/fy20osti/79055.pdf
Saman Dadjo Tavakoli received his M.S. degree
in electrical engineering from Shahid Beheshti Uni-
versity, Tehran, Iran, in 2015. He joined Technical
University of Catalonia (UPC), Barcelona, Spain,
in 2018 to pursue a Ph.D. degree in electrical
engineering as a part of InnoDC project. Since
2022, he is a control engineer at Siemens Energy.
His research interests include modern power system
stability, advanced control system design for power
converters, and DC microgrid.
Mehdi Ghazavi Dozein received M.Sc. degree from
University of Tehran (2014) and Ph.D. degree from
The University of Melbourne (2021). He is currently
an Associate Lecturer in Power Systems at The Uni-
versity of Melbourne. His research interests include
power system dynamics and stability, and modelling
and control of inverter-based technologies, including
hydrogen electrolyzers.
15
Vin´
ıcius A. Lacerda received a B.Sc. and PhD in
Electrical Engineering from the University of S˜
ao
Paulo, S˜
ao Carlos, Brazil in 2015 and 2021. He was
a Visiting Researcher at the University of Strath-
clyde, UK from 2018 to 2019. He is presently a Post-
Doctorate researcher at the Universitat Polit`
ecnica
de Catalunya (CITCEA-UPC), Spain. His research
interests include power systems modelling and simu-
lation, HIL, dynamics of modern power grids, short-
circuit analysis, protection and high-voltage direct
current systems.
Marc Cheah-Mane (S’14-M’18) received the de-
gree in industrial engineering from the School of
Industrial Engineering of Barcelona (ETSEIB), Uni-
versitat Politecnica de Catalunya (UPC), Barcelona,
Spain, in 2013, and the PhD degree in electrical
engineering from Cardiff University, Cardiff, the
U.K. in 2017. From 2017 to 2020 he was a research
associate in CITCEA-UPC, Barcelona, Spain. Since
2020 he is a Lecturer under the Serra H´
unter pro-
gram at the Electrical Engineering Department of
UPC and since 2022 he is co-founder of eRoots
Analytics, which is a spin-off company of CITCEA-UPC. His research
interests include power systems with power electronics, high-voltage direct
current systems, wind and photovoltaic generation.
Eduardo Prieto-Araujo (S’12-M’16-SM’21) re-
ceived the degree in industrial engineering from
the School of Industrial Engineering of Barcelona
(ETSEIB), Technical University of Catalonia (UPC),
Barcelona, Spain, in 2011 and the Ph.D. degree in
electrical engineering from the UPC in 2016. He
joined CITCEA-UPC research group in 2010 and
currently he is a Serra H´
unter Associate Professor
with the Electrical Engineering Department, UPC.
During 2021, he was a visiting professor at the Au-
tomatic Control Laboratory, ETH Z¨
urich. In 2022, he
co-founded the start-up eRoots Analytics focused on the analysis of modern
power systems. His main interests are renewable generation systems, control
of power converters for HVDC applications, interaction analysis between
converters, and power electronics dominated power systems.
Pierluigi Mancarella received the M.Sc. (2002)
and Ph.D. (2006) degrees in electrical energy sys-
tems from the Politecnico di Torino, Turin, Italy. He
is currently the Chair Professor of Electrical Power
Systems at The University of Melbourne, Mel-
bourne, Australia, and Professor of Smart Energy
Systems at The University of Manchester, Manch-
ester, U.K. His research interests include multi-
energy systems, grid integration of renewables, en-
ergy infrastructure planning under uncertainty, and
resilience of low-carbon networks. Dr. Mancarella is
an Editor of the IEEE TRANSACTIONS ON POWER SYSTEMS, an IEEE
Power and Energy Society Distinguished Lecturer, and the Convenor of the
Cigre Working Group C6/C2.34 “Flexibility provision from distributed energy
resources”.
Oriol Gomis-Bellmunt (S’05-M’07-SM’12-F’21)
received the degree in industrial engineering from
the School of Industrial Engineering of Barcelona
(ETSEIB), Technical University of Catalonia (UPC),
Barcelona, Spain, in 2001 and the Ph.D. degree
in electrical engineering from the UPC in 2007.
In 1999, he joined Engitrol S.L. where he worked
as Project Engineer in the automation and control
industry. Since 2004, he has been with the Elec-
trical Engineering Department, UPC where he is
a Professor and participates in the CITCEA-UPC
Research Group. Since 2020, he is an ICREA Academia researcher. In 2022,
he co-founded the start-up eRoots Analytics focused on the analysis of
modern power systems. His research interests include the fields linked with
power electronics, power systems and renewable energy integration in power
systems.
... Moreover, the PEMElz optimally harnesses excess PV power to generate hydrogen, as detailed in [44]. Consequently, the developed system is reinforced by control schemes, incorporating a Maximum Power Point Tracker (MPPT) for the PV system and a grid-forming control scheme for the HS, as elucidated in [45]. These control strategies enhance system stability, effectively minimizing both power and frequency fluctuations. ...
... It helps establish and maintain the voltage and frequency of the electric grid, as system inertia and damping decrease due to the replacement of traditional synchronous generators with power electronics converters [176,177]. Reference [178] explores PEM electrolyzers as grid-forming loads to actively participate in voltage and frequency regulation. The study details a control system that manages dynamic grid conditions and hydrogen production effectively, as validated through simulations and HIL tests. ...
Article
Full-text available
In recent years, global efforts towards a future with sustainable energy have intensified the development of renewable energy sources (RESs) such as offshore wind, solar photovoltaics (PVs), hydro, and geothermal. Concurrently, green hydrogen, produced via water electrolysis using these RESs, has been recognized as a promising solution to decarbonizing traditionally hard-to-abate sectors. Furthermore, hydrogen storage provides a long-duration energy storage approach to managing the intermittency of RESs, which ensures a reliable and stable electricity supply and supports electric grid operations with ancillary services like frequency and voltage regulation. Despite significant progress, the hydrogen economy remains nascent, with ongoing developments and persistent uncertainties in economic, technological, and regulatory aspects. This paper provides a comprehensive review of the green hydrogen value chain, encompassing production, transportation logistics, storage methodologies, and end-use applications, while identifying key research gaps. Particular emphasis is placed on the integration of green hydrogen into both grid-connected and islanded systems, with a focus on operational strategies to enhance grid resilience and efficiency over both the long and short terms. Moreover, this paper draws on global case studies from pioneering green hydrogen projects to inform strategies that can accelerate the adoption and large-scale deployment of green hydrogen technologies across diverse sectors and geographies.
... The benefits of integrating an electrolyzer into the electricity grid to support renewable energy are widely demonstrated in refs. [21,22]. An additional solution is represented by the transport of hydrogen in city gas pipelines, allowed according to European Directive 2021/2168. ...
Article
Full-text available
The majority of urban CO2 emissions come from the transportation sector. To be able to reduce them, it is definitely necessary to replace Internal Combustion Engine (ICE) vehicles with electric ones. In this article, a public transport system is proposed, consisting of a supercapacitor (SC)-powered electric vehicle (EV) charged through a fuel cell-powered (FC) Inductive Power Transfer (IPT) system. The bus runs the usual route and it is charged each time it reaches the terminal, where the charging system is placed. The main advantages of the proposed system are related to the long-term cost of the EV, compared to a classic battery-powered system, to the aspects of ease of use and safety for charging operations and to the possibility of realizing a net-zero-energy transport system thanks to the use of green hydrogen. In addition, the proposed charging methodology allows for better energy utilization avoiding major changes to the main power grid. In this article, the system is presented considering a real case study; it is simulated at system and hardware level, and then validated through the realization of a scaled-down prototype.
... The extracted hydrogen can then be used as a green fuel [4]. Large renewable energy projects have integrated electrolysers, with the aim of extracting hydrogen to drive future power systems, such as in industries and hybrid transportation [5], [6], [7], [8]. Electrolysers are often coupled with renewables to store excess renewable power when renewable energy production is abundant. ...
Preprint
Full-text available
This paper proposes a power system architecture and control for efficient and low-cost green hydrogen production. The proposed system integrates photovoltaic (PV) sources directly with an electrolyser stack, thereby eliminating the need for traditional power converters. With the removal of traditional power converters, maximum power point tracking is achieved through dynamic switching of electrolyser cells in the stack, enabling load variation to maintain optimal voltage for maximum power output. The demonstration methodology involves comprehensive MATLAB Simulink analysis of the integrated system performance through controlled PV-electrolyser interactions.
Article
The recent increased interest in active distribution microgrids has complicated the process of reaching stable generation with the presence of distributed generators (DGs). These challenges can be handled by employing multi-terminal soft open points (SOPs), which can further boost system performance compared to the conventional two-terminal SOP, especially with present system disturbances. At the same time, hydrogen energy storage has drawn increased attraction to strengthen power grid stability and flexibility. This paper uses a hybrid-based energy storage device that employs an electrolyzer and fuel cell means with a hydrogen tank to absorb or generate power through multi-terminal SOP based on desired grid requirements. This work also offers a novel implementation of a hybrid jellyfish search and particle swarm optimization (HJPSO) applied to model predictive controllers (MPCs) to provide a solution regarding frequency control issues for multiple microgrids comprised of hybrid micro-turbines and renewable generators with the presence of storage devices. Furthermore, different forms of nonlinearity and actual data measurements for the implemented renewable generators are incorporated to attain further realistic analysis. The transient behavior of the studied microgrid is substantially augmented via the suggested control scheme. Referring to the modern grid codes and standards, the frequency operation limits in a specific range. This requirement has been preserved throughout the entire simulations, including uncertainty analysis, assessing the capabilities of multi-terminal SOP in providing frequency disturbance ride-through for the operating active hybrid distribution microgrid under various operating conditions.
Article
Full-text available
This paper discusses that the existing grid codes on the fault ride‐through (FRT) operation have certain complications to be implemented in the grid‐forming (GFM) converters. In particular, they do not provide a clear guidelines on the suitable prioritization of the positive‐ and negative‐sequence currents needed for a desired voltage profile during FRT operation. To address this challenge, two FRT controls which are based on (i) voltage balancing priority, and (ii) voltage magnitude priority are investigated for GFM converters. The former gives priority to the elimination of negative‐sequence voltage, while the remaining converter's capacity is used to increase the positive‐sequence voltage. On the contrary, the latter prioritizes the increase of positive‐sequence voltage, while it tries to reduce negative‐sequence voltage by using the remaining converter's capacity. The dynamic performances of these two FRT controls are thoroughly discussed, and the simulation results show that depending on the desired voltage profile during fault, one of them can be implemented.