energies
Article
Frequency Control in Power Systems with Large Share of
Wind Energy
Liang Lu * , Oscar Saborío-Romano






Citation: Lu, L.; Saborío-Romano, O.;
Cutululis, N.A. Frequency Control in
Power Systems with Large Share of
Wind Energy. Energies 2022, 15, 1922.
https://doi.org/10.3390/en15051922
Academic Editor: Mohamed
Benbouzid
Received: 30 January 2022
Accepted: 2 March 2022
Published: 6 March 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Energies 2022, 15, 1922. https://doi.org/10.3390/en15051922
and Nicolaos A. Cutululis
Department of Wind Energy, Technical University of Denmark, Frederiksborgvej 399, 4000 Roskilde, Denmark;
osro@dtu.dk (O.S.-R.); niac@dtu.dk (N.A.C.)
* Correspondence: lilu@dtu.dk
Abstract: Fast frequency response from wind turbines and plants has been playing an increasingly
important role in our modern power system, which has a large share of wind energy integration.
Frequency control from grid-following wind turbines and plants has been extensively investigated
in the literature, whereas grid-forming frequency control methods have aroused people’s attention
in recent years, with relatively little comparison available. This paper provides such a comparison,
illustrating the differences between the two methods of providing frequency support in a power
system. The comparison shows the superiority of grid-forming frequency control methods over
grid-following ones, and gives a glance into power system frequency stability in the transition from
grid-following to grid-forming frequency support supplied by wind power. The results are elaborated
in a benchmark WSCC 9-bus system, which features detailed electromagnetic transient modeling of
the components from a single converter, a wind turbine, and a wind plant to the whole power system.
Keywords: grid-forming control; virtual synchronous machine control; grid-following control; fast
frequency response; wind power integration
1. Introduction
Wind energy is becoming an increasingly significant electricity supplier in modern
society. Wind energy provided 16% of the electricity consumption in the twenty-seven
EU countries and the UK in 2020, and the total installed capacity is predicted to reach
318 GW by the end of 2025 [1]. The ever-increasing penetration of inverter-based-resources
(IBRs) brings an increasingly large challenge in maintaining the frequency stability of
power systems.
To address the challenge, frequency control methods from grid-following (GFL) wind
turbines (WTs) and wind plants (WPs) were proposed at an early stage. For simplicity, these
methods are named ‘grid-following frequency control methods’ (GFL-FCMs). GFL-FCMs
usually include virtual inertial response [2,3], which is proportional to rate of change of
frequency (RoCoF), and frequency-active power ( f − P) droop control. Such methods rely
on phasor-locked loops (PLLs) to estimate the frequency, which consequently causes a time
lag [4] and induces noise to the control system that may lead to instability [5]. Furthermore,
synchronization instability is prone to occur for GFL controls with an increasing number of
paralleled inverters [6] or in weak grids [7–11].
Later, grid-forming (GFM) control was proposed [12], which enables converters to
work as an ideal voltage source with a given amplitude and frequency. Considering
frequency control capability only, grid-forming frequency control methods (GFM-FCMs)
have been investigated and compared with GFL-FCMs in some work. A GFL-FCM and
a GFM-FCM for photovoltaic inverters were compared in [13], but the GFL-FCM only
includes f − P droop control without inertial emulation. Similarly, the GFM-FCM also
only implements droop control for the frequency regulation. Differences in impacts of a
GFM-FCM and a GFL-FCM on system frequency dynamics at various penetration levels of
https://www.mdpi.com/journal/energies
Energies 2022, 15, 1922 2 of 12

inverters and values of mechanical inertia were evaluated in [14]. The considered GFM-
FCM is a droop-based control algorithm with the voltage controller dynamics ignored. Only
a two-source system was investigated, one synchronous generator (SG) and one inverter;
this means that the GFM-FCM and GFL-FCM were compared in separate systems against
the same SG. No dynamics with mixed sources or inter-area interactions were captured.
The modeling in [14] was rather simple: the current controller dynamics of the inverter
were ignored; the power response was modeled as a first-order system and the PLL was
represented as a low-pass filter. Furthermore, no comparison in the system frequency was
given between these two methods. Optimization of the parameters and locations of virtual
inertia devices were investigated in [15] to increase the resilience of a power system. GFM
and GFL controls were compared for fast frequency response (FFR), but only virtual inertia
was implemented for both controls, without considering f − P droop control. Although
time-domain responses of active power and frequency were compared when the virtual
inertia devices were controlled in GFM and GFL modes respectively in a 14-generator,
59-bus South-East Australian system, responses from multiple units were plotted together
in the same figure; therefore, it is not easy to obtain the distinct differences between GFM
and GFL controls. Moreover, in such a large system, no consideration was given to the
scenario in which GFM and GFL units coexist.
In this paper, a GFM-FCM and a GFL-FCM are compared for fast frequency response
(FFR) in a power system with a large share of wind power. The GFL-FCM consists of
virtual inertial response and f − P droop control, while the GFM-FCM is based on a virtual
synchronous machine (VSM) control scheme. These two methods are not only compared
separately in identical systems, but they are also compared when they coexist in the same
system. Therefore, the interactions during FFR between a grid-forming wind plant (GFM-
WP) and a grid-following wind plant (GFL-WP) are revealed. This approach also gives
insights into the frequency stability of a power system in the transition from GFL-WPs to
GFM-WPs. The comparison is conducted in a WSCC 9-bus system [16], which provides
a multi-source, multi-area scenario. A rather detailed electromagnetic transient (EMT)
modeling is implemented, including all transformers and transmission lines. Within a
single WP, collector cables are included and each WT has an average model of its grid-side
converter (GSC) with complete inner controllers.
The rest of this paper is organized as follows. Section 2 describes the GFL-FCM and
GFM-FCM that are considered in this work and their implementations in WPs. Comparison
of these two methods for FFR in a WSCC 9-bus system is illustrated in Section 3, before
concluding in Section 4.

2. Modeling and Control of WPs for FFR

In this section, we describe the GFL-FCM and GFM-FCM compared in this work in
detail. Both methods are implemented at the WT level, more specifically, on the GSC of
each WT. The dynamics of the DC-link, the machine-side converter (MSC), the generator
and the mechanical drivetrain for each WT are beyond the scope of this work. The capacity
of each WT is 10 MVA.
As the discussion in this work is focused on the system level, WP modeling is necessary.
WPs are modeled as 3-machines (WTs) aggregated unit, with a total capacity of 30 MVA.
The terminal voltage of each WT is 690 V. A WT transformer changes the voltage from
690 V to 66 kV, at which the collector cables operate. A 2 km collector cable is modeled
using a π section and its parameters are included in Table A1 in the Appendix A. Each
cable connects a WT transformer (parameters given in Table A2 in the Appendix A) and
the WT behind to the point of connection (PoC) in each WP. The schematic of the WP is
shown in Figure 1.
Energies
Energies2022,
Energies 2022,15,
2022, 15,x1922
15, xFOR
FORPEER
PEERREVIEW
REVIEW 333of
of13
of 12
13

Figure
Figure1.1.Schematic
Schematicof
ofaaWP.
WP.
WP.

2.1.
2.1.Grid-Following
2.1. Grid-FollowingFrequency
Grid-Following FrequencyControl
Frequency ControlMethod
Control Method(GFL-FCM)
Method (GFL-FCM)
(GFL-FCM)
AAGFL-FCM
A GFL-FCMregulates
GFL-FCM regulatesthe
regulates theactive
the activepower
active power
power injection
injection
injection fromfrom
from aaWT a WT
WT or
orWP
or WP
WP based on
based
based ontheonfre-
the the
fre-
quency
frequency deviation
deviation and andRoCoF
RoCoF estimated
estimatedfrom
from the voltage
the voltage atat
quency deviation and RoCoF estimated from the voltage at PoC. Therefore, a frequencyPoC.
PoC. Therefore,
Therefore, aa frequency
frequency
estimator,
estimator,i.e.,
estimator, i.e.,aaaPLL
PLLisisisused
PLL usedtoto
used toestimate
estimatethe
estimate thesystem
the system
system frequency
frequency
frequency and andRoCoF.
and RoCoF.
RoCoF.TheThesynchronous
The synchro-
synchro-
reference
nous frame PLL (SRF-PLL) with a PI compensator is commonly
nous reference frame PLL (SRF-PLL) with a PI compensator is commonly used incontrols
reference frame PLL (SRF-PLL) with a PI compensator is commonly usedusedin controls
in of
controls
converters
of converters [17].
[17].In addition,
In addition, incorporating
incorporatinga low-pass
a low-pass filter (LPF)
filter in
(LPF)
of converters [17]. In addition, incorporating a low-pass filter (LPF) in the loop together the
in loop
the together
loop with
together
awith
SRF-PLL
with helpshelps
aaSRF-PLL
SRF-PLL provide
helps an explicit
provide
provide an estimation
anexplicit
explicitestimationof RoCoF
estimation of
ofRoCoF[18]. [18].
RoCoF [18].
A typical
AAtypical GFL-FCM for offshore wind plants is presented in [19],used
and used as a
typical GFL-FCM for offshore wind plants is presented in[19],
GFL-FCM for offshore wind plants is presented in [19],andand usedas asaabench-
bench-
benchmark
mark for comparisonawith a GFM-FCM.controlThe control scheme is reproducedFigure in Figure 2.
markfor forcomparison
comparisonwith with aGFM-FCM.
GFM-FCM.The The controlscheme schemeisisreproduced
reproducedin in Figure2.2.

Figure
Figure2.
2.The
Thegrid-following
grid-followingfrequency
frequencycontrol
controlmethod
control method(GFL-FCM).
method (GFL-FCM).
(GFL-FCM).

In
In this
In control
thiscontrol
controlscheme,scheme,f n𝑓𝑓is is
scheme, is the
the nominal
nominal
the nominal system
system frequency.
frequency.
system frequency. f esti is𝑓𝑓the estimated
isis the
the estimated
system
estimated
system
frequency frequency
from a from
PLL. a
ThePLL. The
difference
system frequency from a PLL. The difference between difference
between between
f n and 𝑓f and is
𝑓estiand 𝑓 𝑓
the isis the frequencydevi-
frequencythe frequency
deviation ∆f,
devi-
ation ∆𝑓,
ation
which which
whichisby
is filtered
∆𝑓, isfiltered
a low-pass
filtered by
byaalow-pass
filter
low-pass(LPF). filter
The
filter (LPF).
timeThe
(LPF). Thetimetimeconstant
constant T LPF for both
constant 𝑇𝑇 LPFs for
forbothis 2.LPFs
both LPFs
isis2.2. After the LPF, the upper branch in Figure 2 represents the f − P droop control of the
After
GFL-FCM.Afterthe DrLPF,
the is thethe
LPF, upper
droop
the upper branch
branchin
coefficient, Figure
inset to 0.05
Figure 22represents
in this work.
represents the
theThe𝑓𝑓−−lower
𝑃𝑃 droop branch
droop control of
ofthe
represents
control the
the virtual
GFL-FCM. 𝐷𝑟
inertial
is response
the droop of the GFL-FCM,
coefficient, set to 0.05
which is
in proportional
this
GFL-FCM. 𝐷𝑟 is the droop coefficient, set to 0.05 in this work. The lower branch repre- work. The to the
lower derivative
branch of the
repre-
frequency
sents
sentsthe deviation
thevirtual
virtual inertial
inertial∆ fresponse
KGFL is of
.response chosen
ofthe to be 13 under
theGFL-FCM,
GFL-FCM, whichthe
which premise thatto
isisproportional
proportional tothe
the
thecontrols
derivative
derivativeare
in
of
oftheper unit,
thefrequency i.e., f
frequencydeviation , f
n esti
deviation and ∆P
∆𝑓.
∆𝑓. 𝐾 are
𝐾 isischosen
GFL all in
chosentoper unit.
tobebe13 With
13under
underthe this value,
thepremise
premisethatthe best
thatthe frequency
thecontrols
controls
support
are in per performance
unit, i.e.,
are in per unit, i.e., 𝑓 , 𝑓 𝑓 , 𝑓
(using metrics
and
and ∆𝑃 ∆𝑃 suchare as RoCoF and nadir of frequency)
are all in per unit. With this value, the best fre-
all in per unit. With this value, is
the achieved
best fre-
withoutsupport
quency inducingperformance
high frequency oscillations in the active power output ∆PGFL and without
quency support performance (using metrics such (using metrics such as as RoCoF
RoCoF and and nadirnadir of of frequency)
frequency) isis
output ∆𝑃
breachingwithout
achieved the over-current capability of the WT converters.
achieved withoutinducing inducinghigh highfrequency
frequencyoscillations
oscillationsin inthetheactive
activepower poweroutput ∆𝑃
and The active
without power the
breaching components
over-current from the f − Pofdroop
capability the WT control branch and the virtual
converters.
and without breaching the over-current capability of the WT converters.
active power ∆P
the 𝑓𝑓−−𝑃𝑃 droop
inertiaThe branch together make up the overall GFL that is used forthe frequency
Theactive
activepower
powercomponents
componentsfrom fromthe droopcontrolcontrol branch
branchand and thevirtual
virtual
regulation.
inertia branch ∆P together
GFL is added
make toupthe normal
the overall active
active power
power reference
∆𝑃 for
that each
is WT.
used forIn this work,
frequency
inertia branch together make up the overall active power ∆𝑃 that is used for frequency
we assume all WTs have available power of 0.9 pu but operate in a curtailed way, reserving
regulation. ∆𝑃
regulation. ∆𝑃 isis added added to to the
the normal
normal activeactive power
power reference
reference for for each
each WT. WT. In In this
this
10%
work, of
we their nominal
assume all power.
WTs haveHence, the
available active
power power
of 0.9 reference
pu but for
operate each inWTa is 0.8 pu.way,
curtailed
work, we assume all WTs have available power of 0.9 pu but operate in a curtailed way,
 Energies 2022, 15, x FOR PEER REVIEW 4 of 13

Energies 2022, 15, 1922 4 of 12

reserving 10% of their nominal power. Hence, the active power reference for each WT is
0.8 pu.
With thethe
With GFL-FCM
GFL-FCMimplemented,
implemented, thethecontrol
control scheme
scheme of the
of the GSCGSC in WTs
in WTs is shown
is shown in in
Figure 3. The
Figure parameters
3. The parametersare
are given in Table
given in TableA4A4ininthe
the Appendix
Appendix A. A.

Figure
Figure 3. The
3. The GSCGSC controlscheme
control scheme with
with the
theGFL-FCM.
GFL-FCM.

2.2.2.2. Grid-FormingFrequency
Grid-Forming FrequencyControl
Control Method
Method(GFM-FCM)
(GFM-FCM)
AnAn SGSGcancan regulatethe
regulate theterminal
terminal voltage
voltageand andsystem
systemfrequency
frequencyby itsbyexcitation and and
its excitation
governor
governor respectively.This
respectively. Thiskind
kind of
of voltage
voltageand andfrequency
frequency control capability
control is termed
capability as
is termed as
grid-forming (GFM). There are different control methods that all lie in the scope of GFM
grid-forming (GFM). There are different control methods that all lie in the scope of GFM
control [20].
control [20].
A reduced-order-VSM-based frequency control scheme is presented in [21]. It pro-
A reduced-order-VSM-based frequency control scheme is presented in [21]. It provides
vides satisfactory 𝑓 − 𝑃 droop control and virtual inertia with considerable simplicity.
Because off −
satisfactory itsPsimplicity,
droop control and virtual
it contains essential inertia withofconsiderable
elements VSM controlsimplicity. Because of
while excluding
its simplicity,
features that are unique to different realizations of VSM control in the literature [22–25]. that
it contains essential elements of VSM control while excluding features
are Therefore,
unique tothe different
controlrealizations of VSM
scheme in [21], control
as a typical in the
VSM literature
control scheme, [22–25]. Therefore,
is considered as the
control scheme in [21], as a typical VSM control scheme, is considered
the GFM-FCM in this work, with the addition of a virtual impedance, facilitating parallel as the GFM-FCM
in this work, with
operation. the addition
The block diagram of athe virtual
GFM-FCM impedance, facilitating
implemented parallel
on each WT is operation.
shown in The
Figure 4 below. The droop coefficient 𝐷𝑟 is the same
block diagram of the GFM-FCM implemented on each WT is shown in Figure with that in the GFL-FCM, which4isbelow.
0.05. The inertia constant 𝐻 is kept at 4 s in this work. The
The droop coefficient Dr is the same with that in the GFL-FCM, which is 0.05. The values of the remaining pa-inertia
rameters are
2022, 15, x FORHPEER
Energiesconstant in Table
is REVIEW A4 in the Appendix A.
kept at 4 s in this work. The values of the remaining parameters are 5 ofin13

Table A4 in the Appendix A.

Figure 4. The GSC control scheme with the GFM-FCM.

Figure 4. The GSC control scheme with the GFM-FCM.
The block diagram of the virtual impedance is shown in Figure 5. 𝑅 is 0.06 pu, and
𝐿 is 5 × 10 pu.
Energies 2022, 15, 1922 5 of 12

Figure 4. The GSC control scheme with the GFM-FCM.

The
Theblock
block diagram
diagram of
of the
the virtual impedance 𝑅V I is 0.06
impedance is shown in Figure 5. R 0.06pu,
pu,and
and
L𝐿V I is 10−3 pu.
is 55××10 pu.

Figure5.5. Virtual
Figure Virtualimpedance.
impedance.

3. Comparison of the GFM-FCM and GFL-FCM on FFR

3. Comparison of the GFM-FCM and GFL-FCM on FFR
In this section, we compare the influence of GFM-FCM and GFL-FCM on fast frequency
In this section, we compare the influence of GFM-FCM and GFL-FCM on fast fre-
response (FFR) from WPs. In this work, FFR is defined as the same fast frequency response
quency response (FFR) from WPs. In this work, FFR is defined as the same fast frequency
that needs to be delivered in the fast frequency reserve product [26], in terms of technical
response that needs to be delivered in the fast frequency reserve product [26], in terms of
properties (activation, duration, etc.). The comparison is tested in the benchmark WSCC
Energies 2022, 15, x FOR PEER REVIEWtechnical properties (activation, duration, etc.). The comparison is tested in the benchmark
6 of 13
9-bus system [27–29]. The topology is shown in Figure 6. The parameters of the passive
WSCC 9-bus system [27–29]. The topology is shown in Figure 6. The parameters of the
components (transformers and lines) can be found in Table A3 in Appendix A.
passive components (transformers and lines) can be found in Table A3 in Appendix A.

Figure 6. WSCC 9-bus system.

Figure 6. WSCC 9-bus system.

There are three

There are threepower
powerplants
plantsin in
thethe system;
system; 𝐺2a is
G2 is a conventional
conventional thermal
thermal powerpower
plant
plant (TPP), 𝐺1 and 𝐺3
(TPP), G1 and G3 are two WPs that have the same configuration, as shown in Figurein
are two WPs that have the same configuration, as shown 1. Fig-
The
ure 1. The difference between them is the control mode. In the following simulations, three
scenarios are considered. The first is that both WPs are controlled in GFL mode; the second
is that 𝐺1 is a GFL-WP whereas 𝐺3 is a GFM-WP; and the third scenario is that both are
GFM-WPs. It each scenario, the same frequency event is tested. Therefore, the GFL-FCM
Energies 2022, 15, 1922 6 of 12

difference between them is the control mode. In the following simulations, three scenarios
are considered. The first is that both WPs are controlled in GFL mode; the second is that G1
is a GFL-WP whereas G3 is a GFM-WP; and the third scenario is that both are GFM-WPs. It
each scenario, the same frequency event is tested. Therefore, the GFL-FCM and GFM-FCM
described in Section 2 are compared in the same system in the second scenario. More
significantly, these three scenarios show how the system frequency stability is developing
when GFL-WPs are gradually replaced by GFM-WPs.
As mentioned in Section 2, for the two WPs G1 and G3, each consists of three WTs
and each WT has a capacity of 10 MVA. Hence, each WP has a capacity of 30 MVA. For
simplicity, the thermal power plant G2 has the same capacity as each WP, and the three
loads are equal in the active power consumption. The parameters of the TPP can be found
in Table A6 in Appendix A.
An under-frequency event is simulated to check the frequency response of the system.
Initially, all three power plants are operating at 0.8 pu of their nominal power, and the WPs
have a 10% reserve used for frequency support in an under-frequency event. The event
happens at t = 25 s, when each load is increased by 3 MW (equal to 10% capacity of each
power plant). The load increase is equally shared by the three power plants as they have
the same droop coefficient Dr = 0.05. Therefore, for each power plant, their active power
Energies 2022, 15, x FOR PEER REVIEW 7 of 13
output is supposed to increase by 0.1 pu, from 0.8 pu to 0.9 pu. The results of this frequency
event for the three scenarios considered are shown in Figures 7 and 8.

Figure 7. Comparison of GFL-FCM and GFM-FCM on FFR. 1st column: scenario (i) one TPP with
Figure 7. Comparison of GFL-FCM and GFM-FCM on FFR. 1st column: scenario (i) one TPP with
two GFL-WPs; 2nd column: scenario (ii) one TPP with one GFL-WP and one GFM-WP; 3rd column:
two GFL-WPs; 2nd column: scenario (ii) one TPP with one GFL-WP and one GFM-WP; 3rd column:
scenario (iii) one TPP with two GFM-WPs.
scenario (iii) one TPP with two GFM-WPs.
The three columns in Figure 7 represent the three scenarios respectively. In all three
scenarios, 𝐺2 is unchanged, which is the thermal power plant in Figure 6, whereas 𝐺1
and 𝐺3 are varied in the three scenarios: (i) 𝐺1 and 𝐺3 are GFL-WPs; (ii) 𝐺1 is a GFL-
WP while 𝐺3 is a GFM-WP; (iii) 𝐺1 and 𝐺3 are GFM-WPs.
The first row in Figure 7 is the active power output from the three power plants after
the under-frequency event 𝑡 = 25 s. The second row depicts the corresponding grid fre-
quency in each scenario for the same frequency event. Each subfigure plots of the grid
frequency are estimations at these power plants. For the thermal power plant 𝐺2, the ro-
tating speed of the SG is used as an estimation for the grid frequency. For a GFL-WP, the
estimated frequency comes from the PLL. For a GFM-WP, the estimated frequency is the
 the frequency estimations are identical as it is a global variable. However, the most im-
portant part is the nadir of the frequency, and the results show a clear improvement when
GFM-WPs are gradually replacing GFL-WPs in the system.
To have a clearer and more direct comparison of the three scenarios, we plot the sys-
Energies 2022, 15, 1922
tem frequency of the three scenarios in the same figure (Figure 8). The frequency estima-
7 of 12
tion from the rotating speed of the SG is used as the indicator of the system frequency for
all three scenarios.

Figure 8.
Figure 8. Comparison
Comparison of
of GFL-FCM
GFL-FCM and
and GFM-FCM
GFM-FCM on
onFFR
FFR(system
(systemfrequency).
frequency).

The
The three columns in Figure
improvements—both for7the
represent the nadir
frequency three and
scenarios
for therespectively.
RoCoF—areIn all three
very clear,
scenarios, G2 is unchanged, which is the thermal power plant in
leading to the conclusion that by replacing a GFL-WP with a GFM-WP, both the nadirFigure 6, whereas G1 and of
G3 are varied
frequency andinthe
theRoCoF
three scenarios:
are improved(i) G1inand G3 are GFL-WPs; event.
an under-frequency (ii) G1 is a GFL-WP while
G3 is Furthermore,
a GFM-WP; (iii) G1 and G3
to compare are GFM-WPs.
these three scenarios horizontally both in Figures 7 and 8,
The easy
it is also first to
row in Figure
draw 7 is thethat
the conclusion active power output
by replacing a GFL-WPfrom with
the three power itplants
a GFM-WP, takes
after the under-frequency event t =
less time for the system to go through the transients and to reach a new steady state grid
25 s. The second row depicts the corresponding after
frequency
the frequency in each scenario for the same frequency event. Each subfigure plots of the grid
event.
frequency are estimations at these power plants. For the thermal power plant G2, the
rotating speed of the SG is used as an estimation for the grid frequency. For a GFL-WP, the
4. Conclusions
estimated frequency comes from the PLL. For a GFM-WP, the estimated frequency is the
In this paper, we mainly compare two different frequency control methods from
internal frequency within the GFM controller in each WT (ω in Figure 4).
wind turbines and plants: grid-following and grid-forming control. Two control schemes
The first row shows the active power outputs after the frequency event, starting on the
of these two types of controls are introduced, implemented and compared in a WSCC 9-
left with the system with one TPP and two GFL-WPs. As there is a virtual inertial response
bus system. Three scenarios of system configurations are simulated and compared, in
included in the GFL-FCM, at the beginning of the event, the GFL-WPs are able to provide
which grid-following and grid-forming controlled wind power take different shares in
an instantaneous active power support, which is around 0.05 pu in this event of 0.1 pu
pre-event power output and post-event frequency support. This gives an insight into how
load increase. However, this instantaneous active power support tends to be just a pulse;
the frequency stability of power systems changes as grid-forming wind plants gradually
the active power outputs from GFL-WPs drop instantly close to the pre-event level after
replace grid-following ones.
the pulse and then gradually increase. On the contrary, the TPP provides a much larger
Grid-following and grid-forming frequency control methods present rather different
active power support, which is more than 25% of its capacity, immediately after the event
because of support
frequency response
its inherent in active
inertia. More power outputthough
importantly, from wind plants.decreasing
gradually With virtual inertial
after the
peak, this extra power persists above 0.9 pu for half a second before reaching the samecould
response implemented in the frequency control method, grid-following wind plants level
provide
as instantaneous
those from GFL-WPs.power support
Therefore, at the instant
we might conclude of that,
the frequency
in the firstevent, but this
half second power
after the
output cannot sustain long, being almost just a pulse, especially when
frequency event, which can be considered as the (virtual) inertial response period, the main coexisting with
other grid-forming
contributions units (SGs
for frequency or GFM-WPs)
support come from in the
thesystem. The magnitude
TPP instead of the pulse
of the GFL-WPs. Afteris
the (virtual) inertial response period, all three power plants contribute to the frequency
support by outputting their active power at 0.9 pu, as a consequence of the droop control
implemented in them. In addition, low frequency (around 0.5 Hz) interactions are observed
in active power between the TPP and GFL-WPs. These two WPs have almost the same
responses in this event.
In the middle subfigure of the first row, three distinct curves represent the active power
response from three different types of power plants in the system. In this scenario, the
GFM-WP replaces the TPP and takes the main responsibility for the instant power support
at the moment of the event. The instant power support rises up to more than 20% of its
capacity. It needs to be addressed that, in this work, no current limitation is considered for
Energies 2022, 15, 1922 8 of 12

the GSCs in WTs. Furthermore, we assume the WPs have a 10% reserve in active power,
which is meant for the extra power output in the steady state after the event, for example,
t = 29 s in Figure 7. Hence, the extra part of their power output in the first second after the
event, which is beyond the 10% reserve, needs to be supplied from somewhere else. An
energy buffer, either a super-capacitor or battery, at the DC-link is able to fulfil the task,
but this aspect is out of the scope of this work. A DC voltage source is connected at the
DC-link for all the WTs (as shown in Figures 1, 3 and 4), which provides the possibility of
any amount of extra power from WTs.
If we still consider the first half second as the (virtual) inertial response period, in
this stage, the main contributions for frequency support come from the GFM-WP and the
TPP. The GFM-WP has an even larger contribution than the TPP in terms of the velocity of
action and the amount of power. The former might be due to the rapidity of converters,
which is an obvious advantage of GFM-WPs in FFR. For the GFL-WP, its power output at
the instant of the event is very similar to that in the subfigure on the left, which increases
and decreases almost at the same time (t = 25 s). It is also a pulse of around 0.05 pu, but it
is overlapped by the blue curve in the middle subfigure. Afterwards the GFL-WP slowly
increases its active power from pre-event set-point 0.8 pu to post-event steady state 0.9 pu.
Because of the droop control implemented in all three power plants, they all contribute
equally in compensating for the load change. In this scenario, low frequency (around 2 Hz)
interactions are observed in active power from the TPP and the GFM-WP.
The subfigure in the upper right corner gives the results of active power responses in
the system with the TPP and two GFM-WPs. The initial responses are very similar to those
in the middle subfigure. The GFM-WPs contribute more than the TPP in the momentary
overshoot of power at the starting of the event. This is because virtual inertial response
is inherent in the GFM-FCM by emulating the swing equation of SGs. Although inertial
response is inherent in SGs, it seems that the virtual inertial response from a GFM converter
behaves faster and stronger than the inertial response from the mechanical rotating mass of
an SG, under the circumstances that both of them have the same (virtual) inertial constant
values. This is thanks to the rapidity of power electronic devices and their electronic
controllers, which possibly turns out to be an advantage of GFM IBRs over traditional TPPs
in maintaining the power system stability. As a GFM-WP is contributing more than a TPP
in the early stage of FFR (around 1.5 s after the event, including the virtual/real inertial
response period) and there are two GFM-WPs in this scenario, the TPP is contributing
less in this early stage than that in the second scenario (the middle subfigure). Similar
interactions in active power, such as those in the second scenario also exist from the TPP
and GFM-WPs in the third scenario. All three power plants share the same load increase in
the post-event steady state due to the same droop coefficient values in their controls. In
general, from the second and third scenarios, we can conclude that GFM controls enable a
WP to achieve behaviour (FFR in this case) that is more dominant than that of a TPP.
In the second row are the results of the system frequency. The frequency curves are
derived at each power plant, with its own estimation method. As GFL-WPs rely on PLLs
to estimate the system frequency, drastic changes in the frequency values are observed at
the origin of the event in the first two subfigures (from left to right). This illustrates the
transient process in the PLL to track the frequency. These drastic changes in frequency
estimation bring challenges in calculating the RoCoF and introduce disturbances to the
control system. This could be considered as a drawback of using PLL in FFR. In addition,
there are also some minor mismatches between the frequency estimations from the TPP
and GFM-WPs in the early stage of FFR (around 1.5 s after the event). In steady state,
all the frequency estimations are identical as it is a global variable. However, the most
important part is the nadir of the frequency, and the results show a clear improvement
when GFM-WPs are gradually replacing GFL-WPs in the system.
To have a clearer and more direct comparison of the three scenarios, we plot the system
frequency of the three scenarios in the same figure (Figure 8). The frequency estimation
Energies 2022, 15, 1922 9 of 12

from the rotating speed of the SG is used as the indicator of the system frequency for all
three scenarios.
The improvements—both for the frequency nadir and for the RoCoF—are very clear,
leading to the conclusion that by replacing a GFL-WP with a GFM-WP, both the nadir of
frequency and the RoCoF are improved in an under-frequency event.
Furthermore, to compare these three scenarios horizontally both in Figures 7 and 8, it
is also easy to draw the conclusion that by replacing a GFL-WP with a GFM-WP, it takes
less time for the system to go through the transients and to reach a new steady state after
the frequency event.

4. Conclusions
In this paper, we mainly compare two different frequency control methods from
wind turbines and plants: grid-following and grid-forming control. Two control schemes
of these two types of controls are introduced, implemented and compared in a WSCC
9-bus system. Three scenarios of system configurations are simulated and compared, in
which grid-following and grid-forming controlled wind power take different shares in
pre-event power output and post-event frequency support. This gives an insight into how
the frequency stability of power systems changes as grid-forming wind plants gradually
replace grid-following ones.
Grid-following and grid-forming frequency control methods present rather different
frequency support response in active power output from wind plants. With virtual inertial
response implemented in the frequency control method, grid-following wind plants could
provide instantaneous power support at the instant of the frequency event, but this power
output cannot sustain long, being almost just a pulse, especially when coexisting with other
grid-forming units (SGs or GFM-WPs) in the system. The magnitude of the pulse is also
quite limited. Hence, grid-following frequency control methods in fact cannot contribute
much in the virtual inertial response period, in which grid-forming units take the main
responsibility. Overall, in the whole frequency support process (from the beginning of the
event to the post-event steady state), grid-following wind plants behave just like ‘followers’.
By saying so, we mean, they are not as active as grid-forming units in their ability to
quickly and largely increase the power output to make up the power imbalance. On the
contrary, grid-following wind plants increase power output at a slow pace and gradually
reach the steady state with no or little overshoot. Partly this is due to the dominant role of
grid-forming units in frequency and voltage controls.
In contrast, grid-forming wind plants show very similar behavior as synchronous
generator-based power plants in frequency control, which is very beneficial for maintaining
the system frequency stability while at the same time increasing the wind power penetra-
tion in the system. More importantly, grid-forming wind plants demonstrate even more
dominant behavior in the frequency support context than thermal power plants, due to the
rapidity in response of power electronic devices compared to slow dynamics in rotating
machines. This indicates a potentially promising popularization of grid-forming IBRs and
their contributions to forming a reliable 100% inverter-based power system.
From the comparison of the three scenarios, by replacing a grid-following wind plant
with a grid-forming one, the system frequency undergoes less severe RoCoF and higher
nadir in an under-frequency event. This illustrates the superiority of grid-forming frequency
control methods over grid-following ones. This also suggests a roadmap towards increasing
wind power penetration further in the system while maintaining or even improving the
system’s frequency stability.

Author Contributions: Conceptualization, L.L. and N.A.C.; methodology, L.L., O.S.-R. and N.A.C.;
software, L.L. and O.S.-R.; validation, L.L., O.S.-R. and N.A.C.; formal analysis, L.L., O.S.-R. and
N.A.C.; investigation, L.L.; resources, L.L., O.S.-R. and N.A.C.; data curation, L.L.; writing—original
draft preparation, L.L.; writing—review and editing, O.S.-R. and N.A.C.; visualization, L.L.; supervi-
sion, O.S.-R. and N.A.C.; project administration, L.L.; funding acquisition, N.A.C. All authors have
read and agreed to the published version of the manuscript.
Energies 2022, 15, 1922 10 of 12

Funding: The research was funded by the European Union’s Horizon 2020 research and innovation
program under grant agreement No. 727680 (TotalControl) and EUDP under grant agreement No.
12558 (PowerKey). The APC was funded by DTU Wind Energy.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A
The cables in the wind plants are modelled by a three-phase PI section in MATLAB
Simulink. For each cable, the length is 2 km, and the frequency is 50 Hz. The parameters
are listed in Table A1.

Table A1. Parameters of cables in the wind plants.

Parameter Value Parameter Value

Positive-sequence Zero-sequence
0.264 0.713
resistance (Ω/km ) resistance (Ω/km )
Positive-sequence Zero-sequence
4.58 × 10−4 4.55 × 10−4
inductance (H/km ) inductance (H/km )
Positive-sequence Zero-sequence
1.48 × 10−7 1.48 × 10−7
capacitance (F/km ) capacitance (F/km )

The wind turbine transformers are modelled using a Three-Phase Transformer block,
comprising three single-phase transformers. The nominal power is 15 MVA and the
frequency is 50 Hz. The other parameters are listed in Table A2.

Table A2. Parameters of wind turbine transformers in the wind plants.

Parameter Value Parameter Value

Winding 1 connection Yg Winding 2 connection Yg
Winding 1 voltage (LL, RMS) 690 V Winding 2 voltage (LL, RMS) 66 kV
Winding 1 resistance (pu ) 1 × 10−6 Winding 2 resistance (pu ) 1 × 10−6
Winding 1 inductance (pu ) 0 Winding 2 inductance (pu ) 0.0586
Magnetization resistance (pu ) 500 Magnetization inductance (pu ) 500

The parameters of transformers and lines in the WSCC 9-bus system are listed in
Table A3. The values are in per unit. The base capacity is 100 MVA, and the base voltage
is 230 kV. The nominal power for transformers is 30 MVA. We assume all the lines are
100 km.

Table A3. Parameters of transformers and lines in the WSCC 9-bus system.

Name Parameter (pu) Name Parameter (pu) Name Parameter (pu)

T1 XT = 0.0576 Line 1 Z = 0.01 + j0.085, B/2 = j0.08 Line 4 Z = 0.039 + j0.17, B/2 = j0.179
T2 XT = 0.0625 Line 2 Z = 0.032 + j0.161, B/2 = j0.153 Line 5 Z = 0.0085 + j0.072, B/2 = j0.0745
T3 XT = 0.0586 Line 3 Z = 0.017 + j0.092, B/2 = j0.079 Line 6 Z = 0.0119 + j0.1008, B/2 = j0.1045

The parameters of the GSC control scheme implemented in WTs are listed in Table A4.
As for some parameters in per unit, the per unit representation and simulation settings are
given in Table A5.
Energies 2022, 15, 1922 11 of 12

Table A4. Parameters of the GSC control scheme.

Parameter Value Parameter Value

Nominal capacity, Sn 10 MVA Current loop time constant, τc 7.9577 × 10−4 s
Nominal voltage (LL, RMS), Vn 690 V Current loop proportional gain, K pc 0.4
Nominal current (RMS), In 8.3674 kA Current loop integral gain, Kic 0.04
Nominal frequency, f n 50 Hz Voltage loop proportional gain, Kv 0.0667
DC − link voltage, VDC 2000 V Voltage 139.6263
Converter switch − Voltage low −
1 × 10−3 0.0072
on resistance, Ron pass filter time constant, T1
Filter resistance, R f 1 × 10−3 Inertia 4 W·s/VA
Filter inductance, L f 0.1 pu Droop coefficient, Dr 0.05
Filter capacitance, C f 0.05 pu

Table A5. Per unit representation and simulation settings.

Parameter Value Parameter Value

Power base, Sb 10 MVA Inductance base, Lb = Zb /ωb 1.5155 × 10−4 H
Voltage base (phase − ground, peak), Vb 563.3826 V Capacitance base, Cb = 1/Zb /ωb 0.0669 F
Current base, Ib = Sb /Vb × 2/3 1.1833 × 104 A Simulation time step, Ts 160 × 10−6 s
Impedance base, Zb = Vb /Ib 0.0476 Ω Initial power set point of GSC in WTs 0.8 pu
Frequency base, f b 50 Hz Initial power set point of TPP 0.8 pu
Angular speed base, ωb = 2π f b 314.1593 rad/s

A salient pole rotor is modelled in the SG, hence there is no parameter q axis transient
reactance Xq0 , open-circuit (Tqo
0 ) or short-circuit (T 0 ) time constant. The d axis is the short-
q
circuit and q axis is the open-circuit.

Table A6. Parameters of the TPP.

Parameter Value Parameter Value

SG d axis transient short −
Capacity, Sn 10 MVA 1.01 s
circuit time constant, Td0
SG d axis subtransient short −
Nominal voltage (LL, RMS), Vn 690 V 00 0.053 s
circuit time constant, Td
SG q axis subtransient open −
Nominal current (RMS), In 8367.4 A 00 0.1 s
circuit time constant, Tqo
Nominal frequency, f n 50 Hz SG stator resistance, Rs 0.0028544 pu
SG Number of pole pairs 1 SG friction factor, F 0
SG d axis synchronous reactance, Xd 1.305 pu SG inertia constant, H_SG 4s
SG d axis transient reactance, Xd0 0.296 pu Droop coefficient, Dr_SG 0.05
00
SG d axis subtransient reactance, Xd 0.252 pu Governor time constant, Tg 0.2 s
SG q axis synchronous reactance, Xq 0.474 pu Steam turbine time constant, Tch 0.3 s
00
SG q axis subtransient reactance, Xq 0.243 pu SG leakage reactance, Xl 0.18 pu

References
1. Komusanac, I.; Brindley, G.; Fraile, D.; Ramirez, L. Wind Energy in Europe: 2020 Statistics and the Outlook for 2021–2025; WindEurope:
Brussels, Belgium, 2021; Available online: https://windeurope.org/intelligence-platform/product/wind-energy-in-europe-in-
2020-trends-and-statistics/ (accessed on 30 January 2022).
2. Morren, J.; Pierik, J.; de Haan, S.W.H. Inertial response of variable speed wind turbines. Electr. Power Syst. Res. 2006, 76, 980–987.
[CrossRef]
3. Morren, J.; de Haan, S.W.H.; Kling, W.L.; Ferreira, J.A. Wind turbines emulating inertia and supporting primary frequency control.
IEEE Trans. Power Syst. 2006, 21, 433–434. [CrossRef]
4. Teodorescu, R.; Liserre, M.; Rodriguez, P. Grid Converters for Photovoltaic and Wind Power Systems, 1st ed.; John Wiley & Sons, Ltd.:
New York, NY, USA, 2011.
5. Rakhshani, E.; Rodriguez, P. Inertia Emulation in AC/DC Interconnected Power Systems Using Derivative Technique Considering
Frequency Measurement Effects. IEEE Trans. Power Syst. 2017, 32, 3338–3351. [CrossRef]
Energies 2022, 15, 1922 12 of 12

6. Pattabiraman, D.; Lasseter, R.H.; Jahns, T.M. Impact of Phase-Locked Loop Control on the Stability of a High Inverter Penetration
Power System. In Proceedings of the 2019 IEEE Power & Energy Society General Meeting (PESGM), Atlanta, GA, USA, 4–8
August 2019; pp. 1–5. [CrossRef]
7. Durrant, M.; Werner, H.; Abbott, K. Model of a VSC HVDC terminal attached to a weak AC system. In Proceedings of the 2003
IEEE Conference on Control Applications, Istanbul, Turkey, 25 June 2003; pp. 173–177. [CrossRef]
8. Harnefors, L.; Bongiorno, M.; Lundberg, S. Input-admittance calculation and shaping for controlled voltage-source converters.
IEEE Trans. Ind. Electron. 2007, 54, 3323–3334. [CrossRef]
9. Wang, X.; Taul, M.G.; Wu, H.; Liao, Y.; Blaabjerg, F.; Harnefors, L. Grid-Synchronization Stability of Converter-Based Resources—
An Overview. IEEE Open J. Ind. Appl. 2020, 1, 115–134. [CrossRef]
10. Rosso, R.; Andresen, M.; Engelken, S.; Liserre, M. Analysis of the Interaction among Power Converters through Their Synchro-
nization Mechanism. IEEE Trans. Power Electron. 2019, 34, 12321–12332. [CrossRef]
11. Wu, W.; Zhou, L.; Chen, Y.; Luo, A.; Dong, Y.; Zhou, X.; Xu, Q.; Yang, L.; Guerrero, J.M. Sequence-Impedance-Based Stability
Comparison between VSGs and Traditional Grid-Connected Inverters. IEEE Trans. Power Electron. 2019, 34, 46–52. [CrossRef]
12. Rocabert, J.; Luna, A.; Blaabjerg, F.; Rodríguez, P. Control of power converters in AC microgrids. IEEE Trans. Power Electron. 2012,
27, 4734–4749. [CrossRef]
13. Elkhatib, M.E.; Du, W.; Lasseter, R.H. Evaluation of Inverter-based Grid Frequency Support using Frequency-Watt and Grid-
Forming PV Inverters. In Proceedings of the 2018 IEEE Power & Energy Society General Meeting (PESGM), Portland, OR, USA,
5–10 August 2018; pp. 1–5. [CrossRef]
14. Pattabiraman, D.; Lasseter, R.H.; Jahns, T.M. Comparison of Grid Following and Grid Forming Control for a High Inverter
Penetration Power System. In Proceedings of the 2018 IEEE Power & Energy Society General Meeting (PESGM), Portland, OR,
USA, 5–10 August 2018; pp. 1–5. [CrossRef]
15. Poolla, B.K.; Groß, D.; Dörfler, F. Placement and Implementation of Grid-Forming and Grid-Following Virtual Inertia and Fast
Frequency Response. IEEE Trans. Power Syst. 2019, 34, 3035–3546. [CrossRef]
16. Delavari, A.; Kamwa, I.; Brunelle, P. Simscape Power Systems Benchmarks for Education and Research in Power Grid Dynamics
and Control. In Proceedings of the 2018 IEEE Canadian Conference on Electrical & Computer Engineering (CCECE), Quebec, QC,
Canada, 13–16 May 2018; pp. 1–5. [CrossRef]
17. Chung, S.K. A phase tracking system for three phase utility interface inverters. IEEE Trans. Power Electron. 2000, 15, 431–438.
[CrossRef]
18. Golestan, S.; Monfared, M.; Freijedo, F.D.; Guerrero, J.M. Performance improvement of a prefiltered synchronous-reference-frame
PLL by using a PID-type loop filter. IEEE Trans. Ind. Electron. 2014, 61, 3469–3479. [CrossRef]
19. Saborío-Romano, O.; Göksu, Ö. Performance of Ancillary Services Provision from WFs Connected to DR-HVDC. 2018. Available
online: https://www.promotion-offshore.net/fileadmin/PDFs/D3.5_PROMOTioN_Performance_of_ancillary_services_pro-
vision_from_WFs_connected_to_DR-HVDC.pdf (accessed on 30 January 2022).
20. Tayyebi, A.; Grob, D.; Anta, A.; Kupzog, F.; Dorfler, F. Frequency Stability of Synchronous Machines and Grid-Forming Power
Converters. IEEE J. Emerg. Sel. Top. Power Electron. 2020, 8, 1004–1018. [CrossRef]
21. Lu, L.; Saborío-Romano, O.; Cutululis, N.A. Reduced-order-VSM-based frequency controller for wind turbines. Energies 2021,
14, 528. [CrossRef]
22. Beck, H.P.; Hesse, R. Virtual synchronous machine. In Proceedings of the 2007 9th International Conference on Electrical Power
Quality and Utilisation, Barcelona, Spain, 9–11 October 2007. [CrossRef]
23. Zhong, Q.-C.; Weiss, G. Synchronverters: Inverters That Mimic Synchronous Generators. IEEE Trans. Ind. Electron. 2011, 58,
1259–1267. [CrossRef]
24. D’Arco, S.; Suul, J.A. Virtual Synchronous Machines—Classification of Implementations and Analysis of Equivalence to Droop
Controllers for Microgrids. In Proceedings of the 2013 Grenoble PowerTech, Grenoble, France, 16–20 June 2013; pp. 1–7. [CrossRef]
25. D’Arco, S.; Suul, J.A.; Fosso, O.B. Control system tuning and stability analysis of Virtual Synchronous Machines. In Proceedings of
the 2013 IEEE Energy Conversion Congress and Exposition, Denver, CO, USA, 15–19 September 2013; pp. 2664–2671. [CrossRef]
26. ENTSO-E, Fast Frequency Reserve—Solution to the Nordic Inertia Challenge. 2019. Available online: https://energinet.dk/-/
media/762E225FFB894177A37B75A7CF4C894C.PDF (accessed on 30 January 2022).
27. Aggarwal, G.; Mittal, A.; Mathew, L. MATLAB/simulink model of multi-machine (3-Machine, 9-Bus) WSCC system incorporated
with hybrid power flow controller. In Proceedings of the 2015 Fifth International Conference on Advanced Computing &
Communication Technologies, Haryana, India, 21–22 February 2015; pp. 194–201. [CrossRef]
28. Asija, D.; Choudekar, P.; Soni, K.M.; Sinha, S.K. Power Flow Study and Contingency status of WSCC 9 Bus Test System using
MATLAB. In Proceedings of the 2015 International Conference on Recent Developments in Control, Automation and Power
Engineering (RDCAPE), Noida, India, 12–13 March 2015; pp. 338–342.
29. Anderson, P.M.; Fouad, A.A. Power System Control and Stability; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2002. [CrossRef]