Study of Resonance Issues between DFIG-based

Offshore Wind Farm and HVDC Transmission

Yonggang Zhang, Christian Klabunde, Martin Wolter
Institute of Electric Power Systems
Otto von Guericke University Magdeburg
Magdeburg, Germany
yonggang.zhang@ovgu.de

Abstract—Wide-frequency range of resonances can occur in resonances at hundreds or thousands of Hz [7]. Noted that the
HVDC-connected DFIG-based offshore wind farm (OWF), thus risk of above resonances is susceptible to the variation of grid
induce large harmonic distortion or endanger system stability. topology and operating conditions [8], [9].
Previous studies usually treat OWF as a single-machine system
without considering the switching states of individual wind In recent years, impedance based methods have been
turbines and collecting cables, or ignore the converter responses intensively applied in resonance analysis. Accurate impedance
to wind speed variation and generation control order. In this model of DFIG incorporating double-loop PI control
paper, such factors are included in the modelling of OWF. Their dynamics, mirror frequency coupling and DC side dynamics is
impacts on resonance frequency, amplification level and the built in [10], but the model has complicated analytical
occurrence of negative damping are analyzed applying the representation. It has been proved that neglecting outer control
resonance mode analysis method. Harmonic stability and power loop and DC side dynamics is acceptable for resonance studies
quality issues are demonstrated for the frequency range from down to a few Hz [11]. Zong et al [12] try to build an accurate
several Hz to a few kHz. Simulations in MATLAB/Simulink impedance model for wind farm by considering a unified dq-
validate the analytical results. Additionally, resonance damping frame for all WTs, but the involved phase difference between
strategies are discussed regarding the identified resonances in global and local reference frames varies with operating
widely spread frequencies. conditions and needs to be identified each time when
operating point changes, thus make impedance methods lose
Index Terms—DFIG, HVDC, harmonic stability, power quality,
resonance mode analysis (RMA).
the advantage in convenience. Chen et al [13] compare
existing aggregation methods of wind farm and demonstrates
that neglecting the coupling between positive and negative
I. INTRODUCTION sequence impedances may not be effective for quantitative
Resonance issues in converter-interfaced power systems stability analysis, but it is useful for identifying the resonance
have attracted intentions in industry and academia for years. points of wind farm and analyzing their risk qualitatively.
For the application of HVDC connected OWF employing From the above, it is feasible for this work to simplify the
DFIG-based wind turbines (WTs), the interaction among impedance modelling of DFIG-based WT and OWF while
converter-interfaced grid components and with other passive maintaining their accuracy for resonance analysis.
grid components may introduce wide frequency range of Generally, Nyquist and Bode-plot methods are applied to
resonances. The stimulation of such resonances may cause the study resonances at the point of common coupling (PCC),
violation of harmonic distortion limits as stipulated in IEEE which requires equalizing a system into load and source
Std 519-2014 [1] or incur instability when under poor or subsystems [14], [15]. Since the severest harmonic distortion
negative damping conditions. Possible consequences are the can occur at any bus of a system, simply investigating the
tripping of turbines, blocking of HVDC converter, or physical PCC condition is obviously not in favor of identifying
damages of sensitive grid components. Field experiences on resonance source and developing mitigation strategies [16].
such issues are reported in [2]-[4]. Therefore, this work adopts an impedance network method,
When connected with HVDC transmission, DFIG-based i.e. resonance mode analysis (RMA), to address the resonance
WTs are prone to sub-synchronous resonance (SSR) due to the issues related to wind speed variation and grid topology
inductive and capacitive (LC) interaction at SSR frequencies change. The locations where resonances can be most easily
[5]. Damping of the SSR is typically low or negative due to excited are identified through bus participation factor analysis
the negative resistance induced by the negative slip ratio of the [17]. Additionally, traditional impedance-based RMA only
induction generator and converter controls [6]. For large-scale evaluates the amplitude of modal impedance, this work also
OWFs, the widely distributed WT collecting cables have large uses the damping information embedded in phase angle or the
cumulative capacitive effect, thus strongly participate in the real part of modal impedance in assessing stability.

Y. Zhang thanks for the financial support from China Scholarship Council
(CSC File No.201608320229).

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
 The rest of this paper is organized as follows: Section II Since the series impedance of each cable section is small,
introduces the investigated system and its small-signal it is reasonable to assume that in-service wind turbines operate
impedance modelling. Section III presents the impact of OWF under same conditions, i.e. with same terminal voltage and
operating conditions on the risk of resonances and a current injection. Then to relieve computational burden, each
discussion on damping strategies. Section IV concludes the medium-voltage (MV) collector cable system is equalized into
work. a single cable applying the equivalent power-loss method [5],
[19]. Series impedance and shunt admittance of the adopted pi
II. SYSTEM DESCRIPTION AND IMPEDANCE MODELLING model are represented by (1) and (2), where Rm, n , Lm, n and
This paper investigates the resonances of an offshore grid Cm , n denote the parameters of the nth cable section in array m,
composed by a 400 MW DFIG-based OWF and a ±150 kV M denotes the total amount of arrays, and Nm denotes the
HVDC connection. Typical German offshore grid amount of WTs in array m. Parameters of the aggregated WT
configuration is adopted, as shown in Fig. 1 [2], [3]. On the are linearly related to power rating, which can be easily
left side are two 33 kV collector systems, each gathering 40 obtained by dividing/multiplying wind turbine number, as
WTs through 6 wind arrays, in which there are 2 arrays, each given in Table II. When the number of operating WTs varies
consisting of 4 WTs, and there are 4 arrays, each consisting of or grid topology changes, the aggregated models can be
8 WTs. In the middle are two OWF platforms, which are adjusted accordingly.
connected to the right-side HVDC platform through short 155 Nm 2
kV AC cables. This offshore system is connected to main
M
M 
onshore grid through the HVDC link. Setup of collector cables
Z mv ( s )
= ∑∑ n ( R
m 1=
= n 1
2
m,n + sLm , n )  ∑ N m 
 =m 1 
(1)
is given in Fig. 1. Cable data is estimated from ABB user's M Nm
guide for submarine cable systems [18], as presented in Table Ymv ( s ) = ∑∑ sCm , n (2)
I. m 1=
= n 1

2 × 4 WTs + 4 × 8 WTs
AC Aggregated WTs N wt (quantity)
33 kV
OWF
AC cable Onshore Rated Voltage / Power 950 V / 5 MW ⋅ N wt
Wind array 1 platform1 HVDC
155 kV DC cable
platform 300 kV DFIG generator (pu) =Rs 0.023,
= Lσs 0.18,
= Rr 0.016,
= Lσr 0.16,
= Lm 2.9

Wind array 6 8 WTs

~= ~= GSC side reactor (pu)
= Rp 0.003,
= Lp 0.3
PoC
WFVSC GSVSC AC Grid Current controllers GSC:
= K gp 0.83,
= K gi 5, RSC:
= K rp 0.15,
= K ri 2
2 × 4 WTs + 4 × 8 WTs PCC
AC Offshore Terminal filter 600 kvar ⋅ N wt , Q=50 (quality factor)
33 kV
PoC
OWF AC cable
Wind array with 4 wind turbines
Step-up transformer 5.5 MVA ⋅ N wt , 0.95 / 33 kV
= , uk 5%,
= ur 0.17%
Wind array 7 platform2 155 kV
Sec.1 Sec.2 Sec.3 Sec.4
150mm 2 150mm 2 240mm 2 240mm 2 Sampling period Ts= 5 µs
Wind array with 8 wind turbines PoC Switching frequency 1600 Hz
Wind array 12 8 WTs
PoC
Sec.1 Sec.2 Sec.3
150mm 2 150mm 2 240mm 2
Sec.4
240mm 2
Sec.5
500mm 2
Sec.6 Sec.7
500mm 2 500mm 2
Sec.8
500mm 2 PLL K pp 16,
= = K pi 50
MV Collector Systems
TABLE II. PARAMETERS OF SINGLE OR AGGREGATED DFIG-BASED WT
Fig. 1 Structure of the investigated system
Aggregated representation of the investigated system is
Cable Type Cross section Length Electrical Parameters per km shown in Fig. 2, in which grid parameters and the control
1 150 mm2 1 km R= 0.16 Ω, L= 0.41 mH, C= 0.21 μF blocks of DFIG converters and HVDC converter are included.
For DFIG-based wind turbine, there are two back-to-back
2 240 mm2 1 km R 0.098 Ω, =
= L 0.38 mH, C
= 0.24 μF
converters and an induction machine. Grid side converter
3 500 mm2 1 km R= 0.05 Ω, L= 0.34 mH, C= 0.32 μF
(GSC) controls DC side voltage, while rotor side converter
TABLE I. PARAMETERS OF COLLECTOR CABLES (RSC) controls rotor speed and reactive power.
Offshore Wind Farm Offshore AC-grid HVDC Converter Station
P* , p Speed
*
ird
control + H ri ( s ) +
ωr* , ωr +
θm WT1 30 Mvar 10 Mvar =f sw 1350
= Hz, Ts 5 μs
PWM 40 x 5 MW B2 Q=15 Q=15
θ PLL − θ m ird Decoupling mrd B8 – B6 – B4 155 kV Cable
abc term dq RSC Another DFIG branch 27th 54th
WFVSC
ir,abc dq irq abc 420 MVA Z VSC
Decoupling mrq Active 155 / 150 kV
Q* , q *
irq
term
damper uo,abc ic,abc Rc Lc uc ,abc
var/volt + H ri ( s ) + θ PLL − θ m ir,abc 155 kV Cable
udc
us,abc control 33 kV Cable (4 km)
is,abc Z WT Rmv 0.034 Ω
= 210 MVA Rhv 0.089 Ω
=
0.14% pu 0.14 pu
Phase reactor
210 MVA Lmv = 0.191 mH B5 33 / 155 kV Lmv = 0.36 mH B1
udc usd ig,abc 0.95 / 33 kV B7 B3 uk= 14%
* iwt,abc Cmv = 10.52 μF Cmv = 0.23 μF ur = 0.3% uod Active
damping
U dc *
igd +
+ H dc ( s ) + H gi ( s ) us,abc
*
uod
+ Lg + H v ( s) + PWM
AC voltage
θ PLL uk = 12% +
igd mgd GSC uk = 5% ur = 0.5% * controller:
abc Lg
abc ur = 0.17% θ = 2π f1t abc uod icd ucd
ig,abc abc Lc abc K vp = 0.1,
dq PWM icq
Lg dq θ
igq mgq RC Filter, 40 x 0.6 Mvar, Q=30 Fixed clock dq
uoq dq Lc *
dq K = 15,
usd signal ucq vi

* + H gi ( s )
θ PLL abc us,abc + H v ( s) + θ K d = 0.14
igq =0 + u *
θ PLL usq dq oq
usq H PLL ( s ) uoq Active
damping

Fig. 2 Aggregated representation of the investigated HVDC connected DFIG-based OWF

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
 Rg + Lg s −  H gi ( s − jω1 ) − jK gd  Gd ( s − jω1 )
Z gp ( s ) = (3)
{
1 − Gd ( s − jω1 ) + TPLL ( s − jω1 ) Gd ( s − jω1 )  H gi ( s − jω1 ) − jK gd  I g1 / 2 + U g1 / 2 }
Rg + Lg s −  H gi ( s + jω1 ) + jK gd  Gd ( s + jω1 )
Z gn ( s ) = (4)
{ *
1 − Gd ( s + jω1 ) + TPLL ( s + jω1 ) Gd ( s + jω1 )  H gi ( s + jω1 ) + jK gd  I g1 *
/ 2 + U g1 /2 }
Rs + Rr' σ p ( s ) + ( Lls + L'lr ) s + H ri ( s − jω1 )Gd ( s − jω1 ) σ p ( s ) simulation impedance responses well. The adopted two-level
Z rp ( s ) = (5)
1 − TPLL ( s − jω1 )  I r1 H ri ( s − jω1 )Gd ( s − jω1 ) σ p ( s ) − U r1  / 2
WFVSC shows capacitive characteristic at sub-synchronous
frequencies, and such observations can be found for MMC-
Rs + Rr' σ n ( s ) + ( Lls + L'lr ) s + H ri ( s + jω1 )Gd ( s + jω1 ) σ n ( s ) (6) based HVDC converter using the same control strategy [20],
Z rn ( s ) =
1 − TPLL ( s + jω1 )  I r1* H ri ( s + jω1 )Gd ( s + jω1 ) σ n ( s ) − U r1*  / 2 [21], only not for the whole sub-synchronous frequency area.
5 MW wind turbine HVDC converter
Typical double-loop PI controller are applied under 60 60

synchronous reference frame (SRF). Phase-locked loop is used 40

positive
40

to synchronize WT terminal voltage with grid voltage.

sequence
20 20

Amplitute in dB
Neglecting the slow outer loop control dynamics, DC side 0 0

dynamic and sequence-domain impedance coupling, the -20 -20

positive
positive and negative sequence impedances of GSC and RSC
negative negative
-40 -40
sequence sequence sequence

branches are derived applying harmonic linearization [11], as -60 -60

represented by (3) – (6), where 100 100

H gi (=
s ) K gp + K gi / s , H ri (=
s ) K rp + K ri / s ,
Angle in degree
0 0

σp (s) = ( s − jωr ) / s , σn (s) = ( s + jωr ) / s , -100 -100

K ppll + K ipll / s H PLL ( s )

H PLL ( s ) = , TPLL ( s ) = -200 -200

s 1 + U g1H PLL ( s ) 10
1
10
2
10
3
10
1
10
2
10
3

Frequency in Hz Frequency in Hz

I g1 , U g1 are the fundamental components of GSC terminal

Fig. 3 Validation of sequence-domain impedance responses. Solid lines:
current and voltage, and I r1 , U r1 are the fundamental analytical models; Asterisks and circles: numerical simulation results
components of RSC terminal current and voltage. Complete
impedances of the simulated WT, i.e. ZWT in both positive-and As for passive grid components, all cables are represented
negative-sequence as illustrated in Fig. 2, can be obtained by by pi models, long-line correction is not needed due to the
solving parallel circuit impedance. short length of each cable section. Transformers are modelled
with short circuit impedances, and passive filter impedances
For HVDC connection, two-level VSC based converters are calculated according to their capacities and quality factors
are adopted here. Grid side VSC converter (GSVSC) controls [20]. Additionally, the accuracy of the aggregated model of
the voltage of DC link, and constant DC voltage is assumed MV collector system is acceptable for resonance analysis, as
due to the connection with strong onshore grid. Wind farm validated in [5]. Note that all passive grid components have
side VSC converter (WFVSC) is regulated with a single-loop same positive- and negative sequence impedance models.
PI controller to maintain constant AC voltage for the
connected wind farm [15], [20], [21]. A fixed clock signal is III. RESONANCE ANALYSIS
used to generate reference frequency for the regulated AC Based on the sequence impedances of each grid
voltage. Applying harmonic linearization [22], the AC side component and the aggregation of MV collector system, the
sequence impedances of WFVSC as illustrated in Fig.2 can be three-phase test grid in Fig. 2 can be analyzed with its single-
derived as line sequence-component circuits. Without losing generality,
Rc + sLc − jK d K pwmGd ( s − jω1 ) this paper focuses on the resonances in positive-sequence
Z VSC,p ( s ) = (7) system. According to basic circuit theory, the 8-bus
1 + K pwm H v ( s − jω1 ) Gd ( s − jω1 )
impedance network in Fig. 2 can be formulated into an 8-
Rc + sLc + jK d K pwmGd ( s + jω1 ) dimensional nodal admittance matrix YNN ( s ) that relates the
Z VSC,n ( s ) = (8)
1 + K pwm H v ( s + jω1 ) Gd ( s + jω1 ) voltages and currents of the buses from B1 to B8, as
represented by
where Gd =( s ) 1 / (1 + 1.5Ts s ) , H v (=
s ) K vp + K vi / s and K pwm is
the gain of the PWM unit.  I B1  Y11 Y12  Y18  U B1 
 I  Y Y  Y  U 
The derived analytical impedances of single 5 MW DFIG-  B2  =  21 22 28   B2 
(9)
based WT and HVDC converter are validated through           
numerical simulations, as presented in Fig. 3. As observed, the     
 I B8  
Y81 Y82  Y88  U B8 
analytical models of both grid components match their 
YNN ( s )

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
 Inversing and decomposing YNN ( s ) according to matrix damping at SSR frequencies [6], the buses B7 and B8 where
decoupling theory [17] yields the eigenvalue-based impedance DFIG-based WTs are located have larger impact on the risk of
matrix, also called modal impedance matrix this SSR mode. For the 301 Hz, 413 Hz, 2120 Hz and the
2179 Hz resonance modes, the buses B5, B6, B7 and B8 in
 Z m1 0  0  two MV collector systems have the largest PFs, thus are the
 0 main resonance sources. For the 530 Hz and 1026 Hz
Z m2  0  resonance modes, the buses B2, B3 and B4 where 155 kV AC
Z M ( s ) T= YNN ( s ) T 
−1 −1
= (10)
      cables are located play an important role, besides, the
  switching harmonic filters of HVDC converter in bus B1 also
 0 0  Z m8 
participate the 530 Hz resonance. Noted that the PFs in Table
where T is the right eigenvector matrix of YNN ( s )
−1
, III are only valid for the operating condition when all WTs are
in-service and generate full power.
Z m1 , Z m2 to Z m8 are system modal impedances. Fig. 4 shows
the frequency scanning to the obtained modal impedances. A. Impact of WT Rotor Speed
The peaks in magnitude plot indicate resonances. When the WT usually works under the maximum power point
damping of a resonance mode is negative, i.e. corresponding tracking mode. The variation of wind speed to a lower value
modal impedance with a negative real part, system stability is may cause WT output power decreases, and as follows, rotor
at risk. The negative damping is also indicated in the phase speed will be decreased from its nominal value (1.2 pu). This
plot by a larger than 180o phase variation around resonance behavior is included in the detailed DFIG model by following
frequency. For modal impedance with positive damping, a
larger peak value in amplitude plot means greater harmonic ω*r =
− 0.67 P 2 + 1.42 P + 0.51 (11)
amplification effect at resonance frequency. as speed reference, where P denotes the feed-in active power
from WT [25]. The speed reference ω*r slowly tracks changes
in power P with a time constant of approximately 5 seconds.
Fig. 5 shows the impact of rotor speed on the modal
impedances of the investigated system. Only SSR mode is
displayed as the influence on other resonance modes is not
observable. As rotor speed decreases, extra negative resistance
will be introduced into WT at SSR frequencies. When the total
resistance of the system becomes negative, SSR stability will
be endangered. This is what happens under the conditions of
rotor speed as 0.8 pu and 0.7 pu in Fig. 5. The larger than 180º
phase variation around the SSR frequency indicates a negative
part of the modal impedance at resonance frequency, i.e.
negative damping.
Fig. 4 Frequency scanning to the modal impedances of the test grid Rotor speed: 1.2 pu 1.05 pu 0.9 pu 0.8 pu 0.7 pu

20

20

After obtaining resonance frequencies and impedances, the 10

Imaginary part in pu

15
Amplitude in pu

most influential buses to each resonance mode could be

determined through eigenvalue decomposition [17], [24]. The
10
0

evaluation index is defined as participation factor (PF), as 5

listed in Table III.

-10
0
200 20

100 10

Modal fres in Hz 10.5 301 413 530 1026 2120 2179

Angle in degree

Real part in pu

Res. |Zm| in pu 5.7 31 7.8 12.8 59 45 49 0 0

B1 25 3 5 14 0 0 0 -100 -10

B2 15 10 9 16 31 0 0
B3 15 11 9 15 31 0 1
-200 -20
PF
B4 15 11 9 15 31 0 1
5

9
10

11

12

13

14

15

10

11

12

13

14

15

in % B5 8 16 17 9 3 27 26
Frequency in Hz Frequency in Hz

B6 8 16 17 9 3 27 26 Fig. 5 Impact of rotor speed ωr on system modal impedances

B7 7 17 18 11 2 23 23
B8 7 17 18 11 2 23 23
In order to validate this analytical result, simulation
TABLE III. PARTICIPATION FACTORS OF DOMINATING RESONANCE MODES models are built up in MATLAB/Simulink, where the DFIG
system is taken from the standard simulation model developed
For the 10.5 Hz resonance mode, the buses B1, B2, B3 and by MATLAB [25], only the change of parameters and minor
B4 in 155 kV high-voltage side contributes mostly to the modification of control structure are made. In time-domain
formed LC resonance circuit, but as the source of negative simulation, wind speed is varied from 15 m/s to 5 m/s at t=6 s.

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
The variation of wind speed and outputs of the aggregated Apart from the variation of wind speed, wind turbine
wind turbine WT1 are shown in Fig. 6. Diverging oscillation output power could also be changed when wind farm receives
is observed when rotor speeds drops to be around 0.8 pu. a lowering power generation command from its superior level
control center, then as a consequence, rotor speed might be
regulated down and risk SSR stability.

B. Impact of the Variation of In-service Wind Turbines

In real life wind farm, individual WTs can be disconnected
from grid by switching off their terminal breakers, or arrays of
WTs can be disconnected from grid at the 33 kV AC bus, as
shown in Fig. 8. Their impact on resonances will be addressed
in this section separately.
Cable Sec.1 Cable Sec.2 Cable Sec. N1
33 kV
Disconnect Array 1 AC Bus
at WT
WT 155 kV
WT WT Disconnect at
AC Bus
33 kV AC bus
Cable Sec.1 Cable Sec.2 Cable Sec. N m HV Cable

Array m
Another
WT
WT WT branch

Cable Sec.1 Cable Sec.2 Cable Sec. N 6

Array 6

WT
WT WT

Fig. 8 Schematic diagram of WT disconnection at different locations

1) Disconnection of wind turbines at 33 kV AC bus

The change of wind array switching state can happen when
wind farm is started or stopped, or when a lower generation
control order is received from its superior level control center.
The impact on resonance is shown in Fig. 9. At SSR
frequencies, the phase variation around resonance frequency is
far less than 180o for each scenario, so the impact on SSR
stability is quite small. Besides, the amplitudes of modal
impedances are no larger than 10 for all scenarios, thus the
Fig. 6 Time-domain simulation outputs of the aggregated wind turbine WT1 impact on amplification effect is neglectable. At frequencies
above 200 Hz, negative damping has not been observed for
Furtherly, FFT analysis is conducted to the voltage of the
each resonance mode, so only using the amplitude plot is
155 kV AC bus B2, which interconnects OWF and HVDC.
sufficient for evaluating resonance amplification effect. The
The results are shown in Fig. 7. Large SSR frequency
arrows in Fig. 9 b) indicate that as less wind arrays are
component is observed at the starting stage of the diverging
connected into grids, the two resonance modes at above 2 kHz
oscillation, and the mirror frequency component around 90 Hz
move from 2 kHz toward 3 kHz, the resonance mode at
is seen in the frequency spectrum.
around 1 kHz moves from 1100 Hz toward 900 Hz, and the
resonance mode at around 300 Hz moves from 300 Hz toward
400 Hz.
6 arrays:40 WTs 3 arrays: 24 WTs 1 array: 8 WTs (at each MV bus)

6 4
Imaginary part in pu

4 2
Amplitude in pu

2 0
Fundamental (50Hz) = 0.9945

15 0 -2
100
Mag (% of Fundamental)

10 4
50
Phase in degree

Real part in pu

5 0 2

0 -50 0
0 10 20 30 40 50 60 70 80 90 100
5

9
10

11

12

13

14

15

10

11

12

13

14

15

Frequency (Hz)
Frequency in Hz Frequency in Hz

Fig. 7 FFT analysis to the voltage of 155 kV AC bus B2 a) At SSR frequencies

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
 Brk2 Brk3 Brk4
DFIG

Brk1 Cf PoC
Lg Rf

RSC GSC

Fig. 11 Single line circuit of the simulated DFIG-based WT

Assuming that the disconnection of WTs does not

influence the switching states of collector cables, then the
b) At frequencies above 200 Hz
impact of the number of connected WTs on modal impedances
Fig. 9 Frequency scanning of modal impedances when the number of are shown in Fig. 12, both the situations of opening
connected wind arrays are varied transformer breakers and opening GSC and RSC branch
Stimulation of the resonance at around 350 Hz is validated breakers are considered. Since their impact on the SSR
by time-domain simulation. As shown in Fig. 10, after resonance mode is very small, it is not displayed in the figure.
switching off 5 of the 6 wind arrays in each MV collector
system at 4 s, the 350 Hz voltage component on 155 kV AC
bus is greatly amplified. According to the recommended
requirements on harmonic distortions in [1], the 350 Hz
harmonic component violates the limit of 1.5% of
fundamental value, and the total harmonic distortion (THD) of
the 155 kV AC bus voltage violates the limit of 2.5% of
fundamental value.

a) WT disconnected by opening transformer breaker

Fundamental (50Hz) = 0.9911 , THD= 1.39% Fundamental (50Hz) = 1.014 , THD= 3.19%
2.5 2.5
350Hz
2 2

b) WT disconnected by opening GSC and RSC branch breakers

Mag (% of Fundamental)
Mag (% of Fundamental)

1.5 1.5

1 1 Fig. 12 Impact of the number of connected WTs on modal impedances

0.5 0.5
In case WT is disconnected from grid by opening one of
0
0 500 1000 1500 2000 2500 3000
0
0 500 1000 1500 2000 2500 3000
the transformer breakers, the scenarios of 80, 50, 20 and 6
Frequency (Hz) Frequency (Hz) WTs staying connected in grid are shown in Fig.12 a). When
Fig. 10 Waveforms of the 155 kV AC bus voltage and FFT analysis only 6 WTs in the OWF staying connected, the resonance
mode at around 300 Hz moves to be around 350 Hz and have
2) Disconnection of wind turbines at their terminals a very sharp peak. The two resonance modes at above 2 kHz
Wind turbines can be disconnected from grids due to move toward 1 kHz as the number of grid-connected WTs
maintenance, fault or wind curtailment reasons. By quite low reduces. When only 6 WTs is connected in grid, the large
or high wind speed conditions, all WTs can be cut off from resonance peak between 1400 Hz and 1500 Hz falls in the area
grids. The disconnection of WT can be operated through of the switching sideband frequency of HVDC converter
opening the high or low voltage side breakers of their 950 V / (1450 Hz), thus amplification of switching harmonics from
33 kV step-up transformers or the breakers in RSC and GSC HVDC converter will happen. Stimulation of this resonance is
branches, as shown in Fig.11. Opening transformer low- observed in time-domain simulation, as shown in Fig.13. Both
voltage side breaker has almost the same effect as opening 350 Hz and 1450 Hz voltage components violate the
transformer high-voltage side breaker, as transformer shunt requirements on distortion limits as specified in [1].
impedance of the excitation branch is far bigger than its series In case WT is disconnected from grid by opening GSC and
impedances in the primary and secondary windings, there is RSC branch breakers, as shown in Fig.12 b), the number of in-
almost no current flowing into a transformer once one side is service WTs has the greatest impact on the 301 Hz resonance
opened. mode, while the impact on other resonances modes, especially

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
those at above 1 kHz, are negligible. As the number of grid- IV. CONCLUSION
connected WTs decreases to be 50 or down to 20, the 301 Hz This paper presented impedance-based resonance analysis
resonance will be moved to be around 250 Hz, and the typical for an exemplary DFIG-based offshore wind farm with HVDC
5th order harmonic can be slightly magnified. Since the overall connection. Derived impedance models of DFIG-based wind
amplification effect impact in this case is small, no simulation turbine and HVDC converter were validated through
result will be presented. numerical simulations. The aggregated model of medium-
voltage collector system is adjustable in reflecting the
variation of in-service wind turbines and switching states of
collector cables. The impact of different wind farm operating
conditions on resonance frequency, amplification level and the
occurrence of negative damping were analyzed applying the
RMA method. SSR stability and harmonic distortion issues
were demonstrated for the frequency range from 5 Hz to
3 kHz. Considering the involved wide frequency range of
resonances, the coordination of HVDC converter active
control and high-switching frequency active damper is
suggested to be used in mitigating the found resonances.
Fundamental (50Hz) = 1.022 , THD= 8.47%

6
REFERENCES
Mag (% of Fundamental)

4 [1] IEEE Recommended Practice and Requirements for Harmonic Control

in Electric Power Systems, IEEE Std 519-2014 (Revision of IEEE Std
2 519-1992), vol., no., pp.1-29, 11 June 2014.
[2] C. Buchhagen, C. Rauscher, A. Menze, and J. Jung, “BorWin1 – first
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 experiences with harmonic interactions in converter dominated grids,”
Frequency (Hz) in Proceedings of International ETG Congress 2015, 2015, pp. 1–7.
[3] C. Buchhagen, M. Greve, A. Menze, and J. Jung, “Harmonic stability -
Fig.13 Voltage and current waveforms of the 155 kV AC bus B2 and FFT practical experience of a TSO,” in Proceedings of the 15th Wind
analysis Integration Workshop, 2016, pp. 1–6.
[4] M. Brendel and G. Traufetter, “Knall auf hoher See,” Der Spiegel, no.
C. Discussion of Resonance Damping Strategies 35, pp. 71–72, Aug. 2014.
If defining low frequency range as below 50 Hz, middle [5] Y. Zhang, C. Klabunde and M. Wolter, “Frequency-coupled Impedance
Modelling and Resonance Analysis of DFIG-based Offshore Wind
frequency range as between 50 Hz and 800 Hz, and high Farm with HVDC Connection,” IEEE Access, under review.
frequency range as above 800 Hz, then from the analysis of [6] X. Xie, X. Zhang, H. Liu, H. Liu, Y. Li and C. Zhang, “Characteristic
Section III. A and III. B, it is found resonances can occur at Analysis of Subsynchronous Resonance in Practical Wind Farms
any frequency range. The resonance frequencies of low and Connected to Series-Compensated Transmissions,” in IEEE
middle frequency range does not vary much as wind farm Transactions on Energy Conversion, vol. 32, no. 3, pp. 1117-1126,
operating condition changes, but the resonance frequencies of Sept. 2017.
[7] Y. Zhang, C. Klabunde and M. Wolter, “Harmonic Resonance Analysis
high frequency range may moves down to 1 kHz or up to for DFIG-based Offshore Wind Farm with VSC-HVDC Connection,”
3 kHz or higher. 2019 IEEE Milan PowerTech, Milan, Italy, 2019, pp. 1-6.
[8] M. Quester, V. Yellisetti, F. Loku and R. Puffer, “Assessing the Impact
Low and middle frequency resonances could be damped of Offshore Wind Farm Grid Configuration on Harmonic Stability,”
by active controls of WT or HVDC converters [14], [26]. 2019 IEEE Milan PowerTech, Milan, Italy, 2019, pp. 1-6.
Since the switching frequencies of high power rating WT and [9] H. Jiang, J. You, H. Liu, W. Ning and L. Wu, “Impedance
HVDC converters are usually not greater than 2 kHz, their characteristics of DFIGs considering the impacts of DFIG numbers and
capability of controlling high frequency resonances are locations and its application on SSR analysis,” 2017 IEEE Power &
Energy Society General Meeting, Chicago, IL, 2017, pp. 1-8.
limited, so extra active damper with higher switching [10] C. Zhang, X. Cai, M. Molinas and A. Rygg, “Frequency-domain
frequency could be used to address the high frequency modelling and stability analysis of a DFIG-based wind energy
resonance issue [27]. The resonance detection capability of the conversion system under non-compensated AC grids: impedance
active damper in [27] makes it suitable for handling the modelling effects and consequences on stability,” in IET Power
enormous variation of resonance frequency resulted from Electronics, vol. 12, no. 4, pp. 907-914, 10 4 2019.
[11] I. Vieto and J. Sun, “Sequence Impedance Modeling and Analysis of
operating condition change. Considering that the configuration Type-III Wind Turbines,” in IEEE Transactions on Energy Conversion,
of active control in DFIG converters for damping low and vol. 33, no. 2, pp. 537-545, June 2018.
middle frequency resonances may involves dozens of WTs, it [12] H. Zong, J. Lyu, X. Cai, C. Zhang, M. Molinas, F. Rao, “Accurate
is less attractive than the damping configuration in HVDC aggregated modelling of wind farm systems in modified sequence
converter. Therefore, the coordination of HVDC converter domain for stability analysis,” Electric Power Systems Research,
Volume 175, 2019.
damping control and active damper may be a feasible solution
[13] L. Chen, H. Nian and Y. Xu, “Impedance Aggregation Method of
for mitigating the resonances ranging from several Hz to a few Multiple Wind Turbines and Accuracy Analysis,” Energies, 2019, 12,
kHz. 2035.

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
[14] M. Amin, and M. Molinas. “Understanding the Origin of Oscillatory Integration,” in IEEE Journal of Emerging and Selected Topics in
Phenomena Observed Between Wind Farms and HVdc Systems,” IEEE Power Electronics, vol. 6, no. 1, pp. 40-53, March 2018.
Journal of Emerging and Selected Topics in Power Electronics, vol. 5, [22] M. Cespedes and J. Sun, “Impedance modeling and analysis of grid
no. 1, pp.378–392, Mar. 2017. connected voltage-source converters,” IEEE Trans. Power Electron.,
[15] J. Lyu, X. Cai and M. Molinas, “Frequency Domain Stability Analysis vol. 29, no. 3, pp. 1254–1261, Mar. 2014.
of MMC-Based HVdc for Wind Farm Integration,” in IEEE Journal of [23] W. Xu, “Component Modeling Issues for Power Quality Assessment, ”
Emerging and Selected Topics in Power Electronics, vol. 4, no. 1, pp. in IEEE Power Engineering Review, vol. 21, no. 11, pp. 12-17, Nov.
141-151, March 2016. 2001.
[16] Z. Li, H. Hu, L. Tang, Y. Wang, T. Zang and Z. He, “Quantitative [24] C. Amornvipas, “Harmonic Studies and Resonance Analyses in
Severity Assessment and Sensitivity Analysis Under Uncertainty for Electrical Power Systems,” Ph.D. dissertation, Institute of Electric
Harmonic Resonance Amplification in Power Systems,” in IEEE Power Systems, Eng., Univ. Hannover, Shaker Verlag, 2009.
Transactions on Power Delivery. [25] N. W. Miller, J. J. Sanchez-Gasca, W. W. Price and R. W. Delmerico,
[17] Wilsun Xu, Zhenyu Huang, Yu Cui and Haizhen Wang, “Harmonic “Dynamic modeling of GE 1.5 and 3.6 MW wind turbine-generators for
resonance mode analysis,” in IEEE Transactions on Power Delivery, stability simulations,” 2003 IEEE Power Engineering Society General
vol. 20, no. 2, pp. 1182-1190, April 2005. Meeting (IEEE Cat. No.03CH37491), Toronto, Ont., 2003, pp. 1977-
[18] ABB, “XLPE Submarine Cable Systems, Attachment to XLPE Land 1983 Vol. 3.
Cable Systems - User's Guide,” volume Rev 5, 2007. [26] I. Vieto and J. Sun, “Damping of subsynchronous resonance involving
[19] E. Muljadi et al., “Equivalencing the collector system of a large wind Type-III wind turbines,” 2015 IEEE 16th Workshop on Control and
power plant,” 2006 IEEE Power Engineering Society General Meeting, Modeling for Power Electronics (COMPEL), Vancouver, BC, 2015, pp.
Montreal, Que., 2006, pp. 9 pp.-. 1-8.
[20] J. Lyu, X. Cai, M. Amin and M. Molinas, “Sub-synchronous oscillation [27] Y. Zhang, C. Klabunde, M. Wolter, “Harmonic Filtering in DFIG-
mechanism and its suppression in MMC-based HVDC connected wind based Offshore Wind Farm through Resonance Damping,” IEEE ISGT
farms,” in IET Generation, Transmission & Distribution, vol. 12, no. 4, Europe 2019, Bucharest, 2019, pp. 1-5.
pp. 1021-1029, 27 2 2018.
[21] J. Lyu, X. Cai and M. Molinas, “Optimal Design of Controller
Parameters for Improving the Stability of MMC-HVDC for Wind Farm

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020