1858 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO.

2, MAY 2013

General Methodology for Analysis

of Sub-Synchronous Interaction
in Wind Power Plants
Babak Badrzadeh, Senior Member, IEEE, Mandhir Sahni, Yi Zhou, Member, IEEE, Dharshana Muthumuni, and
Aniruddha Gole, Fellow, IEEE

Abstract—This paper presents a general methodology for anal- in the form of sub-synchronous resonance (SSR) which is an
ysis of sub-synchronous interaction in wind power plants. These electromechanical interaction between a series compensated
include appropriate frequency scanning method for the assessment ac line and the rotating masses of the turbine generator. With
of the sub-synchronous control interaction, and calculation of the
electrical damping provided by the wind turbine generator for in-
this definition SSR only applies to type1 wind turbines which
vestigation of the sub-synchronous torsional interaction. A general do not use any power electronic controller during steady-state
formulation of both methods applicable to any given wind turbine conditions (soft starter is only active during the start-up pe-
and network is presented. A dynamic frequency scanning method riod). SSCI, SSTI and SSR are sometimes collectively termed
for the turbine side is developed which takes account of the turbine as sub-synchronous interaction (SSI).
nonlinearities and its active behavior. Various aspects that need A number of technical papers published recently have inves-
to be considered when injecting a voltage or current signal into
the system for dynamic frequency scanning are discussed in detail. tigated the risk of sub-synchronous control interaction in wind
The veracity of these methods is confirmed against electromagnetic power plants. Electromagnetic transient analysis [1]–[3], eigen-
transient analyses. The application of these tools and techniques is value analysis [4], [5], or a combination of the two [4], [5] has
demonstrated on a practical power system comprising type 3 wind been generally adopted.
turbines and series compensated lines. The eigenvalue analysis approach requires a detailed repre-
Index Terms—Electrical damping analysis, electromagnetic sentation of both the mechanical and electrical systems. This
transient simulation, frequency scanning, modeling and simu- implies the need for a very detailed model of mechanical drive
lation, sub-synchronous control interaction, sub-synchronous train including all sub-synchronous torsional modes. A prac-
torsional interaction, wind power plant, wind turbine generator. tical wind turbine generator (WTG) often has several low-order
sub-synchronous torsional modes. Higher order models than the
2nd or 4th order models discussed in the technical literature [4],
I. INTRODUCTION
[5] are therefore required. The model parameters for these high
order models are not readily known.
S UB-SYNCHRONOUS interaction is a relatively newly
experienced phenomena in wind power plants (WPPs)
[1]–[8]. It has been shown that [1] WPPs terminated to series
With exception of [6]–[8], all studies discussed in the tech-
nical literature have been conducted on a small test system
rather than the actual power system to which the wind power
compensated ac lines can exhibit sub-synchronous interaction
plant is connected. When using an actual model of the transmis-
problems in the form of sub-synchronous control interaction
sion system with a large number of lines, identifying the most
(SSCI), or sub-synchronous torsional interaction (SSTI). SSTI
onerous contingencies is of prime importance. For this reason
is an electromechanical interaction between a power electronic
it is essential to first identify a number of critical scenarios
component such as a wind turbine generator (WTG) and the
which will undergo further investigation. Neither an EMT-type
rotating masses of the turbine-generator. An SSCI is a purely
analysis nor an eigenvalue approach is deemed practical for
electrical interaction between a power electronic component
this purpose.
and a series compensated ac line. Compared to the SSTI, SSCI
A precise investigation of SSCI in WPPs integrated into bulk
is a faster developing phenomenon due to the fast action of
power systems require more advanced tools and techniques
the converter control. A third type of interaction can occur
than those generally applicable for the analysis of conventional
power systems without active and nonlinear components. These
Manuscript received June 13, 2012; revised June 30, 2012 and August 27, tools and techniques have been briefly introduced in [6], and
2012; accepted October 08, 2012. Date of publication December 03, 2012; date
of current version April 18, 2013. Paper no. TPWRS-00659-2012. their applications in a practical transmission system with series
B. Badrzadeh is with the Australian Energy Market Operator, Melbourne, compensated lines have been discussed in [7] and [8]. In these
Australia (e-mail: babak.badrzadeh@ieee.org). papers optimal tuning of the frequency scanning and electrical
M. Sahni is with PwrSolutions Inc., Dallas, TX 75207 USA.
Y. Zhou is with Vestas Global Solution and Services, Singapore.
damping analysis methods were achieved using a trial and error
D. Muthumuni is with Manitoba HVDC Research Center, Winnipeg, MB R3P approach. This may not be practicable for other wind turbine
1A3, Canada. designs, and power system configurations. To address these
A. Gole is with the University of Manitoba, Winnipeg, MB R3T 5V6, Canada
Color versions of one or more of the figures in this paper are available online
issues this paper intends to develop a generalized mathematical
at http://ieeexplore.ieee.org. basis for these tools and techniques to streamline their applica-
Digital Object Identifier 10.1109/TPWRS.2012.2225850 tion to any given WTG and power system including multiple

0885-8950/$31.00 © 2012 IEEE

BADRZADEH et al.: GENERAL METHODOLOGY FOR ANALYSIS OF SUB-SYNCHRONOUS INTERACTION IN WIND POWER PLANTS 1859

Fig. 1. Type 3 wind turbine generator model sub-systems.

wind farm applications. Various aspects of the frequency scan- are therefore developed to obviate the need for such data. For
ning and electrical damping analysis methods are elaborated SSCI studies which is a fast growing phenomenon the aerody-
in detail. Validation of these methods against the EMT-type namics and pitch controller can be neglected. For SSTI which
analysis is also presented. grows relatively slower such slow developing dynamics may
For SSCI analysis a frequency scanning method suitable for have an impact and should be included in the overall model.
the WTG with an active and nonlinear behavior is proposed. Fig. 2 shows detailed functional block diagram representa-
For SSTI analysis an electrical damping analysis based on ma- tion of the WTG control system as implemented in the PSCAD/
chine speed perturbation is presented. The dynamic frequency EMTDC model of the turbine used for SSI studies. In the figure:
scanning and electrical damping analysis are derived from the UL: stator line voltage
detailed time-domain representation of the wind turbine model UDC: dc-link voltage
described in Section II, and will therefore include the represen- UTQ: q-axis voltage at tertiary winding of the turbine
tation of control systems and system non-linearities in relevant transformer
detail. Some approximations will, however, be introduced when IGDQ,UGDQ: d-q axis grid current and voltage
computing the harmonic impedance and electrical damping IRDQ,URDQ: d-q axis rotor current and voltage
from the detailed EMT model. In both cases the results obtained UGAB,URAB: grid- and rotor-side phase voltages
should therefore be validated against the EMT-type analysis. SV-PWM: space-vector pulse width modulation
Note that aside from one exception which will be highlighted Omega, Gamma: angular velocity and angular displace-
all studies discussed in this paper are carried out with a type 3, ment of the rotor
1.8 MW, 60 Hz turbine. Looking at Fig. 2 the following sub-systems can be identified:
• Rotor-side converter control
• Rotor side current control
II. WIND POWER PLANT MODELING METHODOLOGY
• Active/reactive power control
In general, a wind turbine generator model comprises the fol- • Power control
lowing sub-systems: • Slope limiters
• aerodynamics; • Filters
• pitch control; • Drive train torsional damping
• mechanical drive train model; • Grid-side converter control
• electrical generator; • Phase-locked loop (PLL)
• measurements; • DC-link voltage control
• converter control. • DC-link current control
The inter-relationship between these sub-systems is shown • Low voltage ride-through control logic (LVRT)
in Fig. 1 where subscripts s, r, el, and LS denote stator, rotor, Additionally correct representation of the turbine protection
electrical and low-speed shaft. The WTG model adopted in this system including the dc choppers is necessary for SSI studies.
paper is simplified such that the mechanical drive train charac- The kHz range PWM switching scheme does not have any
teristics are not included. This is because as discussed in the In- impact on SSI phenomena. To reduce the simulation time an
troduction sufficiently accurate data for the shaft drive train are average switch representation of the voltage source converters is
not often available. The methodologies proposed in this paper adopted instead of the actual converter PWM switching logics.
1860 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

Fig. 2. Converter control model for type 3 turbine used for SSI studies.

Fig. 3. Model of the rotor-side converter control for a type 3 wind turbine including drive train torsional damping.

As an example a detailed control representation of the voltage event is detected. During this period the drive train tor-
rotor-side converter control is shown in Fig. 3. This includes sional damping control is disabled, and rotor current control is
the inner-loop rotor-current control, and outer-loop power given new reference values as determined by the LVRT logic.
controller which in turn consists of active/reactive power To realize the full WPP model an aggregate model of the
controllers, drive train torsional damping, and the power filters. wind turbines is adequate. In majority of practical installations
The PQ filters assure that the power reference for the PQ the park level controller is designed to respond within several
controller will not have discrete jumps in case of a sudden seconds, and can therefore be neglected for SSI studies. The
change in external power reference. The function of drive train system under consideration does not include any STATCOM or
damping is to damp out the generator speed oscillations near other dynamic reactive power support equipment, but as a gen-
resonance frequency of the drive train by injecting power into eral rule such fast acting equipment should be included in the
the grid. This function is included by adding a component onto model whenever applicable.
the active power control output as shown in Fig. 3. As seen in
Fig. 3 the drive train damping has two blocks. The first block III. STUDY TOOLS AND TECHNIQUES
is a high-pass filter acting on the speed difference between the
turbine and the generator. The second block is the drive train A. Frequency Scanning Analysis
damper. The objective of a frequency scanning analysis is to com-
During steady-state conditions the WTG is operated at con- pute the equivalent resistance and reactance looking from the
stant power control with drive train damping being active. The wind turbine into the network, and looking into the wind turbine
generator control switches to current control mode when a low from the network. In the frequency scanning analysis a small
BADRZADEH et al.: GENERAL METHODOLOGY FOR ANALYSIS OF SUB-SYNCHRONOUS INTERACTION IN WIND POWER PLANTS 1861

Fig. 6. Harmonic voltage injection with the WTG as the point of excitation and
Fig. 4. Example of harmonic current injection. measurements.

Using either method the WTG model should be disconnected

from the interconnected network and connected to an ideal
voltage source. This is because in the case of an unbalance in
the network or presence of other power electronic components
non-characteristic harmonics and inter-harmonics can be gen-
erated which may not be readily filtered out. Additionally, the
focus of the investigation is the turbine side impedance scans
whose characteristics are solely dependent on the response of
the WTG and its controls. The remainder of the system can
therefore be represented as an ideal voltage source.
2) Injection Point versus Measurement Point: Having
Fig. 5. Example of harmonic voltage injection. chosen the voltage signal excitation, the next point to consider
is the choice of the injection and measurement points. The first
approach is to apply the excitation signal at the WTG terminal
input voltage (or current) signal is superimposed on the steady- and take the measurements at the same point as shown in
state (fundamental frequency) excitation. Different aspects of Fig. 6. With this arrangement a small-signal voltage or current
the frequency scanning techniques suitable for the WTGs are injected into the system can result in a large disturbance on
discussed below. the steady-state operating point of the system especially the
1) Voltage Injection versus Current Injection: The applica- turbine.
tion of current and voltage injection techniques for determining Another alternative is to inject the signal at a further busbar
the harmonic impedance of the WTG are schematically shown while still making the measurements at the WTG terminals. The
in Figs. 4 and 5, respectively. As an example the figures are application of this method is shown in Figs. 4 and 5 where the
shown with a 34.5/138 kV transformer at the WPP HV substa- injected voltage is applied at the 138 kV transmission line. This
tion. In a current injection method the sub-synchronous current is because injecting the voltage at the WTG terminals would
is injected at the turbine busbar for the entire frequency range of give rise to either steady-state instability of the turbine, or cre-
interest. The measurements can be performed at the same point ates a too small signal which cannot generate the desired har-
as the injection point or at a different point. As an example Fig. 4 monic impedance. To resolve these problems the location of the
illustrates the application of current injection method where the injection was changed while keeping the same magnitude of in-
injection and measurement points are different. This will be jection and keeping the harmonic measurement directly after the
elaborated later in this section. In a voltage injection method a turbine. Note that while there are options in terms of the loca-
small-signal sub-synchronous voltage is injected in series with tion of the injected signal, the measurement has to be performed
the source power system as shown in Fig. 5. The relative mag- at the turbine terminals.
nitude of the injected voltage cannot be directly compared with The methodology shown in Fig. 6 can be extended when sev-
the system voltage. The success criteria for doing so will be dis- eral wind farms are in the vicinity of the series compensated
cussed later in this section. ac line. As an example Fig. 7 shows the application of voltage
In general either a current signal or a voltage signal can be injection method with two wind farms connected to the 138
used as the excitation signal. For the system under consideration kV line. With this configuration both wind farms are located
a voltage injection method is chosen. This has been confirmed at the same side of the series compensated ac line. An onerous
by performing several simulation cases studies which indicate contingency occurring at the transmission network can there-
the following advantages of the voltage injection method: fore simultaneously expose all these wind farms to SSI con-
• a lower degree of sensitivity to variation of the injected ditions. When wind farms are connected at opposite ends of
signal, and hence minimizing the need for a trail and error the series compensated ac line, contingencies that can result in
approach; simultaneous occurrence of SSI conditions on all wind farms
• the need for a smaller excitation signal, and hence creating are not generally considered as credible contingencies. This is
a smaller disturbance on the operating point of the system. because such contingencies would result in disconnection of a
1862 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

This could give rise to differences in precision across the fre-

quency range. Additionally, this method has only been applied
to linear components, and it is not appropriate for calculation of
the harmonic impedance of the wind turbine.
When considering a highly nonlinear and active device such
as a wind turbine, the frequency scanning method utilized must
be able to take account of these nonlinearities. The injected
signal should be a wideband signal such as an impulse or white
noise to avoid the need for multiple simulation runs for each
sub-synchronous frequency. An initial challenge dealing with
the wind turbine or other similar nonlinear elements is that
due to nonlinearities the principle of superposition does not
hold true. The system, however, can still be considered linear
for small changes around an operating point. The harmonic
impedance of a non-linear system, linearized about an oper-
ating point can be determined with an FFT as discussed in
the next sub-section. This forms the basis for the Method 3
frequency scanning technique. Method 3 has been utilized for
all simulations listed in Sections IV and V.
Fig. 7. Harmonic voltage injection method with two adjacent wind farms. 4) Harmonic Impedance Calculation: By measuring the
voltage and current at the measuring point, the impedance seen
from that point at different frequencies can be estimated based
large number of lines where the system integrity cannot be most on Fourier analysis. For example, with the choice of method
likely maintained. shown in Fig. 5 the harmonic impedance of the WTG at the
3) Choice of the Injected Signal Waveform: The most com- sub-synchronous range can be calculated as
monly used frequency scanning techniques include:
• Method 1: Sinusoidal excitation which gives the harmonic
impedance of a linear system at a given frequency
• Method 2: Impulse excitation which gives the harmonic
impedance of a linear system for all frequency range of (1)
interest
• Method 3: White noise excitation which gives the har- where subscript indicates the current and voltage for phase
monic impedance of a nonlinear system for small changes of the system, and denotes the Fourier transform.
around an operating point for all frequency range of in- With the use of the method shown in Fig. 6 the sub-syn-
terest chronous harmonic impedance can be calculated as
The main deficiency of the conventional frequency scanning
(Method 1) is that determining the harmonic impedance for
each particular frequency requires one simulation run, and the
time required to perform the frequency scanning for the entire (2)
sub-synchronous frequency range can be excessive. This is es-
pecially true when considering a very detailed model of wind Note that these calculations are only valid if the WTG it-
turbine with fast acting controllers running at several kHz. An- self does not generate any level of sub-synchronous inter-har-
other disadvantage of this method is that it does not take account monics. This will be confirmed in Section III-A5.
of system nonlinearities, and it is not therefore suitable for deter- 5) Impact of Background Inter-Harmonics: Another point
mining the harmonic impedance of WTG with highly nonlinear to consider when determining the harmonic impedance of the
behavior. Lastly, with this method the power electronic devices WTG is that whether the presence of sub-synchronous inter-har-
such as a wind turbine are treated to be in the off-state condi- monics prior to injection needs to be accounted for. If any appre-
tions. ciable level of inter-harmonics is generated by the WTG prior to
To overcome the limitation of Method 1 in terms of the re- the injection this has to be deducted from the voltage and current
quired processing time, spectral signal injection methods such response of the turbine before deriving the harmonic impedance.
as the impulse and white noise signal injections have been pro- To appreciate this point, a case study was conducted where a 0.1
posed. An impulse injection is essentially the same as the sum of kV voltage is injected at the 138 kV voltage level. The voltage
all sinusoidal waveforms with magnitude of unity over the full injected is then reduced to zero at . As the injected signal
frequency range. The resulting current (or voltage) is then de- is removed at it is expected that the measured current
composed into individual frequency components using a DFT and voltage would drop to smaller values. The new steady-state
or FFT, from which the harmonic impedance of the wind tur- values would be non-zero if any background inter-harmonics
bine can be plotted. This is the basis for the frequency scanning exist in the system. The frequency response of the measured
Method 2 mentioned earlier. One deficiency of an impulse in- current and voltage is shown in Fig. 8. As shown in this figure
jection is that the power injected differs at different frequencies. harmonic current drops to zero when the injected voltage is
BADRZADEH et al.: GENERAL METHODOLOGY FOR ANALYSIS OF SUB-SYNCHRONOUS INTERACTION IN WIND POWER PLANTS 1863

Fig. 9. Impact of voltage injection at the 138 kV voltage level on the WTG
Fig. 8. Investigation of the impact of background sub-synchronous inter-har- current and busbar voltage.
monics.

using a quadratic phase shift [9]. The use of a quadratic

forced to zero. This indicates that under normal operating condi- phase shift spreads out the signal energy and avoids large
tions the WTG does not produce any sub-synchronous inter-har- spikes that could have otherwise caused a disturbance
monics. on the steady-state operating point of the original signal.
Let us take a look at this point from the standpoint of Note that a signal staggering may not be required for all
inter-harmonic generation mechanism. Assuming that any level conditions.
of steady-state sub-synchronous inter-harmonics exists in the c) Sequence transformation: Another point to consider is the
system this should either be generated by the WTG, or caused choice of sequence components for signal injection and
by the network. Background inter-harmonics at the network measurements. In principle, separate injections should be
can be either caused by a voltage unbalance in the system, applied for all three sequences. For the system and turbine
or due to the contribution of other sources of inter-harmonics under consideration injection of a positive sequence com-
such as other power electronic converters in the system. For the ponent will be sufficient due to:
system studied in this paper the three-phase voltages are fully • The control of WTG has been designed to respond
balanced, and no other power electronic components is included to positive sequence quantities only (unless during
in the system. Indeed, for turbine side scans the grid is simply voltage dip conditions). The driving point impedance
represented by an ideal voltage source which does not contain as seen by the WTG when injecting a sub-synchronous
any harmonics or inter-harmonics. The only possible source negative-sequence component is therefore zero
of sub-synchronous inter-harmonic generation is therefore the • The absence of negative- and zero-sequence compo-
WTG. In practice a WTG can generate inter-harmonics due nents at the network side
to: a) voltage unbalance in the supply, b) over-modulation of • The WTG transformer is connected in Delta at the HV
the PWM converters. The as-designed and as-modeled WTG is side, and zero sequence components will not therefore
always operated such that the modulation index is maintained be a matter of interest.
below unity, and no inter-harmonic component is therefore Negative and zero-sequence components can therefore be ne-
generated by the WTG. glected. The simulation studies performed during this study ver-
6) Frequency Scanning Implementation Procedure: Based ified this.
on the discussions presented in the previous sub-sections, a vari- 7) Verification of the Frequency Scanning Method: The suc-
ation of the methods discussed in [8] and [9] is adopted as out- cess criteria for the selection of the injected signal is that it has
lined below: to be large enough to produce the satisfactory response and not
a) White noise signal injection: A small input voltage (or to be lost in the noise, but at the same time small enough not to
current) signal is superimposed on the fundamental fre- disturb the steady-state operating point of the system, i.e., result
quency excitation. A white noise signal is injected which in unstable response of the turbine. These objectives have been
includes several uniformly spaced frequency compo- achieved with application of an excitation signal at the 138 kV
nents, all with the same magnitude. The use of a white voltage level with the magnitude of 0.1 kV and phase angle of
noise signal avoids the need for the multiple run feature in 0 for the turbine under study. As illustrated in Fig. 9 the impact
PSCAD to inject individual sub-synchronous frequencies of this signal on the WTG current and voltage manifest as very
at a time instead of the entire spectrum. small notches highlighted with circles.
b) Signal staggering: The phase angle of the each harmonic Another way to confirm the veracity of the frequency scan-
signal can be staggered by a different value, for example ning method and the injected signal is visual inspection of the
1864 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

Fig. 11. Comparison of the harmonic resistance calculated by methods 1 and 3.

component, and is the component of the torque change

Fig. 10. WTG phase current and voltage for the entire sub-synchronous fre-
quency range. in phase with angular velocity or machine speed termed as
the damping torque component. Note that there is a subtle dif-
ference between (3) and the well-known equation for the syn-
measured currents and voltages for all sub-synchronous fre- chronous machines [12] in which the angular displacement has
quency components. Fig. 10 shows all the measured sub-syn- to be replaced by the rotor angle .
chronous voltage and current components from 1–60 Hz with a The damping and synchronizing torque can be defined both
1 Hz resolution. This figure shows that the impact of injected for mechanical and electrical systems [13]. Equation (3) can
signal is negligible on the steady-state performance of these therefore be generalized as follows:
components as indicated by practically flat lines for all these
components. All these quantities achieve their correct steady- (4)
state value after initialization of the WTG model is complete.
The veracity of Method 3 for calculation of the harmonic where the subscripts and indicate the mechanical and
impedance of the WTG is compared with Method 1 utilizing electrical parts of the synchronizing and damping torque
the multi-run feature to calculate the harmonic impedance for components.
the entire sub-synchronous frequency range. Fig. 11 compares 2) Electrical Damping versus Mechanical Damping: In
the apparent resistance obtained from Methods 1 and 3 for 100% general, the mechanical damping is relatively smaller than the
turbine loading for a type 4 turbine. This figure indicates signif- electrical damping, and precise calculation of the mechanical
icant difference between the two methods although the ampli- damping is not readily possible. It has been shown that the
tude and location of the voltage injection remains the same for net damping, which comprises both electrical and mechanical
both methods. With Method 1 the peak impedance approaches damping, cannot be used to indicate the system stability or in-
4 which is approximately 15 pu. This is a clear indication stability under all conditions [14], [15]. The damping provided
that with Method 1 the signal injected into the system cannot by the WTG can therefore be evaluated solely based on the
be considered as a small-signal disturbance. The use of Method electrical damping. This is a conservative approach because the
1 indicates a negative resistance for frequencies above 45 Hz. mechanical damping can marginally compensate the negative
The type 4 turbine considered is the same as that discussed in electrical damping in certain frequency ranges.
[1] which did not exhibit any level of SSCI. One of prime advantages of the use of electrical damping
alone is the elimination of the need for a detailed mechanical
B. Electrical Damping Analysis torsional model of the drive train required for the eigenvalue
1) Complex Torque Calculation Method: The calculation of analysis as highlighted in the Introduction section.
the electrical torque can be performed by conducting time-do- 3) Proposed Calculation of Electrical Damping: Assuming
main simulation using detailed EMT-type model of the WTG. , (4) can be simplified as
This allows an accurate investigation of the nonlinear system
(5)
components and takes account of all constituting control
systems. Substituting yields
In general, the electromagnetic torque of an induction ma-
chine can be divided into two components [11]:
(6)
(3)
Equation (6) indicates that the relationship between the elec-
where is the component of torque change in phase with trical torque and the machine speed can be utilized to determine
angular displacement referred to as synchronizing torque the damping factor of the WTG, . The electrical damping
BADRZADEH et al.: GENERAL METHODOLOGY FOR ANALYSIS OF SUB-SYNCHRONOUS INTERACTION IN WIND POWER PLANTS 1865

is calculated by injecting a small-signal variation in the tur-

bine speed. Consider a speed modulation signal at frequency
given by

(7)

where is the magnitude of the injected signal component of

frequency , and is the phase angle of the injected signal
component of frequency . For studies discussed in this report
is set to 0.001 pu.
The corresponding electrical torque component, , of
the same frequency can be described as

(8)
Fig. 12. System side frequency scans with various representation of the collec-
tion system for a strongly meshed connection.
The Damping Factor ( or ) is then computed as

(9)
ering low frequency local, and inter-area, and inter-plant modes
of oscillations which are not the point of discussions in this
where is .
paper. This approximate method has not been used in this paper
To determine the electrical damping for all sub-synchronous
for the calculation of the damping factor, but it will be used
frequencies of interest, the excitation frequency is varied be-
to verify the relationship between the damping factors at var-
tween zero and the fundamental frequency.
ious turbine loading conditions as obtained from the simulation
The damping factor must be positive for all sub-synchronous
studies.
range to avoid the risk of sub-synchronous interaction. A neg-
ative damping factor in the sub-synchronous range indicates a
IV. CASE STUDIES
high risk of SSTI.
4) Approximate Estimation of the Electrical Damping: An Having discussed different tools and techniques for the inves-
approximate calculation of the damping factor can be performed tigation of SSI issues, this section briefly describes the applica-
assuming that the synchronizing torque can be neglected. This tion of the three selected tools on a practical power system with
can be justified as follows: For a fixed-speed induction generator significant level of series compensated ac lines [7].
the synchronizing torque can be calculated as function of the
A. Frequency Scanning Analysis
maximum torque and maximum slip [11]:
1) Network Side Scans and Impact of Collector Grid Rep-
(10) resentation: This sub-section discusses the impact of three
different operating conditions on the system side frequency
For low frequency sub-synchronous range the following ap- scans. These include a strongly meshed connection (CTG001),
proximation can be made: a nearly radial connection (CTG007), and a fully radial con-
nection (CTG008). A nearly radial condition means that an
(11) outage of one of the lines makes the system fully radial.
The corresponding frequency scans for these three conditions
The synchronizing torque can therefore be discarded. The looking from behind the wind farm collection system are shown
same conclusion applies to variable speed wind turbines unless in Figs. 12–14, respectively. Note that the network does not
the synchronizing power is emulated through the converter con- include any active or non-linear devices and a simple frequency
trol. scan, i.e., Method 2, has been adopted for the network side.
Neglecting the synchronizing torque (6) can be further sim- As shown in Fig. 12 for system conditions represented by
plified as CTG001, the system impedance across sub-synchronous fre-
quency range of interest is not indicative of any distinct reso-
(12) nant frequency. The CTG007 which represents the near radial
scenario shows a distinct resonant frequency (depicted by the
The electromagnetic torque can be calculated as impedance dip in Fig. 13) at around 22–23 Hz. For CTG008
which represents a completely radial scenario Fig. 14 shows that
(13) the impedance at frequencies around 20–21 Hz drops to practi-
cally zero which is indicative of a classical sub-synchronous res-
where is the nominal generator rotor speed in pu, onance issue. A dip in the magnitude of the network impedance
is the relative variation of the electrical power over the speed implies a negative resistance coincident with a zero reactance.
variation, and the is the generator electrical power in pu. This is consistent with the observations made in [16].
Note that this approximation introduces some errors espe- 2) Turbine Side Scans and Impact of Turbine Transformer
cially when analyzing the upper range of sub-synchronous fre- Impedance: Having investigated the frequency scans at the net-
quencies. However, it provides sufficient accuracy when consid- work side this sub-section discusses the frequency scans at the
1866 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

Fig. 15. Turbine side frequency scans for a nearly radial system.
Fig. 13. System side frequency scans with various representation of the col-
lection system for a nearly radial system.

the turbine side frequency scan this condition does not, how-
ever, hold true. Looking at Fig. 15 the turbine reactance crosses
over at around 40 Hz. This is very far from the network reso-
nant frequency of around 20 Hz shown in Fig. 14. This is unlike
the classical SSR phenomenon where the apparent reactance be-
comes practically zero at the resonance frequencies. This is most
likely attributed to the active control of wind turbines whose be-
havior cannot be represented by passive R, L and C elements.
In general, the following criteria pose a higher risk of SSCI
which are therefore suggested to identify the conditions which
require more detailed studies using the EMT analysis:
1) Any reactance crossovers on the turbine side that coincides
with resonant conditions on the system side even if the
resistance at the sub-synchronous frequency is positive
Fig. 14. System side frequency scans with various representation of the col-
lection system for a fully radial system. 2) Any resonant condition on the system side if the turbine
resistance at that sub-synchronous frequency is negative.
Even though the turbine reactance cross-over does not
turbine side, and aims to correlate the turbine side and network match the system resonant frequency the turbine resistance
side scans to derive conditions which pose a higher risk for po- is negative at the sub-synchronous frequency at which
tential SSI issues. For the turbine frequency scan, the apparent system shows resonant conditions.
impedance is segregated into the resistive and reactive compo-
B. Electrical Damping Analysis
nents which collectively provide information pertaining SSCI
conditions. The traces of R and X for 100% turbine dispatch The frequency scan analysis discussed in the previous sub-
are shown in Fig. 15. These traces are produced looking into section has identified the critical resonance frequencies from an
the high side of the turbine transformer which includes the im- SSCI standpoint. The electrical damping analysis is then con-
pact of transformer reactance and losses. This figure indicates ducted to investigate the risk of SSTI for the specific type 3 tur-
that the turbine apparent resistance is negative across the entire bine used for the studies. An example of the electrical damping
sub-synchronous frequency range; the higher the frequency the analysis performed for CTG007 (near radial condition) for three
larger the negative resistance will be. The turbine apparent re- dispatch levels of 100%, 55%, and 30% has been plotted in
actance is positive up to 40 Hz, and turns negative for the rest Fig. 16. The key observations made here is that the damping
of sub-synchronous range. factor exhibits a clear dip at about 45 Hz. This implies that the
The negative resistance implies that the active component of system resonance frequencies at around 15 Hz could give rise
the converter voltage will decrease with an increase in the ac- to SSTI issues. Note that unlike SSCI, the SSTI issue is only
tive component of the converter current. At such frequencies concerned with a few sub-synchronous torsional resonance fre-
the converter will be operating unstable where some poles and quencies with fixed values.
zeros are in the unstable region, and the required gain margin The impact of different turbine loading on the electrical
and phase margin for the satisfactory operation of the converter damping can be described using (13). In (13) the term
cannot be met. The negative reactance implies that the reactive is nearly constant irrespective of the loading
component of the converter voltage will decrease with an in- conditions. The term will increase with an increase in
crease in the reactive component of the converter current. the turbine loading. This implies a decrease in the electrical
A classical SSR issue can be readily identified from the fre- damping provided by the WTG with an increase in the turbine
quency scans. This occurs when a negative resistance coincides loading. This behavior can be clearly seen in Fig. 16 up to
with a zero reactance derived from the same point. Looking at 52 Hz.
BADRZADEH et al.: GENERAL METHODOLOGY FOR ANALYSIS OF SUB-SYNCHRONOUS INTERACTION IN WIND POWER PLANTS 1867

Fig. 16. Electrical damping plot for three levels of turbine power dispatch.

Fig. 18. Time domain turbine and WPP response for a fully radial system.

Fig. 19. Electrical damping plot for three levels of compensation for simple
Fig. 17. Time domain turbine and WPP response for a nearly radial system. radial systems.

C. EMT-Type Analysis analysis it is therefore necessary to adopt another test system.

1) Validation of Frequency Scanning Against EMT Analysis: For this purpose a simplified version of the test circuit presented
To confirm the veracity of frequency scanning results EMT-type in [1] adopted where the WTG is connected radially to the
simulation studies were conducted. The results for the nearly series compensated line. Results for three different levels of se-
radial and fully radial operating conditions are shown in Figs. 17 ries compensation, namely no compensation and 45% and 75%
and 18. In these cases it is assumed that the bypass breaker for compensation are shown in Fig. 19. For the 45% and 75% series
the series capacitor is opened at and the WPP will compensation level the electrical damping becomes negative
be nearly or fully radial with the series compensated lines. As for a limited sub-synchronous frequency range. The negative
shown in Fig. 17 the nearly radial operating condition does not damping with the 75% compensation is significantly higher
give rise to any SSCI concern and turbine resumes its pre-event than that with 45% compensation. This suggests significantly
steady-state operation within a couple of hundred milliseconds. higher torque oscillations for the 75% case. Results obtained
This is consistent with Fig. 9 which does not indicate any clear from EMT analysis with 45% and 75% compensation levels are
resonance conditions. For a fully radial system Fig. 18 indicates plotted in Figs. 20 and 21, respectively. As an example torque
the occurrence of SSCI condition with subsequent tripping of oscillations between two arbitrary masses on the shaft are
the wind turbine on rotor overcurrent protection (trip 9) which shown. The findings are consistent with the electrical damping
protects the wind turbine from any potential damage. The results analysis where significantly higher and unstable oscillations in
are consistent with the respective frequency scan plot shown in the torque are observed with 75% compensation.
Fig. 14 which clearly indicates a zero impedance point at around
20 Hz. V. CONCLUSIONS
2) Validation of Electrical Damping Analysis Against EMT This paper has discussed methods suitable for the analysis
Analysis: Results shown in Fig. 16 suggest practically no of sub-synchronous interaction in wind power plants. These
negative electrical damping for the entire sub-synchronous fre- include appropriate frequency scanning analysis for wind tur-
quency range. To confirm the veracity of the electrical damping bines with an active and nonlinear behavior, and calculation of
1868 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

[2] G. D. Irwin, A. K. Jindal, and D. A. Woodford, “Sub-synchronous con-

trol interactions between type 3 wind turbines and series compensated
AC transmission systems,” in Proc. IEEE PES General Meeting, De-
troit, MI, Jul. 2011.
[3] M. Bongiorno, A. Petersson, and E. Agneholm, The Impact of Wind
Farms on Sub-Synchronous Resonance in Power Systems, Elforsk re-
port 11:29, Apr. 2011. [Online]. Available: http://www.elforsk.se/Pro-
gramomraden/El--Varme/Rapporter/?rid=11_29_
[4] L. Fan, R. Kavasseri, Z. L. Miao, and C. Zhu, “Modeling of DFIG-
based wind farms for SSR analysis,” IEEE Trans. Power Del., vol. 25,
no. 4, pp. 2073–2082, Oct. 2010.
[5] L. Fan, C. Zhu, Z. Miao, and M. Hu, “Modal analysis of DFIG-based
wind farm interfaced with a series compensated network,” IEEE Trans.
Energy Convers., vol. 26, no. 4, pp. 1010–1020, Dec. 2011.
[6] B. Badrzadeh, M. Sahni, D. Muthumuni, Y. Zhou, and A. Gole, “Sub-
synchronous interaction in wind power plants- Part I: Study tools and
techniques,” in Proc. IEEE PES General Meeting, San Diego, CA, Jul.
2012.
Fig. 20. Torque oscillation with 45% series compensation. [7] M. Sahni, B. Badrzadeh, D. Muthumuni, Y. Cheng, H. Yin, and S.-H.
Hunag, “Sub-synchronous interaction in wind power plants—Part II:
An ERCOT case study,” in Proc. IEEE PES General Meeting, San
Diego, CA, Jul. 2012.
[8] M. Sahni, D. Muthumuni, B. Badrzadeh, A. Gole, and A. Kulkarni,
“Advanced screening techniques for sub-synchronous interaction in
wind farms,” in Proc. IEEE T&D Conf. Expo., Orlando, FL, May 2012.
[9] X. Jiang and A. M. Gole, “A frequency scanning method for the iden-
tification of harmonic instabilities in HVDC systems,” IEEE Trans.
Power Del., vol. 10, no. 4, pp. 1875–1881, Oct. 1995.
[10] M. Mohaddes, A. M. Gole, and S. Elez, “Steady state frequency re-
sponse of STATCOM,” IEEE Trans. Power Del., vol. 16, no. 1, pp.
18–23, Jan. 2001.
[11] P. K. Kovacs, Transient Phenomena in Electrical Machines. New
York: Elsevier, 1984.
[12] P. Kundur, Power System Stability and Control. New York: Mc-
Graw-Hill, 1994.
[13] I. M. Canay, “A novel approach to the torsional interaction and
electrical damping of the synchronous machine, Part I: Theory,” IEEE
Trans. Power App. Syst., vol. PAS-101, no. 10, pp. 3630–3638, Oct.
1982.
[14] L. Harnefors, “Analysis of sub-synchronous torsional interaction with
Fig. 21. Torque oscillation with 75% series compensation. power electronics converters,” IEEE Trans. Power Syst., vol. 22, no. 1,
pp. 305–313, Feb. 2007.
[15] L. Harnefors, “Proof and application of the positive-net-damping sta-
bility criterion,” IEEE Trans. Power Syst., vol. 26, no. 1, pp. 481–482,
electrical damping provided by the turbine at sub-synchronous Feb. 2011.
resonance frequencies. A generalized mathematical formulation [16] R. G. Farmer and A. L. Schwalb, “Novajo project report on sub-syn-
of both methods were presented to allow ease of application chronous resonance analysis and solutions,” IEEE Trans. Power App.
Syst., vol. PAS-96, no. 4, pp. 1226–1232, Jul./Aug. 1977.
for any given wind turbine or network. It has been shown that
the frequency scanning and EMT-type analyses can be used
collectively to confirm the presence of SSCI conditions. The
electrical damping analysis along with the EMT-type analysis
allows identifying the risk of SSTI conditions. It has been
concluded that the EMT-type time domain analysis and the
screening techniques are complementary with respect to the
investigation of sub-synchronous interaction issues, and when
used together allow a significant time saving relative to the use
of an EMT-type program alone. Babak Badrzadeh (S’03–M’07–SM’12) received
the Ph.D. degree from the Robert Gordon Univer-
As the screening methods discussed in the paper provide a sity, U.K., in 2007 in the area of electrical power
distinct advantage for the analysis of large networks an eigen- engineering.
value analysis has not been adopted. This method can, however, After spending a short period as an Assistant Pro-
be considered as a supplementary tool to provide additional in- fessor at the Technical University of Denmark, he
joined Mott MacDonald, Transmission and Distribu-
formation such as participation factors and mode shapes which tion Division, U.K., as a system analysis and network
cannot be obtained by other methods. planning engineer. From March 2010 to March 2012,
he was with Plant Power Systems, Vestas Technology
R&D, Aarhus, Denmark, where he acted as a lead en-
REFERENCES gineer in the area of advanced wind power plant simulation and analysis. Since
May 2012, he has been with the Australian Energy Market Operator as a Tech-
[1] B. Badrzadeh and S. Saylors, “Susceptibility of wind turbine generators nical Specialist in System Capability. His area of interest includes power system
to sub-synchronous control and torsional interaction,” in Proc. IEEE electromechanical and electromagnetic transients, application of power elec-
T&D Conf. Expo., Orlando, FL, May 2012. tronics in power systems, wind power plants, and modeling and simulation.
BADRZADEH et al.: GENERAL METHODOLOGY FOR ANALYSIS OF SUB-SYNCHRONOUS INTERACTION IN WIND POWER PLANTS 1869

Mandhir Sahni received the B.S. degree from Ban- Dharshana Muthumuni received the M.Sc. and
galore University, India, and the M.S. and Ph.D. de- Ph.D. degrees from the University of Manitoba, Win-
grees from the University of Texas, Arlington. nipeg, MB, Canada, both in electrical engineering.
He has over ten years of experience in power He joined the HVDC Centre in June 2001 and is the
market analysis, generation and transmission plan- machines and transformer simulation specialist and
ning in deregulated energy markets in the U.S., head of the PSCAD Technical Sales Department. He
power quality, stability, and protection co-ordination has been involved in numerous engineering studies
analysis with an extensive background in power using a variety of simulation products during his ca-
system analysis. He is currently leading all transmis- reer including PSCAD and PSS/E. Showing leader-
sion planning activities associated with numerous ship in simulation excellence, his technical knowl-
CREZ utilities in ERCOT. He has also served as edge on power system behavior, model development,
an expert witness on behalf of the staff of West Virginia Public Service and simulation studies is highly advanced and is regularly sought out by clients
Commission for the hearing associated with the 765 kV Potomac Appalachian around the world for his expertise.
Transmission Highway (PATH) project in Mid-Atlantic PJM. His research in-
terests include power system planning, integration of renewable generation into
power system grids, power quality, and power system reliability. His research
has been published in numerous referred journals, IEEE P&E magazine, and Aniruddha Gole (F’10) received the B.Tech. degree
conference proceedings. from the Indian Institute of Technology, Bombay, and
the M.Sc. and Ph.D. degrees from the University of
Manitoba, Winnipeg, MB, Canada, all in electrical
engineering.
Yi Zhou (S’05–M’09) received the B.Sc. degree He is a Professor of Electrical and Computer Engi-
from Zhejiang University, China, and the M.Sc. neering at the University of Manitoba. Since 2004, he
and Ph.D. degrees from Delft University, The also has been the NSERC Industrial Research Chair
Netherlands. in Power Systems Simulation. His research interests
He is currently with Power Plant Systems Depart- include the utility applications of power electronics
ment of Vestas, as the team leader of the Plant De- and power systems transient simulation. As an orig-
sign group, working in the field of grid integration inal member of the design team, he has made important contributions to the
study of wind power plant. Before joining Vestas, he PSCAD/EMTDC simulation program.
had worked in Anhui Electric Power Company, State Dr. Gole is active on several working groups of CIGRE and IEEE and is
Grid Corporation of China for 7 years, leading the a Registered Professional Engineer in the Province of Manitoba. In 2007, the
Power System Stability Group and being responsible IEEE Power Engineering Society awarded him the prestigious Nari Hingorani
for the transmission system analysis such as power flow, transient stability, and FACTS Award “..for Contributions to the Education in the Field of Power Sys-
dynamic stability study. He had also worked as the lead engineer in the Global tems and Embedded Power Electronics Apparatus Simulation”. He was elected
Research Center of General Electric, responsible for grid integration of renew- a Fellow of IEEE in 2010 “for contributions to the modeling of power elec-
able power. tronics apparatus”.