1

Wind Farms in Weak Grids

Compensated with STATCOM
Trond Østrem, Research Fellow, Narvik University College

Abstract–In this paper a distribution network in the vicinity II. SIMULATION MODEL
of Narvik is modelled by using an ATP-EMTP application
called ATPDraw. The network has a combination of wind A. Network description
power and hydro power and a weak grid connection. Three
different scenarios are considered: Short circuit on wind
power without hydro, wind fluctuation on a combined
The network of current interest is described Fig. 1, and it
network, and the effect of both wind fluctuation and an arc supplies the town Narvik and its surroundings. This
fault. In each case, the system response with and without a distribution network is connected to the central network
STATCOM is analyzed. through the 132 kV bus bar at Narvik Transformer Station
(Furumoen).
I. INTRODUCTION Most of the distribution network is passive, so the
interesting parts are the hydro generators at Nygård and
THIS document is initiated by a final examination work at Sirkelvann, Nygårdsfjellet wind farm, the overhead lines
The Norwegian University of Science and Technology, and transformers in the area, and the undersea cable to the
Trondheim. 36 kV bus bar at Frydenlund Transformer Station.
The network behind Frydenlund is considered as a stiff
A. Original Problem network in order to simplify the model. The connections
between Nygård and Djupvik and between Nygård and
The original problem was defined as: Simulation of a wind Bjerkvik are quite week connections and are not taken into
farm based on induction generators connected through a account in the simulation model. This network also contains
weak grid and compensated with a static compensation a passive load, Skogvann, but it is just a small company
(STATCOM) device. with a power consumption of 160 kilowatts and is also
Develop a simulation model for a 100 MW wind farm ignored.
using squirrel cage induction generators magnetised with The remaining network consists of an infinite bus at
fixed capacitor banks, and connected to the grid through a Frydenlund, the underwater cable (the weak connection),
weak grid (low short circuit current: strong grid usually generators and transformers at Nygård hydro power plant,
gives a short circuit ratio >20, use this assumption). an overhead line to Sirkelvann, generator and transformer at
1. Model the STATCOM as a voltage source converter. Sirkelvann mini hydro generator, an overhead line to
2. Use decoupled control of active and reactive power for Nygårdsfjell, and finally, Nygårdsfjell wind farm with
controlling the STATCOM. generators, transformers and possibly a STATCOM.
3. Simulate a short circuit and compare the transient
response of the system with and without the STATCOM.
See how much the transient stability margin is improved
with the STATCOM and see what other beneficial effects
the STATCOM has on the system stability.
4. Give a generalised expression for the rating of the
STATCOM as a function of the system characteristics and
contingency level, such as short circuit ratio and voltage
sag.

B. Modified Problem
Since there is a newly developed wind farm close to
existing hydro power generators connected to the grid
trough a weak connection, the decision was made to deviate
somehow from the original problem. The simulation model
is based on the previous mentioned network, rather than the
wind farm described initially.
Since the modified model contains both synchronous and
asynchronous generators, the analysis also contains
evaluations of different voltages and load angles. In order to
limit the size of the work, point 4 in the original problem is
replaced by an analysis of dynamic interaction between Fig. 1. Single line diagram of the distribution network in Narvik.
wind power and hydro power with and without STATCOM.
 2
B. Software One solution is to make a simplified model of the
asynchronous machine, which was done in this work. Since
Due to the suitability of simulation tools from there are two major problems to be simulated, two different
Alternative Transients Program – Electromagnetic overall models were developed. The first problem is a
Transients Program (ATP–EMTP), for power grid matter of how the system handles severe faults, like short
simulation, an ATP–EMTP application called ATPDraw circuits or voltage dips. The second problem is the
was used for this purpose. The simulation tool is based on interaction between different power sources.
Windows and gives a clear picture of the model, much like In the first case, the model consists of a lumped wind
a traditional single line diagram. farm with a STATCOM, the transformers and transmission
For analysis of the simulation results, another ATP– lines, a model of the short circuit, and an infinite bus. This
EMTP application was used, namely ATP Analyzer. This model is shown in Fig. 2, and there is a similar version
program provides different types of plots and several kinds without the STATCOM, not shown in this article. Since
of analysis, e.g. power flow calculations, symmetrical load angle stability is not part of the case, the synchronous
components and Park transform. generators are left out.
These tools are adopted in our institute recently, so parts In the second case a simpler model of the wind farm is
of the work have simply been to learn how to utilize the made, since the main focus of the analysis is to observe the
programs. In ATP-EMTP it is challenging and time stability of the synchronous generators when the wind
consuming to initialize rotating electrical machines, and power production fluctuates. The model, shown in Fig. 3,
especially a combination of synchronous generators and contains, in addition to the pervious mentioned lumped
squirrel cage generator turned out to be an obstacle. So far wind farm model, transmission lines, hydro generators and
one has not succeeded in making a model with a transformers, a model of the short circuit, and an infinite
combination of these elements. bus. The short circuit was added in order to simulate a
combination of fluctuating power and contingencies.

Nygårdsfjellet T2 Overhead lines T1 Undersea Bus bar

wind farm cable Frydenlund
0.75/22kV 36/22kV

VT1 U
I

Infinite bus
3 phase
short circuit

Fig. 2. ATPDraw simulation model of wind farm with STATCOM, connected to the main grid through a weak connection.

Nygårdsfjellet
wind farm

Overhead line
Sirkelvann -
Nygårdsfjellet Overhead line
Nygård -
Sirkelvann

T1 T4
G1
Frydenlund
T2 bus bar U Infinite
Nygård bus bar
G2
hydroelectric Undersea
power plant cable Three phase
short circuit
T3
G3
Nygård
bus bar

Fig. 3. ATPDraw simulation model of wind farm with STATCOM, combined with hydro power and a weak grid connection.
 3
C. Model implementation TABLE III
PARAMETERS FOR TRANSMISSION LINES
Most of the elements in the simulation models are
described in [1], and will not be described in detail in this Length Max. Resistance Inductance Mutual
article. A couple of new elements are developed for this [km] current [Ω] [mH] capacitance
certain purpose and will be thoroughly explained. [A] [µF]
Underwater cable Frydenlund – Nygård
1) Synchronous generators: 7.55 400 1.18 4.01 4.54
These are modelled without control, i.e. the dynamics in Overhead line Nygård – Sirkelvann
turbines and controllers are not included. Saturation is 6.76 454 1.95 8.4 0.61
ignored, and some reactances, time constants and moments Overhead line Sirkelvann – Nygårdsfjell
of inertia are based on assumptions. Table I shows some of 5.02 454 1.29 5.79 0.1
the parameters for the generators.
5) Arc Fault
TABLE I The arc fault is modelled as a Thevenin impedance, a
SOME RATED VALUES FOR THE DIFFERENT GENERATORS
ground recistance and a switch, all three phase. These
components are shown in Fig. 4. The Thevenin impedance
Generator Nominal power Nominal voltage
has a value of (0.1 + j0.4) Ω, and the resistance has a value
[MW] [kV]
of 0.1 Ω.
G1, G2 8.0 4.2
U
G3 9.0 4.2 Infinite
G4 0.7 6.9 bus
G5–G7 (wind farm) 2.3 0.75

2) Squirrel Cage Generators: Three phase

Each generator is equipped with a front end full size short circuit
converter, but the turbine model is implemented as a
squirrel cage generator connected directly to the grid, for
simplicity. Rated values for the squirrel cage generators are
also shown in Table I.

3) Transformers: Fig. 4. The three phase switched voltage divider representing the
The transformers are modelled with phase shifts, temporary arc fault at Frydenlund bus bar.
impedances, vector groups etc, but saturation of the core is
ignored. Table II shows some rated values for the 6) STATCOM
transformers. Data for the wind turbine transformers were The STATCOM is modelled as a voltage source converter.
not available during the work, but transformer impedance is Since the switching period will be much shorter than the
included in the lumped wind turbine model. different mechanical time constants of the system, the
voltage source is implemented simply by an inverse Park
TABLE II transform, as shown in Fig. 5. The other elements of the
SOME RATED VALUES FOR THE TRANSFORMERS STATCOM are voltage and current measurement units, a
Park transform and current controllers for the d-axis and the
Transformer Nominal power Nominal voltage q-axis, respectively. By choosing zero reference for the q-
[MVA] [kV] axis, cos ϕ is forced to be equal to one. By using a PI-
T1, T2 10 35/4.2 controller, the static error disappears. For the d-axis, a P-
T3 10 33.3/4.2 controller is sufficient, since the active power flow changes
T4 1.1 22/6.9 a lot and is limited by the power source. By choosing a high
T5 7.0 32.4/23 reference value, the d-component will be as high as
possible.
Transformer T5 is equipped by a tap changer, but since the
simulations just span a few seconds, the tap changer is
assumed to be in middle position without any action at all.
Inverse
4) Transmission Lines: transform

The transmission system consists of an underwater cable

and a chain of overhead lines with transformers in between.
The line length, maximum current, and the impedance
components at the system frequency (50 Hz) is shown for Park PI-regulator
I_q
PI-regulator
transform I_d
the different lines in Table III.

Current sensors

Fig. 5. STATCOM simulation model for ATPDraw.

 4
7) Simplified Wind Generator Model The dc network to the left represents the mechanical two-
There is still some work to be done in order to initalize mass system of the turbine. The capacitances are equal to
the asynchronous generator in a proper way. By the moment moments of inertia, and the inductance equals the spring
the machine model generates quite heavy false transients, constant of the shaft. The resistances resemble the friction
and these disturbances prevent the simulation model from of the system.
reaching a convergent solution. That is particularly a The generator is modelled by phase equivalents with a
problem in combination with synchronous generators. TACS-controlled rotor resistance expressed by (1), where s
The purpose of the simplified wind generator model is to is the slip of the generator. The frequency measurement of
supply the network with intermittent wind power, which is the stator voltages gives a speed reference ωs for the
done by a voltage source behind an impedance, as shown in electromechanical system.
Fig. 6. The voltage magnitude is controlled by a ‘Transient
Analysis of Control Systems’–signal (TACS–signal). 1− s
The TACS–sources and multipliers to the left in the R2 '⋅ (1)
s
figure generate sinusoidal signals with varying amplitude,
The model does not take into account non-linear and
and the sources to the right convert the TACS-signals into
time variant properties, and the voltage source will have a
phase voltages. As can be seen in Fig. 3, there is an
fixed phase angle even if the grid voltage is slightly phase
impedance behind this three phase voltage source. The
shifted.
model does not show the rotor speed or any other dynamic
properties.
III. SIMULATION RESULTS

Three different scenarios were simulated. First, the

simulation model of Fig. 2 was used, and a short circuit
took place after 3 seconds and lasted for 0.3 seconds.
Voltages, rotor speed and power flow was measured and
analyzed both with and without the STATCOM.
Next simulation used the model in Fig. 3, but without
activating the arc fault. The wind power changes from rated
output to zero in pulses of one second duration. Load
angles, voltages and power flow are measured and analyzed
with and without the STATCOM.
In the final simulation the same model is used, but the
U U U
arc fault at Frydenlund bus bar is activated at 3.3 seconds
and lasts for 0.3 seconds. This is obviously a worst case
scenario, and since the model do not contain any relay
protection or other protection devices, the simulation result
Fig. 6. Simplified wind generator model without rotor speed and turbine in unrealistic. Still, it gives a picture of the forces and
dynamics. energies present under such circumstances. Load angles,
voltages and power flow are measured and analyzed with
8) Complex Wind Generator Model and without the STATCOM.
In some of the simulations, the performance and All simulations went without convergence problems, and
behaviour of the electromechanical wind turbine system is the time span was from zero seconds to five seconds. It is
essential, and for that reason there was a need for a more possible to simulate over a longer period, but it is very time
complex model with a dynamical behaviour. This model is consuming, and this five second interval gives a clear
shown in Fig. 7. picture of cyclic variations of different kinds.
OMEG_S

ROTRES
OMEG_R

U KL

Fig. 7. Complex wind generator model with turbine dynamics.

 5
A. Short Circuit Simulations without Hydro Power 900
Bus bar wind farm without STATCOM (Blk), Bus bar wind farm with STATCOM (Red)

The simulation model for these simulations is presented 800

in Fig. 2. Comparisons were made with and without use of

700

600

STATCOM. 500

Fig. 8. shows rotor speed without STATCOM (red (a)

graph) and with STATCOM (blue dashed graph). It can be 500
Bus bar Nygård without STATCOM (Blk), Bus bar Nygård with STATCOM (Red)

noticed that the fluctuations of the rotor speed are heavier

20000

15000

with the STATCOM, probably due to increased dynamics, 10000

while the rotor speed stabilizes faster after the short circuit 5000

with the STATCOM. The explanation for the ‘ringing’ of (b)

B b F d l d (Blk)

the rotor speed without the STATCOM might be numerical

round-off problems or other numerical problems. Slight
Bus bar Frydenlund (Blk)
25000

20000

changes of the model parameters changed this phenomenon. 15000

Fig. 9. shows rms-values of the wind farm currents. In

10000

5000

(a) one can see the currents from the wind farm into the grid 0
0 1,0 2,0 3,0 4,0 5,0

without STATCOM (black graph) and with STATCOM (c)

(red graph). In (b) the currents from the lumped wind
Fig. 10. rms voltages (a) at the wind farm, (b) at Nygård bus bar, and (c) at
generator model are shown, both without STATCOM (black Frydenlund bus bar, without STATCOM (black graph) and with
graph) and with STATCOM (red graph). It can be noticed STATCOM (red graph).
that the STATCOM increases the current from the wind
farm to the grid during the fault, but it decreases the B. Simulation of Wind Fluctuations with Hydro Power
generator current. Thus, it should increase the stability The simulation model for these simulations is presented
margins of the wind farm. in Fig. 3, but with the arc fault deactivated. Comparisons
Fig. 10. shows rms voltages for the system, the grid were as usual made with and without use of STATCOM.
connection of the wind farm in (a), Nygård bus bar in (b) Fig. 11. shows load angles for G1 and G2 (a), G3 (b),
and Frydenlund bus bar in (c). Black graphs show voltages and G4 (c), without STATCOM (black graph) and with
without STATCOM and red graphs with STATCOM. For STATCOM (red graph). For (a) and (b) there is only a slight
Frydenlund there is no difference with or without increase in the load angle, which indicates a slight increase
STATCOM. It can be seen that the STATCOM improves in the power flow when the STATCOM is applied. For (c)
the voltage condition at the wind farm, but has little effect at there is a significant change, both since G4 is much closer
Nygård. The voltage at Frydenlund drops to approximately to the wind farm and a since it is a much smaller generator.
20% of rated voltage for 0.3 seconds, which is a typical Fig. 12. shows rms voltages at the wind farm (a), and at
voltage drop scenario [2]. Nygård bus bar (b) without STATCOM (black graph) and
with STATCOM (red graph). The figure shows that the
Rotor speed without STATCOM Rotor speed with STATCOM
voltage quality is improved significantly at the wind farm
and only slightly at Nygård bus bar.
180

Fig. 13. shows active (black graph) and reactive power

175

flow (red graph) at Frydenlund bus bar without STATCOM

170

165
(a), and with STATCOM (b). It can be seen that these heavy
160

155
wind farm perturbations causes the active power flow to
150
waver between the wind farm and the grid when the
145
STATCOM is absent, while the use of STATCOM gives a
140
more smooth power flow. One should take into account that
0,0 0,5 1,0 1,5 2,0 2,5
Time in Seconds
3,0 3,5 4,0 4,5 5,0
the simulation model provides a STATCOM without
Fig. 8. Wind generator rotor speed, without STATCOM (red graph) and boundaries and with capability of injecting active as well as
with STATCOM (blue dashed graph). reactive power to the grid.
Load angle G1 and G2 without STATCOM (Blk), Load angle G1 and G2 with STATCOM (Red)
61,2
Wind farm current without STATCOM (Blk), Wind farm current with STATCOM (Red)
10000 61,0
9500 60,8
9000 60,6
8500 60,4
8000
60,2
7500
60,0
7000
6500
6000
(a)
Short circuit Load angle G3 without STATCOM (Blk), Load angle G3 with STATCOM (Red)
5500 57,5
5000 57,0
G t t ith t STATCOM (Blk) G t t ith STATCOM (R d)

(a) 56,5

56,0

55,5
Generator current without STATCOM (Blk), Generator current with STATCOM (Red) 55,0
10000
54,5
9500

9000

8500
(b)
Load angle G4 without STATCOM (Blk), Load angle G4 with STATCOM (Red)
8000 19,5

7500
19,0
7000

6500 18,5

6000 18,0
Short circuit
5500
0 1,0 2,0 3,0 4,0 5,0 17,5
Time in Seconds C:\Diverse\Doktorarbeid\ATP-EMTP\Nygaardsfjellet\VINDNETT2.mdb 0 1,0 2,0 3,0 4,0 5,0
(b)
(c)
Fig. 9. Wind farm rms currents (a) to the grid, and (b) from the generators Fig. 11. Load angles for (a) G1 and G2, (b) G3, and (c) G4, without
without STATCOM (black graph) and with STATCOM (red graph). STATCOM (black graph) and with STATCOM (red graph).
 6

Wind farm voltage without STATCOM (Blk), Wind farm voltage with STATCOM (Red) Load angle G1 and G2 without STATCOM (Blk), Load angle G1 and G2 with STATCOM (Red)
600 66,0

64,0
580
62,0
560
60,0
540
58,0
520 56,0

500 54,0

(a)
480

460

440
,
å å

(a)
Load angle G3 without STATCOM (Blk), Load angle G3 with STATCOM (Red)
62,0
60,0
Bus bar voltage Nygård without STATCOM (Blk), Bus bar voltage Nygård with STATCOM (Red)
20900 58,0
56,0
20850 54,0
52,0
20800 50,0
48,0
20750

20700 (b)
20650

Load angle G4 without STATCOM (Blk), Load angle G4 with STATCOM (Red)
20600 21,0
0 1,0 2,0 3,0 4,0 5,0
20,0
(b) 19,0

18,0

17,0

Fig. 12. rms voltages at (a) the wind farm, and (b) Nygård bus bar, without 16,0

15,0
STATCOM (black graph) and with STATCOM (red graph). 0 1,0 2,0 3,0 4,0 5,0

(c)

6,000E6
Real power at Frydenlund bus bar without STATCOM (Blk), Reactive power at Frydenlund bus bar without STATCOM (Red)
Fig. 14. Load angles for (a) G1 and G2, (b) G3, and (c) G4, without
4,000E6 STATCOM (black graph) and with STATCOM (red graph).
2,000E6

Wind farm voltage without STATCOM (Blk), Wind farm voltage with STATCOM (Red)
0
600

-2,000E6 550

500
-4,000E6
450
-6,000E6
400

(a) 350
B b lt N å d ith t STATCOM (Blk) B b lt N å d ith STATCOM (R d)
Short circuit

Real power at Frydenlund bus bar with STATCOM (Blk), Reactive power at Frydenlund bus bar with STATCOM (Red)
(a)
1,000E7 Bus bar voltage Nygård without STATCOM (Blk), Bus bar voltage Nygård with STATCOM (Red)
25000
8,000E6

6,000E6 20000

4,000E6
15000
2,000E6

0 10000
Short circuit
-2,000E6
5000
-4,000E6

-6,000E6

-8,000E6
(b)
Bus bar voltage Frydenlund (Blk)
0 1,0 2,0 3,0 4,0 5,0
25000
S C O S

(b) 20000

15000

10000
Fig. 13. Active (black graph) and reactive power flow (red graph) at 5000

Frydenlund bus bar, (a) without STATCOM, and (b) with STATCOM. 0
0 1,0 2,0 3,0
Short circuit

4,0 5,0

(c)
C. Simulation of Wind Fluctuations with Hydro Power and
Arc Fault Fig. 15. rms voltages at (a) the wind farm, (b) Nygård bus bar, and (c)
The model for these simulations is presented in Fig. 3, Frydenlund bus bar, without STATCOM (black graph) and with
STATCOM (red graph).
with the arc fault activated. Comparisons were as usual
made with and without use of STATCOM. 1,000E8
Real power from Nygård to Frydenlund without STATCOM (Blk), Reactive power from Nygård to Frydenlund without STATCOM (Red)

Fig. 14. shows load angles for G1 and G2 (a), G3 (b), 8,000E7

6,000E7

and G4 (c), without STATCOM (black graph) and with 4,000E7

2,000E7

STATCOM (red graph). For (a) and (b) there is a significant 0

Short circuit

change in the load angle after the short circuit, and the
-2,000E7

-4,000E7

power will go back and forth for several seconds until the -6,000E7

(a)
swing is completely damped. The STATCOM has very little 1,000E8
Real power from Nygård to Frydenlund with STATCOM (Blk), Reactive power from Nygård to Frydenlund with STATCOM (Red)

influence on this dynamic behaviour. For (c) there is an 8,000E7

6,000E7

increased swing when the STATCOM is applied, probably 4,000E7

because G4 is much closer to the STATCOM and the

2,000E7

STATCOM is able to influence on this relatively small

Short circuit
-2,000E7

-4,000E7

machine. -6,000E7
0
S
0,5 1,0 1,5 2,0 2,5
C
3,0 3,5
f
4,0
O
4,5
S 2
5,0

Fig. 15. shows rms voltages at the wind farm (a), at (b)
Nygård bus bar (b), and at Frydenlund bus bar (c), without
STATCOM (black graph) and with STATCOM (red graph). Fig. 13. Active (black graph) and reactive power flow (red graph) at
Nygård bus bar, (a) without STATCOM, and (b) with STATCOM.
The figure shows that the wind farm voltage to a small
extent is reduced during the short circuit, and surprisingly It should be noticed that the voltage sensors of the
enough, even less without the STATCOM. At Nygård bus STATCOM was connected to the infinite bus through a
bar and Frydenlund bus bar one observes a severe voltage transformer for the model in Fig. 3. This was done to make
drop during the fault contingency. the system stable. If the voltage sensors are coupled straight
Fig. 16. shows active (black graph) and reactive power to the overhead line between Sirkelvann and the wind farm,
flow (red graph) at Nygård bus bar without STATCOM (a), the system is unstable, probably caused by coupling
and with STATCOM (b). The disturbances caused by the between the active and reactive power properties.
short circuit are so heavy that the influence of the
STATCOM is negligible during these violent transients.
 7
IV. CONCLUSIONS This work has shown that a STATCOM is beneficial
both for voltage control and in order to increase transient
This work has been successful in developing suitable stability during fault conditions. In some cases it might
simulation models for different faults and disturbances. The increase the dynamics of the wind farm and its closest
software tools have worked as expected, and the work has surroundings and maybe be a source of instability under
yielded increased insight and understanding in ATP-EMTP certain conditions.
simulations. As can be seen from Fig. 11. and Fig 14, generator G4
The simulation model needs further improvement, and (Sirkelvann mini hydro generator) can be quite heavily
even though it simulates an existing distribution grid, on- perturbed by the wind farm, because it is situated quite
site measurements are required to validate the model. Both close to the wind farm and because it is a relatively small
the wind turbine model and the STATCOM model might machine with a low moment of inertia. During rated wind
have been simplified beyond realistic behaviour. production one should probably keep the hydro production
Even if the existing model should turn out to be of G4 significantly below the stationary stability margin, in
inadequate, it still gives a rough picture of expected order to avoid generator tripping.
dynamic behaviour of the system.
From Fig. 9. it can be seen that the rise of the wind farm V. REFERENCES
current during a short circuit at Frydenlund is not very
dramatical. The reason is obvious: Due to the weak grid [1] L. Prikler, H. K. Høidalen, “ATPDraw Users’ Manual”, SINTEF
Energy Research, 2002
connection, the total transmission impedance limits the [2] T. Ackermann, Wind Power in Power Systems, England: Wiley, 2005
short circuit current of the wind farm.
The benefit of a weak grid connection is by other words
the inherent current limiting properties. The disadvantage is
naturally the reduced stiffness of the network, and thereby
increased dynamics and reduced stability limits.