Академический Документы
Профессиональный Документы
Культура Документы
I HC BCH KHOA H NI
B MN H THNG IN
PHN III
BO V V T NG HA
TRONG H THNG IN
201
Paper
Non-member
Member
Member
This paper presents an improved control strategy for hybrid series active power lter (HSAF) working
with nonlinear and unbalance three-phase three-wire loads. An algorithm based on the Instantaneous Power
Theory is introduced to precisely extract only harmonic component from supply current, even this current is
contaminated with negative sequence component due to the imbalance of load. An improved control strategy
based on that sequence extraction algorithm is proposed and investigated in detail by numerical simulation.
The proposed control method has shown a better performance in mitigating harmonics, especially for the
nonlinear and unbalanced loads.
Keywords: Series active lter, Instantaneous power theory, Unbalanced load, Harmonic isolation.
1. Introduction
The increasing use of power electronics-based loads
(adjustable speed drives, switch mode power supplies,
etc.) is responsible for the rise in harmonic distortion
levels. These nonlinear loads appear to be prime sources
of harmonic distortion in a power distribution system.
Harmonics have a number of undesirable eects on the
distribution system such as the excessive voltage distortion, increasing resistive losses or voltage stresses. In
addition, the harmonic currents can interact adversely
with a wide range of power system equipment such as
capacitors, transformers, and motors, causing additional
losses, overheating, and overloading. Because of the adverse eects that harmonics have on equipments, many
solutions have been developed to deal with harmonic
control (1)(3) .
Besides conventional solutions such as passive lters,
the hybrid series active power lters have proven to be
an interesting alternative to compensate harmonics in
power distribution systems. Compared to passive lters, active lters provide superior ltering performance,
more exible operation and more compact. There are
various series hybrid active power lter topologies reported in literature (4)(6) , but the most common one is
shown in Fig. 1
Figure 1 shows the system conguration of series hybrid active power lter (APF), in which the shunt passive lter consists of one or more single-tuned LC lters and/or a high pass lter (HPF). The hybrid series APF is controlled to act as a harmonic isolator between the source and nonlinear load by injection of a
202
ia
ib (4)
ic
i
i
1
1/2
0
3 2
1/2
3 2
v
v
v
v
i
i
(5)
p
q
=
=
p + p
(6)
q + q
v2 + v2
v
v
v
v
p
q
(7)
plication
2.1 Brief review of Instantaneous Power Theory
The Instantaneous Power Theory (16) is well utilized for control system of active lter. Control strategy
based on this method provides fast response to changes
in power system, good compensating performance and
imposes a little computational burden (17) (18) .
Figure 2 shows the calculation block of this theory:
Firstly, three-phase voltages and load currents are transformed into the stationary - reference frame (Clarke
transformation):
va
2
1
1/2
1/2
v
=
vb (3)
v
0
3 2 3 2
3
vc
203
An Improved Control Strategy for Hybrid Series Active Filter dealing with Unbalanced Load
the resulted dc components p , q of those power quantities in this case are the cross product of only fundamental positive components as shown below:
current
3.1 Problem formulation and proposal
All
the existing control strategies for the series active power
lter determine the reference currents ih by simply subtracting the fundamental positive sequence current from
the supply current as below:
p
q
(11)
here
I+1 : the fundamental positive sequence current
component of the measured supply current.
+1 : phase angle between the auxiliary positive sequence voltage V+1 and the fundamental positive
sequence current component I+1 .
It is clear to see that only fundamental positive sequence
voltage V+1 and current I+1 components contribute to
average value p and q , the negative sequence components does not appear in those power quantities. Next,
the low-pass lter is utilized to extract only those dc
power components p , q . Once those power components
in Eq. 11 is extracted then it is easy to obtain the positive sequence current using the same denition as shown
in Eq. 7.
For extracting fundamental positive sequence component, the amplitude of v+1 and v+1 are not important
and can be chosen arbitrarily due to the fact that they
appear in both direct and inversecalculations (16) .
For simplicity, they are set to unity hence Eq. 10 becomes:
3V+1 sin (1 t)
v+1 =
(10)
v+1 = 3V+1 cos (1 t)
v+1 = + sin (1 t)
(12)
v+1 = cos (1 t)
3.3 Negative sequence current extraction
Similar procedure is employed to extract negative sequence current, however, an auxiliary negative sequence
voltage is considered instead of the auxiliary positive sequence voltage.
The results of - transform of this auxiliary pure fundamental negative sequence voltage is shown in Eq. 13.
v1
v1
=
=
+ sin (1 t)
(13)
+ cos (1 t)
= 3V1 I1 cos (1 )
(14)
= 3V1 I1 sin (1 )
here
I1 : the fundamental negative sequence current
component of the measured supply current.
1 : phase angle between the auxiliary negative sequence voltage V1 and the fundamental negative
sequence current component I1 .
Again, only fundamental negative sequence voltage V1
and current I1 components show up in the average
value p , q even the supply current is distorted and unbalanced. Therefore, if those dc power components are
Next step, this pure fundamental positive sequence voltage is used together with the supply current to calculate
the instantaneous power quantities p, q following Eq. 5.
The supply current may contain fundamental negative
sequence and high order harmonic components, however,
204
=
=
Fig. 5.
lter
4. Control strategy
4.1 Operation principle of series active lter
as harmonic current isolator
It is well known
that series active lters correct current system distortion caused by non-linear load by synthesizing an active
impedance presenting a zero impedance at fundamental frequency and a high resistance K between load and
source at all harmonics frequencies. By inserting a high
resistance K, the series active lter forces the high frequency current ow mainly through LC passive lter
connected in parallel to load (2) (21) .
The equivalent single phase circuit for harmonic compensation is shown in Fig. 4. In this gure, non-linear
load is represented by a harmonic current source Ih and
source voltage is represented by harmonic voltage source
Vsh . The series active lter is equivalent to a controlled
voltage source Vc and shunt passive lter becomes an
equivalent impedance ZF . If the series active lter is
controlled as Vc = K Ish (equivalent to a resistor of K
ohm) then the Ish can be calculated as:
Ish =
Fig. 4.
Positive sequence current detection circuit
5. Simulation setup
Simulation is setup as following:
The investigated system is three-phase, three-wire
system then zero sequence component does not exist.
Sources line to line voltage is 200V (50Hz). Source
impedance is Zs = 0.0280 pu (with the system
base of Ubase = 200V and Sbase = 20kV A).
Active lter rating capacity is 700V A and it is activated at 0.5 [s] during simulation.
PWM converters switching frequency is set at
ZF
Vsh
+
Ih (15)
ZS + ZF + K
ZS + ZF + K
205
Fig. 3.
An Improved Control Strategy for Hybrid Series Active Filter dealing with Unbalanced Load
Isa
150
Isb
Isc
100
50
-50
-100
-150
0.460
Table 1.
5th lter
7th lter
High-pass
Capacitance [F]
340
170
300
0.500
0.510
0.520
0.530
0.540
0.550
0.560
5.0
Q=14
Q=14
R=3
0.0
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
29
30
31
0.0
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
150
Iap
Ibp
Icp
100
50
-50
-100
-150
0.460
0.470
0.480
Fig. 8.
15.0
Ian
0.490
0.500
0.510
0.520
0.530
0.540
0.550
0.560
Icn
10.0
5.0
0.0
-5.0
-10.0
-15.0
0.460
0.480
Fig. 9.
0.500
0.520
0.540
0.560
0.490
Current Spectrum
Inductance [mH]
1.2
1.2
0.26
0.480
Fig. 6.
0.470
Isa
150
Isb
Current Spectrum
Isc
5.0
100
50
-50
-100
-150
0.0
0.460
0.470
0.480
0.490
0.500
0.510
0.520
0.530
0.540
0.550
0.560
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
25
26
27
28
29
30
31
Fig. 10. Harmonic mitigation with improved control strategy for balanced load
Current Spectrum
5.0
Isa
150
Isb
Isc
100
50
-50
-100
0.0
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
-150
0.460
0.470
0.480
0.490
0.500
0.510
0.520
0.530
0.540
0.550
0.560
150
Isb
Isc
50
-50
-100
-150
0.460
0.470
0.480
0.490
0.500
0.510
0.520
0.530
0.540
0.550
0.560
Fig. 11. Waveshape comparison in cases with improved and previous control strategies
trol algorithms
For comparison purpose, the active
lter which is equipped with the previous control strategy is also simulated. Simulation studies for comparison
are setup based on follow assumptions:
In previous control strategy, the negative sequence
component is not excluded from the reference harmonic current.
On the contrary, for the improved control strategy,
the negative sequence component is excluded from
the reference harmonic current.
Comparisons are carried out with unbalanced loads
since the improved control strategy is proposed to
help the series active lter performs better under
unbalanced loading conditions.
All other conditions remains the same for both
cases.
Figure 11 & 12 compare the current waveshapes and
spectra in cases the active lters utilize the improved
and previous control strategies. Under the same testing conditions, the active lter with improved control
strategy shows a better performance. This conclusion
can be claried in below discussion:
The active lter equipped with previous control
strategy will have to handle both harmonic and the
fundamental negative sequence current components,
consequently the load will be forced to be balanced
as shown in Fig. 11b and the PWM converter might
be easily overloaded. Besides exposing the active
207
100
An Improved Control Strategy for Hybrid Series Active Filter dealing with Unbalanced Load
V_active_filter
20.0
VfilterRMS
Zs = 0.02pu
THDi (%)
16.0
Zs = 0.1pu
THDi (%)
Zs = 0.2pu
THDi (%)
12.0
8.0
4.0
0.0
-4.0
-8.0
-12.0
-16.0
0.869893
0.631617
0.497768
(a) Zs = 0.02pu
(b) Zs = 0.1pu
(c) Zs = 0.2pu
-20.0
0.460
0.470
0.480
0.490
0.500
0.510
0.520
0.530
0.540
0.550
0.560
20.0
VfilterRMS
16.0
8.0
4.0
0.0
-4.0
-8.0
-12.0
-16.0
-20.0
0.460
0.470
0.480
0.490
0.500
0.510
0.520
0.530
0.540
0.550
0.560
Fig. 13.
Zs = 0.02pu
THDv (%)
Zs = 0.1pu
THDv (%)
Zs = 0.2pu
THDv (%)
1.06281
2.01309
2.30476
(a) Zs = 0.02pu
(b) Zs = 0.1pu
(c) Zs = 0.2pu
7. Conclusions
The improved control strategy for hybrid series active
lter is already proposed and investigated in detail. This
control method bases on sequence component elimination algorithm to obtain only the harmonic content from
a distorted and unbalanced current set. Consequently,
this control strategy does help the series active lter to
improve its performance, especially when load is unbalanced. Main conclusions can be recognized as follow:
Only harmonic component is precisely extracted to
form the reference current for the series active lter,
therefore, the active lter work with only harmonic
component as expected, even when load is unbalanced.
The proposed control strategy enhances the stability of control process since the imbalance of load will
not have any aect on the reference signal.
Especially, the active lter is protected away from
severer overload conditions because it does not have
to deal with the fundamental negative sequence
component which may occur if load is unbalanced.
The proposed control strategy is well tailored to suit
with all operating conditions such as serving balanced or unbalanced loads. In other words, it can
apply for the series active lter working with any
generic loads.
Finally, all the simulation results have successfully validated the eectiveness and feasibility of this proposal.
this case is about 8V as shown in Fig. 13b. Numerically, the required volt-ampere output of the active
lter now is up to 8V 60A 3 = 1440V A. In this
case, the active lter is about 1440V A/700V A =
2.05 times overloaded. As a result, this will seriously
damage the converter. Apparently, the percentage
of overload depends on the load unbalance factor.
6.3 Inuence of source side impedance on harmonic mitigation eect
From Fig. 4, it is easy to
see that the harmonics generated by nonlinear load are
injecting into source. As consequence, the source voltage will be distorted depending on the value of source
impedance. If the source impedance is low, then the
voltage distortion is low and vice versa. Since this control algorithm utilizes the quantities calculated from
both voltage and current, then it is necessary to examine
the inuence of voltage distortion (or source impedance)
on the compensation eect.
All above simulations run with source impedance set at
0.02pu, then now two more worse scenarios are examined: source impedances are set higher at 0.1pu and
0.2pu. Consequently, the results in Fig. 14 show that
total harmonic distortion (THD) factors of source voltages are 1.06%, 2.01% and 2.3% respectively. Thus, the
higher the source impedance, the worse the voltage distortion.
Figure 15 show the corresponding THDs of source currents after compensation. Noticeably, after series active
lter was started, the remain amounts of harmonic con-
References
(1)
208
Z. Salam, and T. P. Cheng, and A. Jusoh, Harmonics Mitigation Using Active Power Filter: A Technological Review,
12.0
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
(18)
(19)
(20)
(21)
app. Fig. 1.
=
=
va ia + vb ib + vc ic
p3 + p3
vab ia + vcb ic
(A1)
Appendix
Phase locked loop (PLL) circuit
The PLL is one of components of the sequence cur-
209
(2)
An Improved Control Strategy for Hybrid Series Active Filter dealing with Unbalanced Load
Goro Fujita (Member) received the B.E., M.E. and Ph.D
degrees in electrical engineering from Hosei
University, Tokyo, Japan in 1992, 1994 and
1997 respectively. In 1997, he was a research
student of Tokyo Metropolitan University. He
is an Associate Professor of Shibaura Institute of Technology, Tokyo, Japan. His interest is in power system control including AGC
and FACTS. He is a member of the Society of
Instrument and Control Engineers (SICE) of
Japan, the IEE of Japan, and IEEE.
210
Paper
Non-member
Member
Dynamic Voltage Restorer (DVR) is a series custom compensator utilized in power distribution network,
however, due to connected in series with distribution line then DVR would suer from downstream faults.
To limit the ow of large fault currents and protect DVR itself as well, a fault current limiting function is
proposed in the DVR control strategy. Fault current limiting function of DVR will be activated by protection
system and then DVR will start injecting a series voltage to the line in such a way to limit fault current
to an appropriate level (in accordance with required sensitivity of protection systems). The contribution of
this proposal is the utilization of signals from existing feeder protection system to activate DVR. This will
simplify the structure of DVR because the extra build-in fault detection module is not required. Moreover,
it will ensure proper coordination between DVR and protection systems.
1. Introduction
Distribution networks are usually expanded by adding
either extra transformers and/or feeders. This fact may
raise the chances of increasing prospective fault currents
that might exceed circuit breakers interrupting capacity. Moreover, that adding dispersed power sources may
be convinced as other reason for increasing potential
fault current level. Many solutions have been proposed
to deal with high fault duty level; one of these solutions
is utilizing fault current limiting (FCL) devices. Basically, FCL can bring many benets to utilities, such as:
Avoid or delay upgrading existing CBs.
Minimize voltage dip on upstream bus when fault
occurs at downstream feeder.
A larger transformer can be used to meet demand
without upgrading CBs.
Reduce thermal damage due to fault currents and
hence protecting and extending life time of transformers and other equipments.
In this study, a dynamic voltage restorer (DVR) is
proposed as a fault current limiter. Dynamic voltage
restorer is a series custom power device used to protect
sensitive customers from impacts of all voltage disturbances. Dynamic voltage restorer can be implemented
at both low voltage level and medium voltage level. A
topology of typical medium voltage DVR is in Fig. 1.
The basic operation principle of DVR is simple. DVR
will insert a series voltage with appropriate magnitude,
frequency and phase to compensate for any voltage disturbances (especially voltage sags) that may aect the
proper operation of loads. In order to handle that oper
Fig. 1.
211
Fig. 2.
changed to Eq. 2:
current limiter
DVR is known as a multi-function device in distribution system, it mainly used to against voltage sag that
may occur. To minimize the power losses through DVR
then the DVR is held in a null state in normal condition. Once an overcurrent occurs then FCL function is
activated and it starts to react as fast as possible and
inject the required ac voltage to the grid.
Figure 2 and 3 illustrate that operation principle:
When fault occurs, the relationship between voltage at
source and fault current can be expressed as (without
DVR):
US = If Rf + jIf Xf
= URf + UXf = U (1)
where
US : source voltage (behind sources impedance)
Zf : total fault impedance (Zf = ZS + ZLf ault )
If : fault current
ZS : internal impedance of source
ZLf ault : impedance of faulty line section
U : voltage drop on total fault impedance
If DVR is activated, it will inject a series voltage into
line in such a way to reduce amplitude of fault current.
At this time, relationship as shown in Eq. 1 has been
212
Fault Current Limiting Function of Dynamic Voltage Restorer Utilizing Signals from Existing Protective Relays
Sensitivitymin =
If ault min
Ipickup level
Fig. 5.
[pu]
213
E = Pc tprot (3)
where:
Pc : total active power owing into DVR.
tprot : operating time of protection system.
Total active power Pc owing into DVR is determined
by:
Pc = 3 Vinjected Icompensated cos() (4)
UDCmax =
2E
(5)
C
Fault Current Limiting Function of Dynamic Voltage Restorer Utilizing Signals from Existing Protective Relays
Ph
BK
A
1.0
B
Ibsend
0.002 [H]
Iasend
Icsend
Vasend
Vbsend
Vcsend
0.6 [ohm]
Vsc
Vsb
Vsa
#1
#1
#1
#2
P+jQ
#2
#2
Vic_a
Vic_b
Vic_c
Fault Type
3
5
2
g21
g61
1
2
2g51
g31
g41
A
Timed
Fault
Logic
g11
FAULTS
4000.0 [uF]
6
2
2g52
6
2g32
g22
g62
Fig. 9.
Sin
Iaref
D + -
F
Ia
Va
Ia
Vb
Ib
Ic
Vc
PLL
theta
Sin
D + F
Ibref
D + F
Ib
120.0
D + +
Sin
4
2 g12
Icref
D + -
F 120.0
F
Ic
g42
P
Delta_a
P
Delta_b
P
Delta_c
*
1
Limiter
Fig. 10.
Ib
1.0k
0.8k
0.8k
0.5k
0.5k
0.3k
0.3k
0.0
1.0k
Ib
-0.3k
-0.5k
-0.5k
-0.8k
-0.8k
-1.0k
0.000
0.050
Fig. 12.
0.100
0.150
0.200
0.250
0.300
0.000
0.050
Fig. 14.
0.100
0.150
Ia
Ib
1.0k
0.300
Ia
Ib
Ic
0.8k
0.5k
0.5k
0.3k
0.3k
0.0
y
0.250
Ic
0.8k
-0.3k
0.0
-0.3k
-0.5k
-0.5k
-0.8k
-0.8k
-1.0k
0.000
0.200
Ic
0.0
-0.3k
-1.0k
Ia
0.050
0.100
0.150
0.200
0.250
-1.0k
0.300
0.000
0.050
Fig. 15.
0.100
0.150
0.200
0.250
0.300
1.0k
mally higher).
As stated in Sect. 3.1, minimum required sensitivity
of protection system should not be less than 2 pu, therefore, setting level of compensated current should be:
Ia
Ib
Ic
0.8k
0.5k
0.3k
0.0
-0.3k
-0.5k
-0.8k
In other word, when DVR operates as fault current limiter it should reduce fault current magnitude to about
230 [A]. All the faults are assumed to start at 0.03 [s].
5.2 Simulation procedure
Simulation is carried out with various scenarios:
Fault occurs at feeder side: this scenario used
to verify the eectiveness of mechanism to activate/terminate DVR by FPS.
Fault occurs at load side: This simulation is to verify the proposal used to terminate DVR action by
tracing DC link voltage and to verify the proper coordination between FPS and LPS.
5.3 Result and discussion
Simulation results
and discussion are shown below.
Figure 12 shows three-phase fault current (fault takes
place at feeder) with DVR is not activated, fault current is about 700 [A] (about ten times of normal load
current). In case DVR is activated (Fig. 13), DVR
takes full eect just after one cycle and fault current is
reduced to 230 [A] (rms) as expected. In this case, DVR
is activated and terminated by FPS.
Comparing Fig.12 with Fig. 13, it can be seen that
the operation of FPS is not interfered even while DVR
is operating, FPS cleared fault after 0.2 [s] based on its
own setting. Similarly, Fig. 14 and Fig. 15 show the
case of phase-phase fault.
In Fig. 15, it can be seen that, DVR performed single phase control perfectly, only two faulty phases (a
and b) are compensated while healthy phase (c) remain
constant. Moreover, fault clearance time of relay still
-1.0k
0.000
Fig. 16.
DVR
0.050
0.100
0.150
0.200
0.250
0.300
is 0.2 [s] that means DVR operation did not aect the
protection system.
In next section, the cases with three-phase fault at
load side (as shown in Fig. 7) is investigated. Firstly,
DVR is not activated. In this situation, fault current is
570[A] (Fig. 16) and it is smaller than that in case fault
took place at feeder because now fault location is moved
far away the source. LPS cleared the fault after 0.1 [s] as
expected and whenever fault is cleared hence current get
back to normal load current (current did not go down to
zero because only faulty part is eliminated and healthy
part is untouched). Apparently, restored load current is
lower than that of pre-fault condition because the faulty
section is eliminated.
Secondly, DVR is utilized. In this case DVR is activated by FPS because when fault occurs hence both
FPS and LPS pick up. Figure 17 shows that DVR again
fully perform its eect just after one cycle as shown in
other cases.
When fault occurs, DC link voltage goes up (Fig. 18)
since DVR is absorbing energy from upper source. Next,
LPS eliminates the fault after 0.1 [s] (at the same time
FPS reset to standby), DVR immediately reverse its action this leads to DC link voltage goes down suddenly.
216
Fault Current Limiting Function of Dynamic Voltage Restorer Utilizing Signals from Existing Protective Relays
CURRENT (in Ampere)
Ia
Ib
Ic
(5)
0.8k
0.5k
0.3k
(6)
0.0
-0.3k
-0.5k
-0.8k
(7)
-1.0k
0.000
0.050
0.100
0.150
0.200
0.250
0.300
(8)
(9)
Appendix
Main : Graphs
12.0
DC voltage
Parameters in simulation
- Source & Load (Fig. 9)
10.0
8.0
6.0
y
4.0
2.0
0.0
-2.0
0.000
Fig. 18.
0.050
0.100
0.150
0.200
0.250
0.300
- Coupling transformer
At the time DC link voltage drops down then consequently DVR is terminated and fault current get back
to normal level. This simulation again validates the theoretical analysis shown in Sect. 3.3.
- Fault resistance
The value of fault current can be adjusted by changing fault resistance.
6. Conclusion
In this paper, the dynamic voltage restorer is proposed
to use as a fault current limiter. A new method to activate that fault current limiting function by the signals
from existing protection system is proposed and tested.
Moreover, the coordination between existing protection
system and fault current limiting function of dynamic
voltage restorer is investigated and implemented. This
proposal oers many advantages such as the reduction
in the cost of DVR device; the proper coordination between protection systems is ensured regardless of the
fault location and even when the fault current has been
altered. Finally, all the simulation results have validated
the eectiveness of proposal.
References
(1)
(2)
(3)
(4)
Fujita Goro (Member) was born in January 1970. He received the B.E., M.E. and Ph.D degrees in
electrical engineering from Hosei University,
Tokyo, Japan in 1992, 1994 and 1997 respectively. In 1997, he was a research student of
Tokyo Metropolitan University. Since 1998, he
is in Shibaura Institute of Technology, Tokyo,
Japan as an associate professor. His interest
is in power system opeartion and control. He
is a member of the Society of Instrument and
Control Engineers (SICE) of Japan, the IEE of Japan, and IEEE.
217
1.0k
Paper
Non-member
Member
Member
This paper describes a new proposal to deal with imbalance phase loading phenomenon in power distribution system. Discrete switched passive shunt compensators such as reactors or capacitor banks are considered
as the means to compensate for that imbalance phenomenon. Discrete switched passive compensator oers
advantages in term of installation cost and simplies the maintenance process.
A new algorithm is developed to calculate the size of discrete compensators, this algorithm incorporates
the unbalance power ow calculation module and optimal compensator sizing module. In addition, the algorithm is written in MATLAB language and tested on an actual three phase-three wires distribution feeder.
Extensive tests have validated the eectiveness of the proposal and shown that this proposal can be a useful
tool for any electrical utilities.
gories:
Rearrange feeders or redistribute the loads in such a
way the system becomes more balanced (3)(5) . However, the utilities usually cannot aord too many
load swapping due to the long time interruption of
customers and cost of labors to implement those operations.
Install compensators (power quality conditioners)
to compensate for any imbalance quantities (6)(9) .
They seem to be the most possible solutions but another concern raised by utilities is the capital cost
of these solutions.
This paper focuses on the second mitigation technique
and will establish an appropriate solution in term of capital cost. For imbalance compensation purpose, active
compensator such as Distribution Static Compensator
(DSTATCOM) is widely introduced in literature, the
power electronic solutions are elegant, however, they include a greater degree of cost. Cost has been the major factor in limiting the application of power electronic
solution over power distribution level. Based on that
fact, solutions those involve passive compensators such
as switched capacitor or switched reactor banks would
seem to be the most cost eective solution and these will
be investigated in this paper.
Moreover, a new algorithm is developed to calculate
the size of compensators. This algorithm includes unbalance power ow calculation module and discrete optimal compensator sizing module. The objective of this
algorithm is to nd the optimal size of compensators so
that minimize the unbalanced power ow through the
main feeder. The algorithm allows the user to specic
the type and the maximum number of available taps
for each compensator. In addition, detailed model of
1. Introduction
In power system, voltages (and currents) are expected
to be sinusoidal and equal in magnitude, with the individual phases 1200 apart. However, the utilities usually experience the imbalance phenomenon in both voltage and current. The nature of imbalance phenomenon
includes unequal magnitude and phase angle deviation
among phases. A major cause of voltage and current imbalances is that loads are not uniformly spread among
the three phases and load peaks are non-coincident due
to diversity of load categories. Additional causes of
power system imbalances can be asymmetrical distribution feeder impedances possibly caused by incomplete
transposition of feeder lines.
The inuence of imbalance voltage and current on
power system has been well investigated in literature.
V. J. Annette et al. (1) and L. F. Ochoa et al. (2) concluded that unbalanced voltages and currents can result
in adverse eects on equipment and on the power system, for example a small unbalance in the phase voltages can cause a disproportionately larger unbalance in
the phase currents. Under unbalanced conditions, the
power system will incur more losses and heating eects.
The eect of voltage unbalance can also be severe on
equipment such as induction motors, power electronic
converters and adjustable speed drives.
Many mitigation techniques have been developed to
deal with imbalance phenomenon in distribution system,
generally those solutions can be divided into two cate
218
Keywords: Phase Loading Imbalance, Discrete Optimization, Passive Compensation, Distribution System.
2. Compensation principle
Figure 1 shows the general unbalanced three-phase
load fed from a three-phase, three-wire source. Load
and compensator are connected in delta therefore zero
sequence component
will . equal to zero. The compen.
.
sator currents (I ac , I bc , I cc ) is added to load currents
.
.
.
(I al , I bl , I cl ) and then the following equation is satised
(written for phase A, similarly for phase B and C):
.
I a = I al + I ac
.
.
.
.
.
.
= (I a1l + I a2l + I a0l )+(I a1c + I a2c + I a0c )
.
0
V A = V 0 ,
0
V C = V 240
calculated as
V
V
(4)
.
.
0
V B = V 120 ,
I = Y ab (1 a2 ) Y ca (a 1)
. ac
.
.
I bc = Y bc (a2 a) Y ab (1 a2 )
.
.
.
I cc = Y ca (a 1) Y bc (a2 a)
Now considering some particular cases where the compensators are single-phase elements only. The possible
congurations are shown in Fig. 3.
Case (a): Single-phase compensator is connected
across phase A and phase B (Ybc = 0, Yca = 0).
In order to generalize the results, the values used in
this section will be in per unit.
Substitute (Yab , Ybc = 0, Yca = 0) into Eq. 5, current sequence components produced by this single
phase compensator are:
(3)
219
Fig. 1.
Phase Load Balancing In Distribution Power System Using Discrete Passive Compensator
I a2(AB)
.
I a2(BC)
.
I a2(CA)
Fig. 5. Decomposition of required negative current vector into two nearest component vectors
Compensator
Capacitive Inductive
1300
11500
1900
1900
11500
1300
Fig. 4. Vector diagram of negative current components generated by various single phase compensator congurations
.
I. a0(AB) = 0.
(6)
= Y ab V
I
.
. a1(AB)
2
I a2(AB) = a Y ab V
Here subscript AB denotes current quantities resulting from the single phase compensator connected across phases A&B.
If the compensator is capacitive element with
impedance
of XC
.
.
2
I a2(AB) = a Y ab V
V
300 = 1300 (7)
=
XC
If the compensator is inductive element with
impedance
of XL
.
.
2
I a2(AB) = a Y ab V
V
1500 = 11500 (8)
=
XL
Case (b) & (c): Similar calculations are applied for
case (b) and case (c).
system
4.1 Problem formulation
The implementation
of load balancing or load compensation in distribution
level may involve some of follow aspects:
Technical aspects:
Compensation algorithm must be able to apply to
calculate for not only individual load but also for
feeder with several connected loads. The problem
of unbalanced phase loading is not new in power
systems, however, the previous proposed solutions
are to deal with single individual unbalanced load
only and they did not gure out how to deal with
the feeder which has several connected loads (6)(9) .
The contribution of this paper is to propose an algorithm which can apply to solve the unbalanced
phase loading phenomenon for not only single load
but also for the feeder accommodating several loads
Table 1 and Fig. 4 show the results of sequence components contributed by various single phase compensator
congurations.
Assume that the compensator must
generate a required negative sequence current as shown
in Fig. 5. This required vector will be decomposed into
two nearest available vectors:
Case a: If the compensator is capacitive then the required vector can be decomposed into the two nearest component vectors on CA and BC axes with
220
Fig. 3.
Representative feeder
as seen in practice.
Algorithm must produce as more accurate results
as possible in comparison with other proposed solutions: in literature, all the calculations were
based on assumption that voltages are perfectly balanced (6) (7) , but actually voltages at load terminal
may not be always balanced due to many reasons.
In order to overcome this limit and to provide a more
accurate solution, this proposed algorithm will use
the actual node voltages (not assumed voltages) in
all calculations.
Cost advantage aspects: The key contribution of this
paper is that proposing an economically justiable compensation solution.
In order to reduce investment cost and make it applicable to power distribution level then only either
pure capacitor bank or pure reactor bank is considered to form the compensator as stated in Sec. 3,
mixed capacitive and reactive compensator bank is
not an option.
As stated in Sec. 1, power electronic compensators
or active compensators are a perfect choice but these
solutions are far more expensive than that if the passive compensators are selected.
For that consideration, the discrete switched passive compensator banks (such as discrete capacitor
or discrete reactor bank) will be utilized as the compensator.
4.2 Phase load balancing algorithm
The imbalance phase loading phenomenon does not immediately show its inuence over system and equipments due
to the fact that thermal inertias of equipments are quite
long. Based on that analysis, it is not really imperative
to instantly correct the imbalance phase loading situation since it occurs and the balancing action can be
carried out on averaged data over a time interval. In
other word, if daily loading curve is known then it can
be stripped into smaller intervals and the compensation
action will be calculated based on the unbalance factor
averaged over each that time interval.
From above view point, load balancing algorithm is proposed as follow: Considering the representative feeder
with m loads as shown in Fig. 6.
Step 1 : Determine the rst loading level (n = 1) to assess for imbalance compensation (here n stand for loading level order).
Step 2 : Run three-phase, unbalance power ow calculation with the given loading level from step 1.
This unbalance power ow module has ability to handle
for various types of loads such as the load with constant power; load with constant impedance or load with
constant current (11) . Furthermore, mutual impedance of
.
.
1 .
(I ali + a2 I bli + aI cli )
3
= Ci + jDi (9)
I a2li =
.
.
1 .
(I aci + a2 I bci + aI cci )
3
= Ei + jFi (11)
I a2ci =
Ei (Y ) + Ci = 0
Fi (Y ) + Di = 0
221
Fig. 6.
Phase Load Balancing In Distribution Power System Using Discrete Passive Compensator
Fig. 7.
16
Before
Fig. 8.
Percentage (%)
14
After
12
10
8
6
4
2
16
14
After
Time Interval
10
11
12
Fig. 10. Negative currents before & after compensation through Line 2
12
10
8
Before
50
Time Interval
10
11
Percentage (%)
2
0
12
After
40
30
20
10
0
pacitive compensators are used then the power factor will become worse.
Compensators are switchable with maximum 10
steps and each step is 25kVAR; 15kVAR and 5
kVAR for compensator 1; compensator 2 and compensator 3 respectively.
In this paper, time interval for unbalance assessment is
selected as 2 hours, in other word, each compensator will
have 12 time-switching patterns over every 24 hours or
a day.
The load balancing algorithm is written in MATLAB.
This algorithm consists of two main modules.
The rst is unbalance power ow module. This unbalance power ow module has ability to handle for
various types of loads such as the load with constant
power; load with constant impedance or load with
constant current. Furthermore, mutual impedance
of distribution line is also taken into account (11) .
The second module is implementation of brute-force
search algorithm. This search algorithm is simple (10) , however, calculation time will grow signicantly when the size of problem increases. In power
distribution level, the discrete switched capacitor
bank (or reactor bank) has only a few number of
taps so that if even brute-force search is applied then
the calculation time will not be a matter.
5.2 Result and discussion
5.2.1 Eectiveness of compensation algorithm
The eectiveness of compensations are illustrated in
Fig. 9, 10 and 11. Based on results of negative current
before and after compensation, it is clear that the proposed compensation strategy shows a signicant load
balancing eect, the magnitudes of negative currents
that represent the imbalance degree may reduce to less
than 1% in some cases. However, the full compensation eect (resulting in zero negative current) cannot be
expected due to the discrete characteristic of compensators.
In addition, the node voltages become balanced as a
Time Interval
10
11
12
Fig. 11. Negative currents before & after compensation through Line 3
Negative voltage at node 1
0.8
Before
Percentage (%)
0.7
Percentage (%)
Before
After
0.6
0.5
0.4
0.3
0.2
0.1
0
Time Interval
10
11
12
0.8
Before
Percentage (%)
0.7
After
0.6
0.5
0.4
0.3
0.2
0.1
0
Time Interval
10
11
12
Phase Load Balancing In Distribution Power System Using Discrete Passive Compensator
Before
0.7
After
0.6
0.5
0.4
0.3
0.2
Time Interval
10
11
6600
6400
6200
5600
Phase A
Phase B
Phase C
Time Interval
10
11
12
300
6600
AB
BC
CA
250
6400
6200
6000
200
150
100
50
5800
Time Interval
10
11
0
1
12
Fig. 18.
Time Interval
Phase B
Phase C
Phase A
After
6600
6400
11
12
160
6800
10
180
7000
BC
CA
140
120
100
80
60
40
20
6200
0
1
6000
Fig. 19.
5800
Time Interval
10
11
12
Time Interval
10
11
12
60
AB
BC
CA
50
40
30
20
10
0
1
Fig. 20.
P LI =
Time Interval
10
11
12
P Lbf P Laf
100
P Lbf
(%)
where P Lbf and P Laf are total power losses before and
after compensation.
According to results of simulation, power losses improvement can be more than 70% and in the the least
remarkable situation it is about 16%. The possible reasons for power losses improvement can be convinced as
follow:
As stated in Ref. (2), under unbalance loading situa224
After
6800
After
7000
5600
Phase C
5800
12
5600
Phase B
6000
0.1
0
Phase A
6800
Percentage (%)
7000
0.8
80
10
With Incorporation
Without Incorporation
60
Percentage (%)
Percentage (%)
70
50
40
30
20
6
4
2
10
Fig. 21.
Time Interval
10
11
12
Line 1
Line 2
Line 3
Time Interval
Fig. 22. Average negative currents after compensation with & without iteration technique
Appendix
1. Case study parameter
app. Table 1.
Impedance ()
Length (km)
app. Table 2.
Phase
Value
Phase
Value
Phase
Value
Line 1
0.5557+j0.9982
3.629
Line 2
0.3781+j0.4154
1.024
Line 3
0.3803+j0.558
1.211
Load 1
BC
153.46-j200.63
Load 2
AB
BC
95.59-j140.78 83.43-j122.45
Load 3
AB
BC
27.13-j2.48
14.27+j3.29
AB
183.78-j222.8
CA
150.84-j233.9
CA
80.04-j141.96
CA
16.71-j11.32
0
1
Phase Load Balancing In Distribution Power System Using Discrete Passive Compensator
(9)
(10)
(11)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
226
References
(1)
B. Singh, S. Anuradha, and D. P. Kothari, Power factor correction and load balancing in three-phase distribution systems, IEEE region 10 international conference on global
connectivity in energy, computer, communication and control, (1998)
P. Venkataranman: Applied optimization with MATLAB programming, John Wiley & Sons, Inc. (2002)
H. K. William: Distribution system modeling and analysis,
CRC Press (2007)
%jL WRiQ E WL X F{QJ VXW SKQ NKiQJ FKR OL LQ SKkQ SKL KuQK WLD Oj EjL WRiQ WuP Y WUt
Yj GXQJ OQJ WL X FKR FiF W E VDR FKR KjP WQ WKW LQ QQJ XF JLP WL D %jL WRiQ WL X
Y WUt Yj GXQJ OQJ E Oj PW EjL WRiQ TX\ KRFK SKL WX\Q YL FiF ELQ QJX\rQ Yj WKF KQ KS k\
Oj PW EjL WRiQ SKF WS F ELW NKL iS GQJ FKR FiF OL SKkQ SKL OQ WUrQ WKF W 7URQJ EjL EiR
Qj\ FiF WiF JL [XW PW SKQJ SKiS FKLD QK EjL WRiQ WL X WKjQK KDL EjL WRiQ FRQ EjL WRiQ
[iF QK GXQJ OQJ WL X F JLL EQJ SKXQJ SKiS [S [ KyD OLrQ WLS TX\ Y EjL WRiQ TX\
KRFK WX\Q WtQK Yj WKXW WRiQ ORL E FiF W E Fy yQJ JyS tW QKW FKR KjP PF WLrX &iFK WLS FQ
Qj\ FKR SKpS V GQJ FiF F{QJ F JLL EjL WRiQ TX\ KRFK WX\Q WtQK Fy KLX QQJ FDR QK &3/(;
JLL EjL WRiQ [iF QK GXQJ OQJ W E 7KXW WRiQ [XW m F WK QJKLP WKjQK F{QJ YL
PW YjL OL SKkQ SKL WKF W FD 9LW 1DP YL NW TX WW
, ,1752'8&7,21
7KH XVH RI FDSDFLWRUV LQ SRZHU V\VWHPV
KDV ORQJ EHHQ DQ HIIHFWLYH PHDVXUH WR HQKDQFH
WKH V\VWHP YROWDJH TXDOLW\ UHGXFH HQHUJ\
ORVVHV DQG LPSURYH V\VWHP VWDELOLW\ HWF 7KHUH
DUH VHYHUDO REMHFWLYHV WKDW FDQ EH FRQVLGHUHG LQ
D FDSDFLWRU VL]LQJ DQG SODFHPHQW QDPHO\
FDSDFLWRU DOORFDWLRQ SUREOHP )RU WKH
WUDQVPLVVLRQ JULGV WKH FDSDFLWRU DOORFDWLRQ LV
LPSOHPHQWHG PDLQO\ WR HQKDQFH WKH V\VWHP
YROWDJH SURILOH WR SURYLGH ORFDO UHDFWLYH SRZHU
VXSSRUW DQG LQ VRPH FDVHV HQKDQFH V\VWHP
VWDELOLW\ )RU WKH GLVWULEXWLRQ JULGV WKH
FDSDFLWRUV DUH SODFHG PDLQO\ WR UHGXFH HQHUJ\
ORVVHV >@ $SDUW IURP ORVV UHGXFWLRQ PDQ\
RWKHU EHQHILWV DUH REWDLQDEOH E\ FDSDFLWRU
SODFHPHQW VXFK DV FDSDFLW\ UHOHDVH UHGXFWLRQ
227
70 77
4RXW N 4N [N 3N 4N 9N 4/ N
9N UN 3N [N 4N
9N
ZKHUH
3N 4N DFWLYH DQG UHDFWLYH SRZHU IORZV LQ WR WKH
UHFHLYLQJ HQG RI EUDQFK N
9N 9N YROWDJH DW QRGH N DQG N
UN[N UHVLVWDQFH DQG UHDFWDQFH RI EUDQFK N
3RXWN 7KH VXP RI DFWLYH SRZHU IORZV WKURXJK
DOO EUDQFKHV FRPLQJ RXW RI EXV N
4RXWN 7KH VXP RI UHDFWLYH SRZHU IORZV
WKURXJK DOO EUDQFKHV FRPLQJ RXW RI EXV N
3/N 4/N DFWLYH DQG UHDFWLYH ORDGV DW EXV
N
,Q DGGLWLRQ WR WKH EUDQFK IORZ
HTXDWLRQV ZULWWHQ IRU 0 EUDQFKHV ZH DOVR KDYH
WKH WHUPLQDO FRQGLWLRQ LH WKH YROWDJH DW WKH
PDLQ IHHGHU 9 LV NQRZQ 7KHUHIRUH IRU D
UDGLDO QHWZRUN RI 0 EXVHV WKHUH DUH 0
XQNQRZQ >9N3N4N@ N 0 DQG 0 EUDQFK
IORZ HTXDWLRQV 7KH EUDQFK IORZ HTXDWLRQV FDQ
EH ZULWWHQ LQ D PRUH FRPSDFW IRUP DV IROORZV
[N J N [N
ZKLFK FDQ EH VROYHG XVLQJ WKH EDFNZDUG
IRUZDUG VZHHS PHWKRG 7KH GHWDLOV RI WKLV
PHWKRG FDQ EH IRXQG LQ >@>@ 2QH DGYDQWDJH
RI XVLQJ WKH EUDQFK IORZ HTXDWLRQV LV WKDW
WULJRQRPHWULF FDOFXODWLRQV DUH QRW LQYROYHG
7KHUHIRUH WKH FRPSXWDWLRQ VSHHG LV JHQHUDOO\
PXFK IDVWHU WKDQ WKH VROXWLRQ RI FRQYHQWLRQDO
SRZHU IORZ HTXDWLRQV >@ 1RWH WKDW OLQH
FKDUJLQJ LV QRW H[SOLFLWO\ VKRZQ LQ WKH EUDQFK
IORZ HTXDWLRQ +RZHYHU WKH OLQH FKDUJLQJ
HOHPHQWV FDQ EH DSSUR[LPDWHO\ DFFRXQWHG IRU
LQ WKH UHDFWLYH SRZHU GHPDQG DW VHQGLQJ DQG
UHFHLYLQJ EXVHV
% 3UREOHP IRUPXODWLRQ
6LQFH WKH REMHFWLYH LV WR PLQLPL]H
HQHUJ\ ORVV LW LV QHFHVVDU\ WR WDNH LQWR DFFRXQW
WKH ORDG YDULDWLRQ IRU D JLYHQ SHULRG RI WLPH 7
:H DVVXPHG WKDW WKH ORDG YDULDWLRQ IROORZV D
GLVFUHWH SDWWHUQ DV LQ )LJ 7KH ORDG GXUDWLRQ
FXUYH LV QRUPDOO\ JLYHQ IRU D SHULRG RI RQH
\HDU K +RZHYHU IRU FDSDFLWRU
DOORFDWLRQ FDOFXODWLRQ RQH FDQ H[SDQG WKH ORDG
228
/RDG OHYHO
$ &DSDFLWRU VL]LQJ
7LPH K
PLQ I [
L
PLQ 'I M
X &N
& N
N
L
N
L 1
> [ [ [ 1 @7
[L
L
L
[LM [PLQ
d '[ L d [PD[
[ LM
J L [ L L 1
L N
1&
VXFK WKDW
ZKHUH
'3L SRZHU ORVVHV LQ WKH WLPH LQWHUYDO L
N
I M I M d H
229
,9 5(68/76
7KH DOJRULWKP SURSRVHG LQ WKH SUHYLRXV VHFWLRQ
KDV EHHQ VXFFHVVIXOO\ WHVWHG ZLWK VHYHUDO UHDO
GLVWULEXWLRQ QHWZRUNV RI 9LHWQDP 7ZR FDVH
VWXGLHV ZLOO EH SUHVHQWHG KHUH 7R VROYH WKH
VXFFHVVLYH /3 SUREOHP ZH XVH WKH ZHOO
NQRZQ KLJK SHUIRUPDQFH &3/(; VROYHU
GHYHORSHG E\ ,%0 >@
$ *LHQJ 'D\ N9 GLVWULEXWLRQ QHWZRUN
230
[
[
2EMHFWLYH
2EMHFWLYH
,WHUDWLRQ
,WHUDWLRQ
9 &21&/86,21
,Q WKLV ZRUN DQ DOJRULWKP WR VROYH WKH
FDSDFLWRU DOORFDWLRQ IRU UDGLDO GLVWULEXWLRQ
QHWZRUN LV SURSRVHG 7KH VDWLVIDFWRU\ UHVXOWV
KDYH SURYHG WKDW WKH SURSRVHG DOJRULWKP FDQ
ZRUN ZHOO ODUJH UHDO GLVWULEXWLRQ QHWZRUNV
)XUWKHUPRUH LWV DSSOLFDWLRQ LV SRVVLEO\
H[WHQGHG IRU PXFK ODUJHU V\VWHPV
5()(5(1&(6
5$ *DOOHJR $- 0RQWLFHOL DQG 5 5XEHUR 2SWLPDO &DSDFLWRUV SODFHPHQW LQ UDGLDO GLVWULEXWLRQ
QHWZRUNV ,((( 7UDQVDFWLRQV RQ SRZHU V\VWHPV YRO
0% /LX &$ &DQL]DUHV DQG : +XDQJ 5HDFWLYH SRZHU DQG YROWDJH FRQWURO LQ GLVWULEXWLRQ
V\VWHPV ZLWK OLPLWHG VZLWFKLQJ RSHUDWLRQV ,((( 7UDQVDFWLRQV RQ 3RZHU 6\VWHPV YRO
231
To cite this article: Huy Nguyen-Duc, Louis-A Dessaint & Aim Francis Okou (2011): Analysis and
Design of a Robust Power System Damping Controller Considering Time Delay, Electric Power
Components and Systems, 39:3, 254-270
To link to this article: http://dx.doi.org/10.1080/15325008.2010.526995
232
1. Introduction
In todays power systems, exible AC transmission system (FACTS) devices are increasingly utilized. Their primary application is to provide a tool for better control of power
ows, as well as an additional reactive power support. On the other hand, FACTS devices
can be used to enhance the damping of power system electromechanical oscillations.
The electromechanical oscillations are a common problem in power systems. Traditionally, damping of these oscillations is done with power system stabilizers (PSSs),
based on the phase compensation technique [1]. A FACTS damping controller can also be
designed based on this principle [2]. However, an important drawback of using FACTS
as a damping controller is that the open-loop system seen by FACTS is very dependent
of power system operating conditions [3]. In fact, it is particularly challenging to design
a FACTS that performs satisfactorily over a wide range of operating conditions [3].
Recognizing this fact, much research has been directed toward designing robust FACTS
damping controllers. The application of H1 control theory to design a thyristor controlled
series capacitor (TCSC) damping controller was discussed in [4]. A general guideline
Received 12 January 2010; accepted 9 July 2010.
Address correspondence to Professor Louis-A Dessaint, Dpartement de gnie lectrique,
cole de Technologie suprieure, 1100 Notre Dame Ouest, Montral, Quebec, H3C 1K3, Canada.
E-mail: louis.dessaint@etsmtl.ca
254
233
255
of choosing weighting functions for a robust FACTS controller was discussed in [5].
A FACTS-based stabilizer using robust linear matrix inequality (LMI) techniques and
-synthesis have also been studied [6, 7].
The changes in system operating condition can be treated as a model uncertainty,
which is described in terms of a structured perturbation. This approach has been studied
in [8]. The advantage of this approach is that it allows very little conservative evaluation
of robust stability. The results in [8] have been successfully implemented in many power
system robust control analysis [7, 9, 10].
The choice of FACTS location and input signal for damping control is based on
controllability/observability analysis of various input/output signals with respect to the
oscillation mode of concern. It is possible that the best input and output signals are not
geographically close to each other, which leads to a wide-area controller. In such a case,
the problem of dealing with communication delays must also be resolved. Many works
on wide-area control incorporate a xed time delay in the open-loop system [11, 12].
An evaluation of how time delay affects wide-area controller performance was provided
in [13]. In [12], time delay was compensated in the controller using the Smith predictor
technique.
One important issue with the design of FACTS damping control is unwanted interactions with other power system controllers. A FACTS damping control circuit is intended
only to enhance damping of a few electromechanical oscillation modes; it is not desired
for the control loop to interact with other system modes. Unfortunately, high-frequency
interactions are common with many FACTS controllers [14]. The problem is further
complicated by the presence of time delay, as will be shown in this article. Therefore, an
ideal FACTS control loop should have good controllability/observability with the most
critical electromechanical oscillations while having very low controllability/observability
with other modes. Such an ideal control loop does not exist, and thus, the controller must
be designed in such a way as to reduce the interaction problem as much as possible.
In this article, a robust control design based on the -synthesis method, which takes
into account all of the above issues, is presented. One main contribution of the article
is the development of a time delay model for use with robust control synthesis and a
detailed analysis of how time delay affect the interaction between a FACTS controller
and other system controllers. In addition, common knowledge in power system damping
control design is used throughout the -synthesis process; thus, the design parameters
can be done in manner other than trial and error.
This article is organized as follows. Section 2 provides a background on power system damping control design and the framework for robust control synthesis in a power
system. Section 3 presents a detailed analysis of the Kundur power system [15], where
the drawbacks of conventional approaches to damping control are revealed. Section 4
presents a robust control approach using -synthesis. Section 5 gives some discussions
and conclusions.
2. Background
2.1.
Most power system damping controllers are synthesised based on a linearized model of
an open-loop power system. Among many linear control approaches, the residue-based
phase compensation technique is probably the most widely used in power system damping
control synthesis.
234
256
H. Nguyen-Duc et al.
Given the open-loop power system between measurement y and reference input
u, denoted by G.s/, where the system exhibits a weakly damped mode i . Let Ri
be the residue associated with mode i ; a small change in the eigenvalue i caused by
the feedback gain K is given by
i
D Ri K:
(1)
To have a damping effect on mode i , the controller should move the pole i to the
left half of the complex plane. This can be done by shaping the phase of the controller
transfer function with a phase lead compensation. The compensation angle needed to
move mode i to the left (parallel with the real axis) is
i D 180
Ri :
(2)
The phase lead required in (2) is implemented by a controller with the transfer
function
Tw s
1 C T1 s 1 C T3 s
K.s/ D K
:
:
;
(3)
Tw s C 1 1 C T2 s 1 C T4 s
which consists of two lead-lag blocks and a high-pass (wash-out) lter to prevent the
controller from reacting to changes in steady states. The phase characteristic of the openloop transfer function can be derived from the linearized model of the power system. For
PSS designs, it can also be measured with appropriate eld test procedures [16, 17]. This
explains the popularity of the phase compensation technique among the many approaches
to designing power system damping controllers.
The phase compensation technique can also be used to design FACTS damping
controllers. However, FACTS damping controllers are often less robust than PSSs [3].
The main reason for this is that a PSS acts through a generator with generally known
parameters, whereas a FACTS controller is closer to the transmission system, which
can change considerably with load ow conditions. Therefore, many robust control
approaches have been studied to enhance FACTS controller performance.
2.2.
For the study of electromechanical oscillations, the generator stator and network transient
can be safely neglected. As a result, the power system dynamics can be represented by
a system of differential-algebraic equations [15]. The differential equation represents
the dynamics of generator rotors, excitation, and speed-regulator systems. The very fast
network transient can be represented by algebraic equations. Likewise, in the FACTS
control model used for electromechanical oscillation study, it is not essential to model
in detail the electronics circuit dynamics, such as the gate pulse units. The simplied
voltage control scheme for the SVC is depicted in Figure 1 [18].
In Figure 1, the main control loop is a voltage control loop, with local input
signal VSV C and input lter Hm . The voltage controller is represented by a gain KR
and a time constant TR . The dynamics of electronic circuits and gate pulse units are
represented by a time constant Tb and a small delay Td . For the purpose of damping
electromechanical oscillations, a supplementary control loop is added to the main control
loop. To effectively improve the oscillation damping, the input signal u should be chosen
so that it predominantly contains the electromechanical modes of concern. The control
structure of Figure 1 is indeed similar for various type of FACTS controllers [14].
235
257
2.3.
The general framework for many robust control studies is based on an LFT, which
is depicted in Figure 2, which conveniently represents how the uncertainty affects the
input/output relationship of the system under study.
In Figure 2, M is a complex transfer matrix, partitionned as
"
#
M11 M12
M D
2 C.p1 Cp2 / .q1 Cq2 / ;
(4)
M21 M22
and 2 Cq1
p1
M11 / 1 M12 :
(5)
The LFT structure has a useful intepretation: M22 represents the nominal, unperturbed
system and is perturbed by . The transfer matrices M12 , M21 , and M11 reect a prior
knowledge as to how the perturbation affects the nominal tranfer matrix.
236
258
H. Nguyen-Duc et al.
The transfer matrix represents all sources of uncertainties, which could be real
parametric uncertainties or neglected dynamics. The structured uncertainty thus may
contain blocks of repeated real scalars or full complex blocks [19]. Hence, it can be
assumed that takes the form of
where i .s/ 2 R;
j 2 Cmj
mj
; (6)
which means there are rs repeated real scalar blocks and F full complex blocks.
Given the interconnection system, as depicted in Figure 2, the SSV (or ) is dened
as the smallest structured uncertainty , measured in terms of its maximum singular
value ./, which makes det.I M/ D 0:
.M /
WD minf ./ W 2 ; det.I
M/ D 0g:
(7)
.M / D 0.
Power systems experience constant changes in load level and generation level at different
buses. Besides, the system conguration can also change during operation. These changes
can be represented as parametric uncertainties and should be considered in power system
robustness analysis. One method that has been used successfully to treat these uncertainties as real perturbation blocks is based on the linear state space uncertainty model. A
linear state space uncertainty model is described as follows:
2
3
m
m
X
X
i Ai B0 C
i B i 7 "
"
# 6A0 C
#
6
7 x.t/
x.t/
P
i D1
i D1
6
7
D6
7
m
m
6
7 u.t/
y.t/
4C C X C D C X C 5
0
i i
0
i i
"
D
i D1
A0
C0
"
i D1
X
Ai
B0
C
i
D0
Ci
i D1
m
Bi
Di
#! "
#
x.t/
:
u.t/
(8)
237
259
Figure 3. Robust stability analysis framework with time delay and real parameter uncertainty.
structure from this approach is normally of very high dimension, even when the order
reduction method has been applied [8, 21]. Therefore, it is not suitable for robust control
synthesis, since the computation burden would be very heavy (e.g.., with D-K iteration
methods). In this work, this framework is used only to evaluate the robust stability of the
closed-loop system. For control synthesis, a simpler uncertainty model is adopted.
2.5.
Ts
Ts C 1
:
Ts C 1
(9)
Thus, a varying time delay can be represented by a transfer function as in Eq. (9), with
the term T being perturbed by a real parameter , with 2 1I 1, as follows:
T D
Tmin C Tmax
Tmax Tmin
C
:
2
2
(10)
This approach has been used in some previous studies [22, 23]. The problem with this
approach is that the transfer function Eq. (9) has a pole at 1=T , which crosses the
imaginary axis as reaches 1. This will be interpreted by the Analysis and Synthesis
Tool Box [24] as a stability border. The problem can be avoided by using a higher value
for Tmin , but it would, in turn, create some conservativeness in the robustness evaluation.
In this work, a slightly different representation of the varying time delay is proposed
for use, as follows:
T D Tmin C .Tmax
Tmin / 2 ;
(11)
238
260
H. Nguyen-Duc et al.
so that T is always positive for every . It is quite easy to derive the LFT model of varying
time delay in Eq. (11). The varying time delay block is represented by the structure in
Figure 4.
In Figure 4, the LFT representation of 1=T is as follows:
2
3 2
3
3 2
z1
b=a 1
b=a 1
!1
6
7 6
7
7 6
1
1 5 : 4 !2 5 ;
(12)
4 z2 5 D 4 0
y1=T
and T D
1=a
; a D Tmin ; b D Tmax
1=a
x1=T
Tmin .
3. A Case Study
In this work, an SVC controller is designed for the Kundur power system [15]. The system
consists of four generators, located in two areas, as depicted in Figure 5. This small system
retains many properties of large power system dynamics regarding electromechanical
oscillations. It exhibits one weakly damped inter-area mode at 3.81 rad/s.
The control design purpose is to improve the stability of this inter-area mode by
using additional damping control for an SVC at bus 8. This bus is situated at the middle
of a long interconnection line, where an SVC would be needed for better control of
voltage prole. If an additional damping controller can be successfully designed here,
239
261
bus 8 would be a very good place to install an SVC, considering both voltage support
and damping enhancement.
The active loads at two important busesP7 and P9 vary from 867 MW to
1267 MW and from 1267 MW to 1567 MW, respectively. The values of P7 and P9
are combined to obtain 25 different operating conditions. The pole-zero map of this
system in various operating conditions is plotted in Figure 6.
Figure 6 reveals that for damping enhancement of the inter-area mode, the chosen
control loop is, in fact, a less than ideal choice. The inter-area mode is quite close to zero,
which would limit maximum achievable damping effect. In addition, there is another zero
that moves to the right half-plane (RHP) in some operating conditions. This could have
a destabilizing effect to the nearby generator local modes. There are two exciter modes
at 16 rad/s and 18 rad/s, which appear to be well damped.
A conventional SVC controller at bus 8 is designed based on the phase compensation
technique. For the chosen control loop, the open-loop system residue angle varies from
72 to 92 throughout the operating modes. Figure 7 presents the root locus analysis of
the conventional SVC controller for the nominal system in two scenarios: without a time
delay and with a delay of 200 ms. For the root locus analysis, time delay is approximated
by a second-order Pad approximation.
It can be observed that the time delay has a similar effect as an RHP zero. Without
a time delay, the exciter mode at 18 rad/s is affected but not destabilized by feedback
control. However, when there is a delay, the RHP zeros make this mode move toward the
RHP as control gain increases. Adverse interactions have been observed between the SVC
and other power system controllers [14], but it is interesting to nd that these interactions
can be worsened by time delay. The increase of high-frequency mode activity in the
240
H. Nguyen-Duc et al.
Figure 7. Root locus of the SVC damping controller: (a) without delay and (b) with 200-ms delay.
Kundur system, due to the delay, was also observed but not explained in [22]. This result
shows that model order reduction for robust control synthesis of wide-area controllers
should be done more judiciously; one may assume it is safe to ignore the 18 rad/s mode
in Figure 6, while, in fact, it is not if time delay is considered.
Damping performance of the inter-area mode is also affected. Without delay, a
damping factor of 0.3 can be achieved, whereas with a 200-ms delay, the best attainable damping factor is 0.1. The above analysis shows that bus 8 is not a good place for
an SVC damping controller. The damping performance is limited due to the open-loop
system zero location and high-frequency interaction. To improve the system stability with
an SVC at this bus, a robust control approach is needed.
262
241
263
(13)
Wu D 0:06
sC4
;
9s C 1
(14)
W D 0:03
10s C 1
;
s C 10
(15)
Wp D 3
sC3
;
3s C 1
controller
(16)
242
H. Nguyen-Duc et al.
264
243
265
244
266
Simulation Results
To conrm the above analyses, non-linear simulations are carried out for several operating
modes, with a 100-ms three-phase fault at bus 8. The two most representative results are
shown in Figures 12 and 13. The power rating of the SVC is 100 MW. In operating
mode I, where the active power transfer level is 370 MW, the controller performance is
much better than the conventional controller. It can also be observed that the conventional
controller performance has reached its limits, as the high-frequency mode starts to appear
at a 200-ms delay.
Operating mode II represents a worst case scenario, where the open-loop system
becomes unstable, as the power transfer level is high. In mode II, the performance of
4.3.
H. Nguyen-Duc et al.
(a)
(b)
Figure 12. System responses at operating mode I: (a)
245
267
(a)
(b)
Figure 13. System responses at operating mode II: (a)
both controllers has degraded. The controller, however, can still stabilize the system,
even with a time delay of 200 ms, while the conventional controller failed.
246
268
H. Nguyen-Duc et al.
This work was supported by the Hydro-Qubec Transnergie Chair on Simulation and
Control of Power Systems.
References
1. Larsen, E., and Swann, D., Applying power system stabilizerpart I, II, III, IEEE Trans.
Power Apparatus Syst., Vol. 100, No. 6, pp. 30173041, 1981.
2. Larsen, E., Sanchez-Gasca, J., and Chow, J., Concepts for design of FACTS controllers to
damp power swings, IEEE Trans. Power Syst., Vol. 10, No. 2, pp. 948956, May 1995.
3. Rogers, G., Power System Oscillations, Norwell, MA: Kluwer Academic Publishers, 2000.
4. Zhao, Q., and Jiang, J., Robust SVC controller design for improving power system damping,
IEEE Trans. Power Syst., Vol. 10, No. 4, pp. 19271932, November 1995.
5. Klein, M., Le, L., Rogers, G., Farrokhpay, S., and Balu, N., H1 damping controller design in
large power systems, IEEE Trans. Power Syst., Vol. 10, No. 1, pp. 158166, February 1995.
6. Pal, B. C., Coonick, A., Jaimoukha, I., and El-zobaidi, H., A linear matrix inequality approach
to robust damping control design in power systems with superconducting magnetic energy
storage device, IEEE Trans. Power Syst., Vol. 15, pp. 356362, 2000.
7. Zhu, C., Robustness Analysis for Power System Based on the Structured Singular Value and
the Gap Metric, Ph.D. Dissertation, Iowa State University, Ames, IA, 2001.
8. Djukanovic, M., Khammash, M., and Vittal, V., Application of the structured singular value
theory for robust stability and control analysis in multimachine power systems, I. Framework
development, IEEE Trans. Power Syst., Vol. 13, No. 4, pp. 13111316, 1998.
9. Ramanathan, B., and Vittal, V., Small-disturbance angle stability enhancement through direct
load control part Iframework development, IEEE Trans. Power Syst., Vol. 12, No. 2,
pp. 773781, 2006.
10. Ramanathan, B., and Vittal, V., Small-disturbance angle stability enhancement through direct
load control part IInumerical simulations and results, IEEE Trans. Power Syst., Vol. 12,
No. 2, pp. 782790, 2006.
11. Dotta, D., Silva, A. S., and Decker, I. C., Wide-area measurement-based two-level control
design considering signal transmission delay, IEEE Trans. Power Syst., Vol. 24, pp. 208216,
2009.
12. Chaudhuri, B., Majumder, R., and Pal, B. C., Wide-area measurement based stabilizing control
of power system considering signal transmission delay, IEEE Trans. Power Syst., Vol. 19,
No. 4, pp. 19711979, 2004.
13. Wu, H., Tsakalis, K. S., and Heydt, G. T., Evaluation of time delay effects to wide-area power
system stabilizer design, IEEE Trans. Power Syst., Vol. 19, pp. 19351941, 2004.
14. Mathur, R. M., and Varma, R. K., Thyristor-based FACTS Controllers for Electrical Transmission systems, EPRI Power Systems Engineering Series, Piscataway, NJ: Wiley-IEEE Press,
2002.
15. Kundur, P., Power System Stability and Control, EPRI Power Systems Engineering Series,
New York: McGraw Hill Professional, 1994.
16. Murdoch, A., Venkataraman, S., and Lawson, W. P. R. A., Integral of accelerating power
type PSS, part 1theory, design, and tuning methodology, IEEE Trans. Energy Conversion,
Vol. 14, No. 4, pp. 16581663, 1999.
Acknowledgment
247
269
17. Murdoch, A., Venkataraman, S., Lawson, R., and Pearson, W., Integral of accelerating power
type PSS, part 2eld testing and performance verication, IEEE Trans. Energy Conversion,
Vol. 14, No. 4, pp. 16641672, 1999.
18. Pai, M. A., Sen-Gupta, D. P., and Padiyar, K., Small Signal Analysis of Power Systems, Harrow,
Middlesex, UK: Alpha Science International, 2005.
19. Skogestad, S., and Postlethwaite, I., Multivariable Feedback ControlAnalysis and Design,
Bafn Lane, Chichester, West Sussex, England: John Wiley & Sons Ltd, 1996.
20. Castellanos, R., Messina, A., and Sarmiento, H., Robust stability analysis of large power
systems using the structured singular value theory, Intl. J. Elect. Power Energy Syst., Vol. 27,
pp. 389397, 2005.
21. Nguyen-Duc, H., Dessaint, L., and Okou, A., Power system robust stability analysis using
structured singular value theory and model reduction method, Proceedings of the IEEE PES
General Meeting, Calgary, AB, Canada, 2529 July 2009.
22. Wu, H., Robust Control Design Considering Time Delay for Wide Area Power Systems, Ph.D.
Dissertation, Arizona State University, Tempe, AZ, 2004.
23. Zhang, Y., Design of Wide-area Damping Control Systems for Power System Low-frequency
Inter-area Oscillations, Ph.D. Dissertation, Washington State University, Pullman, WA, 2007.
24. The MathWorks, -Analysis and Synthesis Toolbox, Natick, MA: Author, 2001.
25. Zhou, K., Doyle, J., and Glover, K., Robust and Optimal Control, Englewood Cliffs, NJ:
Prentice Hall, 1996.
Suppose a linear uncertain state space model of a power system has been obtained, as
in
i The system in Eq. (8) has n hstates, niy outputs, and nu inputs. Therefore,
h Eq. (8).
Bi
Ai
Bi
i
.nCny / .nCnu /
. Let ri WD rank A
Ci
Di . Each matrix will be factored, using
Ci
Di 2 <
singular values decomposition, as
"
# " #
Ai Bi
Ei
D
Gi Hi ;
(A1)
Fi
Ci Di
where
"
Ei
Fi
#
2 <.nCnu /
ri
Gi
Hi 2 <ri
.nCny /
The singular values decomposition helps to eliminate the weak inuences of parametric
uncertainties to the system states. The system in Eq. (8) can then be represented by
2
32 3
3 2
x
A0 B0 E1
x.t/
P
Em
6
76 7
7 6
Fm 7 6 u 7
6 y 7 6 C0 D0 F1
6
76 7
7 6
6 7
6 z1 7 6 G1 H1 0
0 7
(A2)
6
7 6 w1 7 ;
7D6
6 : 7 6 :
7
:: 7 6
::
::
:: 7
::
6
6 : 7 6 :
7
:
: 54 : 5
:
:
4 : 5 4 :
Gm Hm 0
zm
wm
0
where the perturbation has a structure as follows:
D fdiag1 Ir1 ; : : : ; m Irm W i 2 <g:
(A3)
248
270
H. Nguyen-Duc et al.
-controller:
10s
10s C 1
Gc .s/ D 0:076
10s
10s C 1
s C 2:11
s C5
s C 2:11
:
s C 0:1
(B2)
249
I. INTRODUCTION
978-1-4577-1002-5/11/$26.00
2011
B mn H thng in - i hc Bch
KhoaIEEE
H Ni
250
Huy Nguyen-Duc, Member, IEEE, Amel Zerigui, Louis-A. Dessaint, Senior Member, IEEE, Xiaoping
Tu, Member, IEEE, and Camilo Apraez
Vmin d V d Vmax
(3)
G (t ) M C1 M k G k (t ) M C1 M jG j (t )
kC
(5)
Qg min d Qg d Qg max
(6)
kC
dZi
dt
1
( Pgi Pei )
Mi
; MC
jC
Z (t ) M C1 M k Zk (t ) M C1 M jZ j (t )
kC
(13)
jC
Zi
Pe (t )
dG i
dt
generators; M C
(4)
Pg min d Pg d Pg max
(12)
jC
(7)
Pm (t )
where :
(8)
where:
MCMC
(16)
MC MC
Pei
'T t 1
(10)
(Zi Zit 2) 0
2
'T
Zit 1 Zit
(2 Pgi Peit 1 Peit ) 0 (11)
2M i
where 'T is the integration time step. The constraints
G it 1 G it
Elec. power
Mech. power
120
100
B
80
60
Power
140
40
20
0
-20
-40
20
40
60
80
100
Delta
120
140
160
180
QG QL Q(V ,T )
Voltage
Angle
Pg
Qg
1.06
-0.000
158.2
22.2
1.05
-3.19
87.9
1.3
1.04
-3.44
71.5
-17.7
1.05
-4.70
1.02
-8.77
1.03
-7.85
1.05
-6.05
1.04
-8.05
1.05
-5.64
G (tu cct )
CCT: 512ms
1.1
Gen1
1.08
Gen2
Gen3
1.06
Speeds
1.04
1.02
0.98
0.2
0.4
0.6
0.8
1
Time (sec)
1.2
1.4
1.6
1.8
252
TABLE II
OPERATING CONDITION AFTER FIRST TSC-OPF ITERATION
Bus
Voltage
Angle
Pg
Qg
1
2
1.06
0.00
157.51
23.17
1.05
-4.08
73.22
-0.04
86.86
-16.85
1.04
-2.37
1.05
-4.68
1.02
-9.90
1.04
-7.62
1.05
-7.47
1.04
-8.04
1.05
-5.04
Bus
Voltage
Angle
1.02
-14.67
350
-25.8
1.06
1145.52
501.31
1.04
-10.48
458.33
263.11
1.00
-14.12
468.79
81.59
1.01
-14.18
504.94
137.08
1.06
-8.91
749.99
206.99
1.06
-10.13
375.9
33.69
1.03
-12.63
435.93
20.36
1.03
-14.16
588.63
-31.95
1.05
-18.44
1100
45.46
CCT: 576ms
Qg
Gen1
1.07
Gen2
1.06
Gen3
10
1.05
1.04
Speeds
Pg
CCT: 208ms
1.03
1.01
1
0.99
0.98
0.2
0.4
0.6
0.8
1
Time (sec)
1.2
1.4
1.6
1.8
10
150
Elec. power
Mech. power
Power
100
B
Power
Elec. power
0
Mech. power
-2
-20
20
40
60
80
100
120
140
50
160
Delta
Figure 4 P-G curve of the 3 machine system after the initial OPF.
-50
20
40
60
80
100
Delta
120
140
160
180
200
Figure 5 P-G curve of New England system after the initial OPF
1.02
Bus
Voltage
Angle
Pg
Qg
1.02
-14.73
350
-27.39
1.06
1145.55
501.68
1.04
-10.33
469.11
267.05
1.00
-14
476.71
83.97
1.01
-14.04
509.98
137.69
1.06
-8.89
750
210.36
1.06
-9.9
389.07
35.15
1.03
-11.99
488.62
22.51
1.03
-17.23
499.37
-36
10
1.05
-18.48
1100
48.97
CCT: 245ms
Power
60
40
20
REFERENCES
[1]
-20
Elec. power
-40
-60
Mech. power
20
40
60
80
100
120
140
Delta
Figure 6 P-G curve of New England system after first TSC-OPF iteration
V. CONCLUSION
The problem of Transient Stability Constrained Optimal
Power Flow is studied in this work. The TSC-OPF algorithm
is based on the framework in [6], with a slight modification of
the transient stability constraint. Besides, the loss
minimization is considered as the objective function. The
TSC-OPF framework can successfully tackle both cost
minimization and loss minimization problems, with good
convergence property.
It is found that a slight modification of transient stability
constraint can give a better convergence property. In both test
B mn H thng in - i hc Bch Khoa H Ni
TABLE IV
OPERATING CONDITION FOR NEW ENGLAND SYSTEM, FIRST TSC-OPF
Camilo Apraez received his B.S. degree from de Los Andes university in
Colombia. Currently he is a master student at cole de Technologie
suprieure, Quebec, Canada.
255
AbstractThis paper studies the various aspects of input/output selection for wide-area damping controllers, targeted
at the power system inter-area oscillation modes. The objectives is
to choose a control loop which has large impact to a few selected
modes, while does not affect the other system modes, hence the
interactions problem can be greatly reduced. For this purpose, a
new modal interaction index is proposed. Non linear simulations
are also carried out to validate the modal analysis results.
Index TermsPower system control, inter-area oscillations,
Hankel Singular Value.
I. I NTRODUCTION
LECTROMECHANICAL oscillation is a common problem in large electric power sytems. These oscillations
limit the power transfer capacity, and at times, may cause
severe grid failures [1]. Damping of these oscillations is therefore important for a secure operation of power systems. The
conventional damping device is the Power system Stabilizer
[2]. Emerging electronic based devices (FACTS) can also be
equipped with auxiliary control mode to provide oscillation
damping [3][4].
With the advent of synchronized phasor measurement units,
power system controllers can now utilize input signals from
many different locations of the grid. This provides a much
greater possibility for choosing control loops. A wide-area
control loop is inherently more effective than a local loop
to damp inter-area modes of oscillations. Many works have
been directed towards designing wide-area controllers [5][6].
Throughout many studies in the litterature, a common approach is to use wide area control loop as an additional
loop that works in parallel with the local loop [7][8]. This
approach allows to improve damping of inter-area modes,
while still maintains local modes damping performance in case
of communication failures.
One important problem in designing the wide-are control
loop is the selection of input/output signals. Ideally, a widearea loop should have good controllability/observability to
the inter-area modes of concern, while does not affect other
system modes. Although there are numerous works on the
(1)
256
Huy Nguyen-Duc, Student Member, IEEE, Louis-A Dessaint, Senior Member, IEEE,
Aim Francis Okou, Member, IEEE and Innocent Kamwa, Fellow, IEEE
moi = cos((C, ei )) =
|B T .fi |
||fi ||.||B||
|ei .C|
||ei ||.||C||
C. Measures of interactions
(2)
y(s) = G(s).u(s)
(8)
(3)
For control loop selection, one common measure for interactions is the Relative Gain Array (RGA). Given a power
system with m input and n output, denoted by:
n
i=1
Ri
s i
(4)
ij =
(5)
(6)
uk =0,k=j
1
= gij (s).gij
(s)
(9)
yk =0,k=j
(10)
Large joint controllability/observability measure are essential for an effective control loop. In some works [17], it is
the single modal index being used to evaluate control loops.
However, a loop with weaker joint controllability/observability
measure may be better, in terms of interaction measure, to
loops having large joint controllability/observability. Therefore, the joint controllability/observability measure should be
used in conjunction with other criteria, and not as a preliminary selection step.
B. Interaction measure
As presented in section II, interactions measures in the
litterature are mostly "loop based", i.e these measures are
used to determine control interactions between different control loops. A new interaction measure which determines the
relative contribution of the critical inter-area mode, compared
to the total contribution of all system modes to the loops, is
proposed in this work. The proposed interaction measure is
determined by the following steps:
Given a candidate loop, the state matrix A, B, C, D
of the open loop system are determined. The balanced
realization of this system, denoted by state matrices
Ab , Bb , Cb , Db is determined. In this realization, the
rst state has the most contribution to the input/output
relation, followed by the second state, and so on. The
HSV of the balanced system {i : i = 1, .., N } are also
obtained.
Suppose the wide-area loop is designed to improve damping of an inter-area mode k . The contribution of this
mode to the input/output relation ship of the candidate
loop can be determined based on the participation factor
of each balanced state to mode k [18]. Denote pi (k )
the participation factor of the state variable i (in the
balanced realization) to mode k . The contribution of
each state i to the input/output relationship is weighted
by the corresponding HSV, i.e. i . Therefore the total
contribution of mode k to the input/output relationship
of the candidate loop is:
k r
Ck =
Ik =
Ck
j
Cj
N
i=1
j
i2 .pi (k )
N
i=1
i2 .pi (j )
(12)
(13)
i2 .pi (k )
1 sin( mk )
1 + sin( mk )
1
T1 =
k
T2 = T1
(11)
N
i=1
i=1
i2 .pi (k )
i2 .pi (k )
N
i=1
Ikm =
(15)
1.003
1.002
1.001
Speed (pu)
1
0.999
0.998
0.997
Fig. 2.
15
10
20
30
25
Fig. 1.
0.5
0
8
9
10
b) Interaction index
11
13
16
9
8
10
c) Phase mismatch
11
13
16
11
13
16
Table I
E LECTROMECHANICAL MODES OF N EW E NGLAND SYSTEM
Modes (rad/s)
0.035
2.55
3.2
3.95
4.97
6.06
6.59
7.36
7.87
7.88
8.31
9.44
9.55
9.55
11.34
Participating generators
1-16
1-16
14,16
1-9,12,13
14,15,16
4,5,6,9
2,3,4,5,6,7
4,5,6,7
1,2,3
1,2,3,8,9,10,12
1,8,10
4,5,6,7,8
4,5,6,7
1,8
10,11,12
-0.559+9.55i
-0.45+7.36i
-0.172+3.95i
-0.0653+2.55i
1
0.5
0
200
100
0
10
Input signal
Fig. 3.
A. SVC at bus 17
Mode
Mode
Mode
Mode
Mode
Mode
-0.559+9.55i
-0.277+6.6i
-0.287+6.06i
-0.45+7.36i
-0.172+3.95i
-0.0653+2.55i
0.5
10
11
13
16
11
13
16
11
13
16
b) Interaction index
1
0.5
10
c) Phase mismatch
200
150
1
5
11
16
100
50
Speed (pu)
10
Input signal
Fig. 5.
0
10
15
20
b) SV C17 , with 16 as input signal
25
30
1
5
11
16
1.002
Speed (pu)
Speed (pu)
1.002
1.0015
1.001
1.0005
1
0.9995
0.999
0.9985
0.998
10
15
Time (sec)
20
25
1.001
1
0.999
0.998
30
10
6
8
12
14
b) SV C21 , with 1 as input signal
16
18
20
16
18
20
Speed (pu)
1.002
Fig. 4.
1.001
1
0.999
0.998
Fig. 6.
10
12
Time (sec)
14
B. SVC at bus 21
The SVC 21 is electrically not very far from bus 17, but this
is enough to make a difference. With this SVC location many
loops now have large joint measures for the 7.36 rad/s mode
(Fig. 5). For this SVC location, 16 is still the best input, in
terms of interaction measure. Speed signals from generator 1,
5, 7, 8 have the largest joint measures to the 2.55 rad/s mode,
but signal 5 should not be used because of its large phase
mismatch with mode 7.36 rad/s.
The above analysis is tested by non linear simulations.
Fig 6a shows some speed signals when 5 is used as input.
Speed responses with 1 as input signal are shown in Fig.
6b. The results clearly show that using 5 as input signal will
destabilize the 7.36 rad/s mode. This mode can be easily seen
in Fig. 6a, and Prony analysis is also carried out to verify its
frequency. The control loop with input 1 does not interact
with this mode, therefore 1 is a better input signal to pair
with SV C21 .
B mn H thng in - i hc Bch Khoa H Ni
1.002
1.0015
1.001
1.0005
1
0.9995
0.999
0.9985
0.998
Mode
Mode
Mode
Mode
Mode
Mode
0.2
-0.422+11.3i
-0.277+6.6i
-0.287+6.06i
-0.45+7.36i
-0.172+3.95i
-0.0653+2.55i
0.1
0
8
9
10
b) Interaction index
11
13
16
11
13
16
0.5
10
c) Phase mismatch
200
10
11
13
16
Input signal
Fig. 7.
1
5
11
16
Speed (pu)
1.002
1.001
1
0.999
0.998
20
10
15
b) T CSC147 , with 8 as input signal
25
30
25
30
Speed (pu)
1.002
1.001
1
0.999
0.998
Fig. 8.
10
15
Time (sec)
20
R EFERENCES
100
Aim Francis Okou (M04) received the Dipl.Ing. degree in electrical engineering from cole Suprieure Interafricaine de llectricit, Cte dIvoire,
in 1993, the M.Eng. degree and the Ph.D. in electrical engineering from
cole de Technologie Suprieure (TS), Montreal, QC, Canada, in 1996 and
2001, respectively. During 2002-2005, he was a Postdoctoral Researcher in
the Department of Electrical Engineering at TS. Since 2005, he joined The
Royal Military College of Canada where he is currently an Assistant Professor
in the Electrical and Computer Engineering Department. His research interests
include the application of robust and nonlinear control techniques to largescale systems
262
I. INTRODUCTION
263
Huy Nguyen-Duc, Student Member, IEEE, Louis-A Dessaint, Senior Member, IEEE,
Aim Francis Okou, Member, IEEE, and Innocent Kamwa, Fellow, IEEE
1919
II. BACKGROUND
A. Transfer Function Residue and Location of Zeros
To effectively control an oscillation mode using feedback, the
chosen input and output signal must have, respectively, large
controllability and observability of the mode. These two measures can be combined to give the transfer function residue associated with the mode of oscillation [19]. Let
be the residue
of a control loop with respect to a mode
, a sufciently small proportional feedback control causes a change
in , as follows:
many robust control design approaches, the analysis using structured singular value theory [21] is an effective method to evaluate the robustness of power system controllers.
The general framework for many robust control studies is
based on a linear fractional transformation (LFT), depicted in
Fig. 1. It represents how the uncertainty affects the input/output
relationship of the system under study.
In Fig. 1,
is a complex transfer matrix, partitioned as follows:
(2)
respect to
(1)
(3)
represents
The LFT structure has a useful interpretation:
the nominal, unperturbed system, and is perturbed by . The
transfer matrices
,
, and
reect a prior knowledge
as to how the perturbation affects the nominal transfer matrix.
The transfer matrix represents all sources of uncertainties,
which could be real parametric uncertainties or neglected dynamics. The structured uncertainty thus may contain blocks
of repeated real scalars, or full complex blocks [21]. Hence, we
can assume that takes the following form:
(4)
Given the interconnection system, as depicted in Fig. 1, the
structured singular value (s.s.v. or ) is dened as the smallest
structured uncertainty , measured in terms of its maximum
singular value
which makes
:
(5)
B. Robustness Evaluation Using -Analysis
Conventional power system control designs are based on linearized models of the system about some selected operating
points. As a consequence, the performance of the controller may
not be robust over a wide range of operating conditions. In recent
years, several researches have studied the application of robust
controlBtechniques
to design power system controllers. Among
mn H thng in - i hc Bch Khoa H Ni
.
If no such structure exists, then
Since
is the size of the smallest structured perturbation
that makes the system
singular, it can be used as a measure
of robust stability. For power system application, -analysis can
be used to evaluate the system robustness with uncertainty in
operating conditions. As the operating condition changes, the
state space matrices of the linearized power system also264
change.
The dependency of the state matrix coefcients on the parameter variations can be captured to formulate a structured perturbation model. The linear state space model of a power system,
subjected to parameter changes
, can be written
as follows:
(6)
where
,
, , and
are the state space matrices of the
nominal system. Matrices , , , and
are determined by
solving a set of overdetermined linear equations. In robust stability studies, the perturbations are normalized so that
. The parameter uncertainty representation in (6) can be transformed to an LFT form, suitable for robust analysis. More details on this framework are reported in [22] and [23].
With this framework, changes in system operating conditions
are treated as parameter uncertainties. Besides, time delay can
be considered as a varying quantity, rather than a xed value,
which is used in many studies [24], [25].
C. Control Using Bang-Bang Modulation
For a simple power system model, the concept of FACTS control to improve power swing damping can be derived, without
using complicated modal analysis tools. Consider the single machineinnite bus system, depicted in Fig. 2, where an SVC is
placed at the midpoint of the interconnection line. For this power
system, the system damping can be enhanced if the midpoint
voltage is modulated as a function of
[8], as follows:
(7)
where is a constant. An intuitive explanation of (7) is that it
reduces the kinetic energy built up in the rotor during transient
[8]. To improve damping of an oscillation , (7) suggests that
the modulated SVC voltage should lead the observed oscillation in by 90 degrees at frequency . Alternative signals for
can also be used, e.g.,
or machine speed [2], [3].
Control of
is done by reactive power modulation.
To achieve maximum damping, a bang-bang control method
can be used, in which the SVC reactance is switched to its
maximum capacitive/inductive value in synchronization with
the rate of change of . In our previous work [18], we have successfully implemented a bang-bang control scheme based on the
control rule (7), where the switching time is determined based
on the identied frequency and phase of critical power oscillations. B
The
proposed control scheme in [18] can be summarized
mn H thng in - i hc Bch Khoa H Ni
(10)
(9)
1921
use a xed value for , with given parameter and time delay
uncertainties. On the other hand, the damping performance of
the hypothetic xed-phase controller, as in Fig. 4, does not represent the performance of the proposed controller which uses
bang-bang control action. The latter should have better damping
effect, since maximum power rating of FACTS is used at each
switching instance.
1923
Fig. 10. System response with SVC and PSS. (a) Tie line power, PSS or SVC
alone. (b) Tie line power, with PSS and SVC.
Fig. 8. System response to a three-phase fault at bus 8.
swings, the PSS has better damping effect. However when the
bang-bang mode is activated, the proposed controller catches up
with the PSS and nally has better settling time. On the other
hand, the two controllers can be used simultaneously to give a
much better response [see Fig. 10(b)].
The proposed control scheme also works well in the case of
reversed power ow. Simulation results for this case were presented in [18].
B. New England System
The proposed control scheme is now applied to the New England power system, depicted in Fig. 11. Complete system description and data can be found in [6]. This system has a weakly
damped mode at 2.39 rad/s, in which generators 113 oscillate
against generators 1416. The proposed control scheme is thus
designed for this mode. All other electromechanical modes of
this system are presented in Table I.
Mode shape analysis reveals that generators 113 are virtually in phase for the 2.39 rad/s mode. For the proposed control
scheme, this means any speed signal
can be268
used as
TABLE I
ELECTROMECHANICAL MODES OF NEW ENGLAND SYSTEM
TABLE II
PARAMETER VARIATION RANGES
for the open loop system, and for the closed loop systems with
several different control loops. The open loop only shows two
peaks at 2.39 rad/s and 3.9 rad/s. All closed loop systems have
a lower peak at 2.39 rad/s, since the controller is designed to
damp this mode. However, the controllers have excited various
modes at different frequencies. It can be seen that only the loop
(SVC at bus 21, with as input signal) is free from
interaction problems. Using different input signals for the SVC
results in exciting the inter-area modes at 3.9 and 7.3 rad/s. The
TCSC on line 147 tends to excite the inter-area modes at 11.34
rad/s and 8.34 rad/s. It should be noted that the large -peaks in
Fig. 12 are largely due to the time delay. The peaks for closed
loop systems without time delay (not shown here for the sake of
brevity) are less pronounced than those in Fig. 12.
This result suggests that with a continuous control scheme,
it is not possible to change the input signal without having to
redesign the controller: Control interactions occur at various
frequencies, depending on the input signal. On the contrary, if
the proposed control scheme is used, the identication based
bang-bang control action will maximize the damping effect.
When the oscillation has been reduced considerably, a continuous controller with low feedback gain will be used. This kind
of gain scheduling control allows to achieve better damping
performance while avoiding interaction issues.
To validate the above analysis, closed loop performances of
these control loops are analyzed with nonlinear simulations,
with time delay varying from 0300 ms. High feedback gains
are used with the continuous controllers, in order to match the
proposed control scheme damping performance.
Fig. 13 presents damping performances of three control
loops:
in Fig. 13(a) and (b),
in
Fig. 13(e) and (f), and
in Fig. 13(c) and (d),
when a three-phase fault at bus 42 is applied at
for
200 ms. With each control loop, damping performances of
the continuous controller and the proposed control scheme are
compared. The continuous controllers parameters are given in
Appendix B.
It can be seen that with the loop
, the continuous control scheme achieves a fairly similar damping performance to the proposed control scheme. In fact, in this
large
269
Fig. 12. Robust stability for various control loops, New England system.
Fig. 13. Generator speed responses with various control loops. (a) SV C
! , continuous control. b) SV C ! , proposed control scheme. (c) T CSC
, continuous control. (d) T CSC
! , proposed control scheme. (e) SV C
! , continuous control. (f) SV C
! , proposed control scheme.
power system, a single controller could not have as large controllability as in the Kundur system. Therefore, the difference in
damping performance between two controllers is smaller. For
all other control loops, however, the interaction problems predicted by the -analysis can be easily observed. Using continuous feedback control, the TCSC destabilizes the 11.34 rad/s
mode in Fig. 13(c), while the SVC destabilizes the 3.9 rad/s
mode in Fig. 13(e).
On the other hand, the damping performance of the proposed
control scheme is consistent with different control loops in
Fig. 13(b), (d), and (f). Therefore, it is possible to change
the input signal (e.g., in case of a communication failure/bad
measurement), while still maintaining the damping performance. Besides, the damping performance can be further
improved if several controllers are activated simultaneously.
Fig. 14 presents the system response when three controllers are
used simultaneously: the SVC at bus 21, with power rating of
, the TCSC on line 147, and a UPFC on line 127.
For damping purpose, the series voltage for UPFC is modulated
at a magnitude of
. The maximum compensation ratio
for the TCSC on line 147 is
. For the 2.39 rad/s mode,
the required phase leads for the two controllers are 5 and
30 , respectively. Compared to the system response with one
controller, the system damping has been clearly improved.
The performance of the proposed control scheme is also
examined for more severe contingencies. Fig. 15 shows the
system response with three controllers, with the same fault at
bus 42,Band
lines 4748 and 30-9-36 tripped out. Compared to
mn H thng in - i hc Bch Khoa H Ni
1925
It has been shown in [15] that the switchings of a small percentage of system load, if well synchronized with the disturbing
event, could have highly stabilizing effects. The same conclusion is drawn from this work: In all test cases, the FACTS power
ratings required are very small, compared to the system total active power. A large power rating means the scheme can damp
the oscillation faster, but also means that it cannot be used when
there is a small oscillations (e.g., in smaller contingencies). It
would be better to use a small value for FACTS power rating
for this type of control, so that it can be used in many contingencies, without violating physical constraints.
APPENDIX A
MODELING OF VARYING TIME DELAY UNCERTAINTY
damp inter-area oscillations, or ideally the global mode of oscillation (where all machines in the system participate), since the
number of FACTS that can be used is maximized. The proposed
scheme damps one oscillation at a time, but it can be designed
for more than one mode. For each mode, there is an associated
group of FACTS that can be used.
One advantage of the control scheme is the lack of interactions with other system dynamics. As can be seen in all
test cases, the proposed control scheme can achieve very
good damping performance while a conventional controller
using the same input-output signals only achieves moderate
damping, due to interactions with other system modes, which
can be either another electromechanical mode (New England
system), or a exciter mode (Kundur system). A high order
controller, designed using a robust control synthesis method,
might achieve the same performance as the proposed control
scheme. The proposed control scheme, however, presents a
simple alternative to various robust control design approaches.
It requires knowledge of only the open loop residue angle for
the mode of interest. Besides, it has the exibility of switching
easily between different input signals. A controller can also be
added or removed from the control scheme without having to
change any control parameter.
This work reveals that communication delay in the control
loop can have a drastic effect on control interactions. Time delay
has a similar effect as RHP zeros [21], so that when a mode is
stabilized by the control action, another mode will be destabilized. The proposed analysis method in this paper, using structured singular value theory, can reect very well this problem.
More detailed comparative studies between the proposed robust
stability analysis framework and conventional modal analysis
will be presented in a future publication.
One drawback of the proposed approach is that it is not
possible to quantify the improvement in damping, since the
bang-bang control mode is not a continuous feedback controller. However, the hypothetical xed-phase controller
introduced in Section III-B can be used to evaluate the robustness of the proposed control scheme. The xed phase controller
can also provide a lower bound of obtainable performance for
the proposed
control scheme.
B mn H thng in - i hc Bch Khoa H Ni
(12)
where
,
(nominal transfer function). Since
, the relative uncertainty should satisfy
(13)
Since
have
, and
, we
(14)
By varying from
to
and in the range of frequencies of interest, we can obtain the upper bounds for the lefthand side of (14), in function of frequency. Then a weighting
function for
can be derived by approximation, using a least
square curve tting routine.
APPENDIX B
CONTINUOUS CONTROLLER PARAMETERS
The structure of the conventional continuous controller in this
work is as follows:
(15)
Parameters for the New England system controllers:
:
,
,
,
:
,
,
.
;
,
REFERENCES
[1] A. Olwegard, K. Walve, G. Waglund, H. Frank, and S. Tornseng,
Improvement of transmission capacity by thyristor controlled reactive power, IEEE Trans. Power App. Syst., vol. PAS-100, no. 8, pp.
39303939, Aug. 1981.
[2] A. Hammad, Analysis of power system stability enhancement by static
var compensators, IEEE Trans. Power Syst., vol. 1, no. 4, pp. 222227,
Nov. 1986.
271
Fig. 15. Damping performance with three controllers, lines 4748, 3036 out
of service.
The problem of representing a time delay by an input multiplicative uncertainty is to nd an uncertainty weighting function
, so that the time delay can be represented as
Huy Nguyen-Duc (S06) received the B.S. degree in electrical engineering from
Hanoi University of Technology, Hai Ba Trung, Vietnam, in 2001 and the M.S.
degree from Sherbrooke University, Sherbrooke, QC, Canada, in 2003. He is
currently pursuing the Ph.D. degree in the Electrical Engineering Department
of the cole de Technologie suprieure, Montral, QC, Canada.
His areas of research interest are simulation and control of power system dynamics.
Louis-A Dessaint (M88SM91) received the B.Ing., M.Sc.A., and Ph.D. degrees in electrical engineering from the cole Polytechnique de Montral, Montral, QC, Canada, in 1978, 1980, and 1985, respectively.
Currently, he is a Professor of electrical engineering at the cole de Technologie Suprieure, Montral. From 1992 to 2001, he was the Director of the
Groupe de Recherche en lectronique de Puissance et Commande Industrielle
(GREPCI), a research group on power electronics and digital control. Since
2002, he has held the TransEnergie (Hydro-Qubec) Chair on Power Systems
Simulation and Control. He is an author of the SimPowerSystems simulation
software.
Dr. Dessaint received the Outstanding Engineer Award from IEEE-Canada in
1997. He is also an associate editor for the IEEE TRANSACTIONS ON CONTROL
SYSTEMS TECHNOLOGY.
Aim Francis Okou (M04) received the Dipl.Ing. degree in electrical engineering from the cole Suprieure Interafricaine de llectricit, Cte dIvoire,
in 1993, and the M.Eng. degree and the Ph.D. in electrical engineering from the
cole de Technologie Suprieure (TS), Montreal, QC, Canada, in 1996 and
2001, respectively.
Since 2005, he joined The Royal Military College of Canada where he is currently an Assistant Professor in the Electrical and Computer Engineering Department. His research interests include the application of robust and nonlinear
control techniques to large-scale systems.
Innocent Kamwa (S83SM98F05) received the B.Ing. degree and the Ph.D
degree in electrical engineering from Universit de Laval, Qubec city, QC,
Canada, in 1984 and 1988, respectively.
Since then, he has been with the Hydro-Qubec Research Institute, where
at present, he is a Principal Researcher with interests broadly in bulk system
dynamic performance. Since 1990, he has held an Associate Professor position
in electrical engineering at Universit de Laval.
Dr. Kamwa is a recipient of the 1998 and 2003 PES Prize Paper Awards and is
currently serving on the System Dynamic Performance Committee, AdCom. He
is also the acting Standards Coordinator of the PES Electric Machinery Committee and an associate editor for the IEEE TRANSACTIONS ON POWER SYSTEMS.
He is a member of CIGR.
272
[3] E. Lerch, D. Povh, and L. Xu, Advanced SVC control for damping
power system oscillations, IEEE Trans. Power Syst., vol. 6, no. 2, pp.
524535, May 1991.
[4] E. Larsen, J. Sanchez-Gasca, and J. Chow, Concepts for design of
FACTS controllers to damp power swings, IEEE Trans. Power Syst.,
vol. 10, no. 2, pp. 948956, May 1995.
[5] P. Kundur, Power System Stability and Control, ser. EPRI Power Systems Engineering. New York: McGraw-Hill, 1994.
[6] G. Rogers, Power System Oscillations. Norwell, MA: Kluwer, 2000.
[7] S. Ray, B. Chaudhuri, and R. Majumder, Appropriate signal selection
for damping multi-modal oscillations using low order controllers, in
Proc. IEEE PES General Meeting, 2008.
[8] R. M. Mathur and R. K. Varma, Thyristor-Based FACTS Controllers
for Electrical Transmission Systems, ser. EPRI Power Systems Engineering. New York: Wiley, 2002.
[9] A. Heniche and I. Kamwa, Control loops selection to damp inter-area
oscillations of electrical networks, IEEE Trans. Power Syst., vol. 17,
no. 2, pp. 378384, May 2002.
damping
[10] M. Klein, L. Le, G. Rogers, S. Farrokhpay, and N. Balu,
controller design in large power systems, IEEE Trans. Power Syst.,
vol. 10, no. 1, pp. 158166, Feb. 1995.
[11] X. Yu, Robustness analysis and controller design for static var
compensators in power systems, Ph.D. dissertation, Iowa State Univ.,
Ames.
[12] J. F. Hauer, Robustness issues in stability control of large electric
power systems, in Proc. 32nd Conf. Decision and Control, 1994, vol.
3, pp. 23292334.
[13] S. C. Cigre, Task force 17, Advisory Group 02, Advance angle stability
control, 1999, CIGRE Technical Brochure.
[14] O. Samuelsson and M. Akke, On-off control of an active load for
power system dampingTheory and eld test, IEEE Trans. Power
Syst., vol. 14, no. 2, pp. 608613, May 1999.
[15] I. Kamwa, L. Grin-Lajoie, D. Asber, J. Gingras, and G. Trudel,
Large-scale active-load modulation for angle stability improvement,
IEEE Trans. Power Syst., vol. 14, no. 2, pp. 582590, May 1999.
[16] J. Chang and J. Chow, Time-optimal series capacitor control for
damping inter-area modes in interconnected power systems, IEEE
Trans. Power Syst., vol. 12, no. 1, pp. 215221, Feb. 1997.
[17] J. Chang and J. Chow, Time-optimal control of power systems requiring multiple switchings of series capacitors, IEEE Trans. Power
Syst., vol. 13, no. 2, pp. 367373, May 1998.
[18] H. Nguyen-Duc, L. Dessaint, and A. Okou, A power system damping
controller based on identication of critical oscillations, in Proc. IEEE
PES General Meeting, Jun. 2007.
[19] N. Martins and L. Lima, Determination of suitable locations for
power system stabilizers and static var compensator for damping
electromechanical oscillations in large scale power systems, IEEE
Trans. Power Syst., vol. 5, no. 4, pp. 14551469, Nov. 1990.
[20] X. Yang and A. Feliachi, Stabilization of inter-area oscillation modes
through excitation systems, IEEE Trans. Power Syst., vol. 9, no. 1, pp.
494502, Feb. 1994.
[21] S. Skogestad and I. Postlethwaite, Multivariable Feedback ControlAnalysis and Design. New York: Wiley, 1996.
[22] M. Djukanovic, M. Khammash, and V. Vittal, Application of the structured singular value theory for robust stability and control analysis in
multimachine power systems. I. framework development, IEEE Trans.
Power Syst., vol. 13, no. 4, pp. 13111316, Nov. 1998.
[23] H. Nguyen-Duc, L. Dessaint, and A. Okou, Power system robust stability analysis using structured singular value theory and model reduction method, in Proc. IEEE PES General Meeting, 2009.
[24] D. Dotta, A. S. Silva, and I. C. Decker, Wide-area measurement-based
two-level control design considering signal transmission delay, IEEE
Trans. Power Syst., vol. 24, no. 1, pp. 208216, Feb. 2009.
[25] F. Okou, L.-A. Dessaint, and O. Akhrif, Power systems stability enhancement using a wide-area signals based hierarchical controller,
IEEE Trans. Power Syst., vol. 20, no. 3, pp. 14651477, Aug. 2005.
[26] U. Mhaskar and A. Kulkarni, Power oscillation damping using FACTS
devices: Modal controllability, observability in local signals and location of transfer function zeros, IEEE Trans. Power Syst., vol. 21, no.
1, pp. 285294, May 2006.
[27] J. Hauer, C. Demeure, and I. Scharf, Initial results in Prony analysis
of power system response signals, IEEE Trans. Power Syst., vol. 5, no.
1, pp. 8089, Feb. 1990.
[28] J. Quintero and V. Venkatasubramanian, A real-time wide-area control framework for mitigating small signal instability in large electric
power systems, in Proc. IEEE 38th Hawaii Int. Conf. System Sciences,
2005.
1927
LECTROMECHANICAL oscillation is a common problem in large electric power systems. This problem involves many groups of large generators oscillating against each
other. As the level of power transfer increases, the damping
factors of electromechanical oscillations tend to decrease.
Thus, as nowadays power systems are operating closer to
their limits, the improvement of power oscillations damping
is becoming more and more important.
The conventional approach in the design of power system
damping controllers is based on linearization of power system
model around a nominal operating point. The controller is
tested at different operating points to evaluate its performances. Even though this assessment procedure is simple, it
doesnt guarantee the controller robustness [1]. Real power
systems continuously experience several different perturbations and changes of operating conditions. Moreover, there
is always some uncertainty in the power system model due
to inaccurate knowledge of its parameters, or neglected dynamics. It is therefore essential to rely on a robust method
when analyzing the closed loop performance of a real power
system. This is particularly true for the emerging power
system controllers, such as FACTS controllers using wide area
signals. FACTCS controllers are in general less robust than
Power System Stabilizer (PSS) [2]. Moreover, the utilization
of wide area signals introduces additional uncertainties, in
communication delays for instance.
In recent years, several researches have been directed towards the application of robust control theory for the design of
power system controllers. Many of them include the utilization
This work was supported by the Natural Sciences and Engineering Research
Council (NSERC).
H. Nguyen-Duc and L-.A. Dessaint are with the Dpartement de gnie
lectrique, cole de Technologie suprieure, Montral, QC H3C 1K3 Canada
(email: huy.nguyen-duc.1@ens.etsmtl.ca; dessaint@ele.etsmtl.ca).
A.F. Okou is with the Department of Electrical and Computer Engineering,
Royal Military College of Canada, Kingston, ON K7K 7B4 Canada.
II. BACKGROUND
A. Linear Fractional Transformation and Structured Singular
Value
The general frame work for many robust control studies
is based on a Linear Fractional Transformation (LFT), de-
I. I NTRODUCTION
978-1-4244-4241-6/09/$25.00
B mn H thng in - i 2009
hc Bch IEEE
Khoa H Ni
273
Huy Nguyen-Duc, Student Member, IEEE, Louis-A Dessaint, Senior Member, IEEE,
and Aim Francis Okou, Member, IEEE
(2)
m
i=1 i Ai
m
i=1 i Ci
m
i=1 i Bi
m
i=1 i Ci
x(t)
u(t)
A Bi
x(t)
B0
m
=
+ i=1 i i
.
u(t)
D0
Ci Di
(5)
In Eq (5), each i represents a specic parameter variation.
The matrices A0 , B0 , C0 , D0 are state space matrices of the
linearized power system at the nominal operating point, which
consists of n states, nu inputs, and ny outputs. The state space
matrices for other operating points are described as a linear
approximation of those obtained at the nominal operating point
and the variables i . The linear state space uncertainty can then
be transformed into a LFT representation. We rst decompose
Ai Bi
, using singular value decomposition.
the matrices
Ci Di
A Bi
Let ri := rank i
, then
Ci Di
x(t)
=
y(t)
A0 +
C0 +
A0
C0
Ai
Ci
Ei
Bi
=
Di
Fi
B0 +
D0 +
Gi Hi ,
(6)
where:
Ei
R(n+nu )ri ,
Fi
Fig. 2.
Gi Hi Rri (n+ny ) .
A0 B0 E1 Em
x
x(t)
y C0 D0 F1 Fm u
z1 G1 H1
0
0
w1 , (7)
=
.. ..
.
..
..
.
.
..
..
..
. .
.
.
zm
wm
0
G m Hm 0
where the perturbation has the following structure:
= {diag[1 Ir1 , .., m Irm ] : i R}.
(8)
Fig. 1.
pi =
pmax pmin
pmax
+ pmin
i
i
i
+ i
i ,
2
2
(9)
In robust stability studies, the perturbations i are normalized so that |i | 1. Therefore, if we have m parameter
to pmax
,
variations p1 , .., pm , with each pi ranging from pmin
i
i
then each parameter pi is expressed as function of i as
follows:
gsi (s) =
(b1 )s (b1 b2 )
.
(s )2 + 2
(11)
By comparing (10) and (11), the new values for bi must be:
i
i
b1 = ci1 bi1 + ci2 bi2 ; b2 = ci1 bi2 bi1 ci2 .
For a rst order subsystem, the transformation is simple: A
subsystem
x = i x + bi u
(12)
y = ci x
is replaced by:
x = i x + b u
y=x
(13)
where b = bi ci .
As an example, the state matrices of a reduced 5th order
Kundur power system [13], illustrated in Fig. 4, at its nominal
operating point are:
Fig. 3.
(ci1 bi1 + ci2 bi2 )s (ci1 bi1 ci1 bi2 + bi1 ci2 + ci2 bi2 )
.
(s )2 + 2
(10)
0.0485
3.7054
a = 0.1086
0.1496
0.0075
3.7054
0.0122
0.0571
0.0740
0.0037
0.1086
0.0571
3.1087
16.4597
0.5830
0.1496
0.0740
16.4597
11.2099
0.7124
0.0075
0.0037
0.5830 ;
0.7124 .
0.0488
The
reduced
system
has
ve
eigenvalues:
0.0113, 7.1790 15.9572i, 0.0294 3.7062i.
After being transformed to modal canonical form, rearranged
and normalized (for c), the state space matrices become:
0.0294
3.7062
0
a=
0
0
3.7062
0.0294
0
0
0
0
0
7.1790
15.9572
0
0
0
15.9572
7.1790
0
0
;
0
0.0113
1.4
0.8
0.6
0.4
0.2
100.3
Fig. 5.
100.4
100.5
100.6
100.7
Frequency(rads/s)
100.8
100.9
1
Full order system
7th order system
0.9
0.8
0.7
0.6
0.5
upper bound
1.2
upper bound
0.4
0.3
0.2
0.1
100.3
Fig. 4.
Kundur system
Fig. 6.
100.5
100.7
Frequency(rads/s)
100.9
277
7 0.33
0.26
0.19
0.13
0.06
6.5
6
Gen. 3 mode
6
5.5 0.41
Table I
PARAMETER VARIATION RANGES
Parameter
P9
P33
P47
4.5
0.5
Inter-area mode
4
0.59
3
-2
Fig. 7.
-1.5
-1
-0.5
0.3
0.9
Full order system
7th order system
0.8
15th order
18th order
29th order
32nd order
0.25
upper bound
3.5
Range
40MW-1240MW
31MW-1303MW
40MW-1240MW
0.2
0.15
0.5
0.1
0.4
0.05
0.3
0
0.2
0.1
100.3
Fig. 8.
100.5
100.7
Frequency(rads/s)
100.9
Fig. 9.
100.1
100.2
100.3
100.4
Frequency(rads/s)
100.5
100.6
100.7
Fig. 10.
Table II
S.S.V COMPUTATION TIME FOR N EW E NGLAND POWER SYSTEM MODELS
Comp. time (sec)
53
74
199
251
15th order
18th order
29th order
32nd order
0.25
upper bound
Model order
15
18
29
32
0.3
0.2
0.15
0.1
0.05
0
100.1
Fig. 11.
100.3
Frequency(rads/s)
100.5
100.7
V. C ONCLUSION
Robustness analysis of power systems with parameter uncertainties has been studied in this work. A method to obtain
a reduced order LFT structure is also proposed for SISO
systems, in order to reduce the high computation cost, which
is inherent to analysis. It has been shown that reduced LFT
models can accurately retain the behaviour of system most
dominant modes, with respect to parameter variations. The two
case studies also show a strong coherence between structured
B mn H thng in - i hc Bch Khoa H Ni
Aim Francis Okou (M04) received the Dipl.Ing. degree in electrical engineering from cole Suprieure Interafricaine de llectricit, Cte dIvoire,
in 1993, the M.Eng. degree and the Ph.D. in electrical engineering from
cole de Technologie Suprieure (TS), Montreal, QC, Canada, in 1996 and
2001, respectively. During 2002-2005, he was a Postdoctoral Researcher in
the Department of Electrical Engineering at TS. Since 2005, he joined The
Royal Military College of Canada where he is currently an Assistant Professor
in the Electrical and Computer Engineering Department. His research interests
include the application of robust and nonlinear control techniques to largescale systems
B mn H thng in - i hc Bch Khoa H Ni
280
Surachai Chaitusaney
I.
INTRODUCTION
1642
281
I dp ,ref
d
1
V pq
2 Vp
3 V pq V pd
Pref
Qref
d
Iinv
, ref
d
Iinv
I dp
I qp ,ref
q
Iinv
, ref
q
Iinv
abc
U PWM
, ref
I qp
dq
I inv
, ref ,lim
dq
I inv
,ref
dq
I inv
, ref
dq
Iinv
,ref ,lim
dq
Iinv
, sat
abc
I inv
, ref
Power controller
I dp ,ref
q
I p ,ref
2
=
3 Vd
V pd
q
V p
( ) ( )
2
+ V pq
V pq Pref
V pd Qref
(1)
q
I inv , ref
I pd , ref
=
I qp , ref
I d I dp
+ inv
q
I inv
I qp
2 f c
=
s
2 f c
+
I dp ,ref
q
I
p ,ref
I d I pd
+ inv
q
I inv
I qp
(3)
V pabc
(2)
q
I inv,ref
abc
I inv
GPR = k p +
2ki
(4)
s + 2
2
1643
282
V pabc
, ref
V pabc
I abc
p
abc
I inv
, ref
abc
I inv
, ref ,lim
abc
Iinv
, sat
abc
Iinv
, ref
abc
I inv
, ref , lim
L
N
Reset
1644
I inv,ref =
Pref jQref
3V p*
+ jC f
Vp
(5)
283
-Ithres
Va1
Va0
Va2
Phase B Phase C
Fault
instant
V p*
= V pYpp +
q =1, p q
YpqVq
p = 1, n q p (6)
k=k+1
Bus count p=1
PV bus?
Yes
Q kp +1 = imag V p*
( )
I abc = AI 012
(7)
a ,
a2
1
3
a= + j
2
2
n
k
k
V p Y pp + Y pqVq
q =1
p q
slack bus?
Yes
I inv,ref I thres ?
Yes
n
1 Pp jQ p p 1
V pk +1 =
YpqVqk +1 YpqVqk
Y pp V * k
q =1
q = p +1
( )
Calculate change in
voltage of bus p?
V pk = V pk +1 V pk
V pk +1 =
V pk > max V k ?
p 1
n
1
I inv, sat YpqVqk +1 YpqVqk
Y pp
q =1
q = p +1
Yes
Set max V k = V pk
No
p=p+1
The end bus?
p > n?
Yes
No
Replace V pk by V pk +1
p = 1, n p s
Yes
No
max V k > ?
1 1
A = 1 a 2
1 a
Yes
Q p = Q kp +1
PQ bus?
(8)
Pp jQ p
1645
284
VApositive p.u.
Bus 2
0.820.02
I Apositive
I Azero
0.472.99
Bus 3
rad
rad
4.48 0.98
Bus 4
0.42 0.06
rad
I Anegative
rad
4.182.08
0.81 0.39rad
IF
rad
1.412.99rad
ing IBDG. However, the increases are small (48.8 (A) in phase
B and 48.15 (A) in phase C) because the power of IBDG is
small and the contributed current is limited due to the current
limiter in the control system of the IBDGs inverter. The peak
values of fault currents after installing IBDG are 1090 (A) in
phase B and 945 (A) in phase C.
CONCLUSION
1646
285
APPENDIX
Parameters of the simple system in Section 4:
[1]
1647
286
I. INTRODUCTION
Along with benefits such as renewable energy usage and
power quality improvement, installing DG in distribution
systems causes some detrimental impacts. Many of them are
concerned with protection system operations. These impacts,
which can limit DG, comprise reach reduction, miscoordination of protective devices, false tripping, and reclosing [1]-[3].
Therefore, a number of researchers have tried to prevent them
so that DG owners can install the maximum DG without any
troubles to utility systems [4]-[5]. However, the reach
reduction or the relay sensitivity degradation problem,
which definitely happens at any added DG, still has not
received much concern. Most of discussions are based on the
bolted three-phase fault and use descriptive analyses [2]-[3].
This paper not only mathematically analyzes the protection
reach reduction problem with all fault types, but also presents
an applicable algorithm to obtain the maximum DG.
Distribution systems are usually grounded and protected by
overcurrent protection schemes. Normally, all utility relays are
set to detect a fault occurring inside a required area with one
source supplying, e.g. the utility substation. Adding a new
source such as DG can decrease the fault current flowing from
the substation through utility relays, causing their sensitivity to
degrade [6]. When the fault current is decreased too much and
less than the relay pick-up value, the required protected area
becomes smaller and electric elements in unprotected section
of the system may be damaged. The larger the DG, the more
possibilities for this issue occurs.
To analyze the effects of DG in the protection reach
reduction problem, both phase and ground faults must be
included. This involves the consideration of DG transformer
connection. Section II will analyze this issue, and then an
appropriate connection type will be proposed. After analyzing
impacts of DG on the reach of relay in Section III, a practical
and robust algorithm for maximizing DG can be achieved in
Section IV using sensitivity-based method. This method
gradually increases DG until the constraints of system
operating limits and protection reach reduction are reached. It
is then applied to the IEEE 34 Node Test Feeder in Section V.
B
mn H thng in - i hc2011
Bch Khoa
978-1-4577-0255-6/11/$26.00
IEEEH Ni
1010
287
TENCON 2011
1011
288
is still able to sense the fault. This means the relay sensitivity,
represented by the ratio of fault current through the relay to the
corresponding relay pick-up current, is quick enough (larger
than 1) for the relay to operate. This can be achieved by
limiting DG because of the inverse relationship between DG
capacity and its transient reactance.
In practice, the POF of utility relay is set based on the
current under maximum load condition. Before adding DG,
this current can be estimated from the load flow program with
maximum system load condition. For the DG relay at the point
of common coupling (PCC), the POF is set based on the
current generated at nominal DG power. Thus, the pick-up
current of DG relay will increase if the DG increases. Fig. 1
illustrates the relation between DG and the pick-up current.
To detect the fault successfully, the sensitivity of utility
relay must be larger than one. This is done if DG is smaller
than DGmax as shown in Fig. 1.
[P
... Q
]T
J1
J 2
J3
J 4
... V
T
....... (5)
1012
289
1013
290
If the full load condition is applied for limiting XN, the value of
86 is the maximum reactance that ensures the effectively
grounded condition for the system under unintentional
islanding condition.
As a summary after using this algorithm, the maximum DG
that can be installed in the system at Bus 15 is 1.1 MW and a
grounding reactance which has the value ranging from 2 to 86
should be used to mitigate the operation of protective
devices.
VI. CONCLUSIONS
REFERENCES
[1] P.P. Barker and R.W. De Mello, "Determining the impact of distributed
generation on power systems. I. Radial distribution systems," in Proc.
2000 IEEE Power Engineering Society Summer Meeting, vol. 3.
[2] R.C. Dugan and T.E. McDermott, "Operating conflicts for distributed
generation on distribution systems," in Proc. 2001 IEEE Rural Electric
Power Conference, pp. A3/1 - A3/6.
[3] S. Chaitusaney and A. Yokoyama, "An Appropriate Distributed
Generation Sizing Considering Recloser-Fuse Coordination," in Proc.
2005 IEEE Transmission and Distribution Conference and Exhibition:
Asia and Pacific Conf., pp. 1-6.
[4] Calderaro, V.; Piccolo, A.; Siano, P., "Maximizing DG penetration in
distribution networks by means of GA based reconfiguration, " in Proc.
2005 International Conference on Future Power Systems.
[5] Celli, G. ; Ghiani, E. ; Mocci, S. ; Pilo, F. , "A Multi-Objective Approach
to Maximize the Penetration of Distributed Generation in Distribution
Networks," in Proc. International Conference on Probabilistic Methods
Applied to Power Systems, 2006. PMAPS 2006.
[6] Dao Van Tu; Chaitusaney S., "Maintaining the Reach of Protective
Devices in Distribution System with Penetration of Distributed
Generation," ECTI 2011 Conf., Thailand.
[7] Bloomquist, Walter C., "Select the Right Transformer Winding
Connection for Industrial Power Systems," IEEE Trans. Industry
Applications, vol. IA-11, pp. 641-645, 1975.
[8] Mozina, C.J., "Interconnect Protection of Dispersed Generators," in
Proc. 2001 IEEE/PES Transmission and Distribution Conference and
Exposition, pp. 709 - 723 Vol.2.
[9] Arritt, R.F.; Dugan, R.C., "Distributed generation interconnection
transformer and grounding selection, in Proc. 2008 IEEE Power and
Energy Society General Meeting - Conversion and Delivery of Electrical
Energy in the 21st Century.
[10] Zhang, W.Y.; Zhu, S.Z.; Zheng, J.H.; Zhang, H., "Impacts of Distributed
Generation on Electric Grid and selecting of Isolation Transformer," in
Proc. 2005 IEEE/PES Transmission and Distribution Conference and
Exhibition: Asia and Pacific.
[11] IEEE Guide for the Application of Neutral Grounding in Electrical Utility
Systems - Part 1 - Introduction, IEEE Standard C62.92.1, 2000.
[12] IEEE Standard for Interconnecting Distributed Resources with Electric
Power Systems, IEEE Standard 1547.2-2008.
1014
291
ACKNOWLEDGEMENT
I. INTRODUCTION
Distributed Generation (DG) has been more and more
widely interested due to its system support benefits, e.g.
efficient usage of energy resources, especially renewable
resources, improvement of power quality like voltage
regulation, reliability, etc. [1]. However, the interconnection of
DG to a distribution system, which is normally designed for
radial feeder structure operation, may cause problems.
Before DG installation, the setting values of feeder
protective devices are typically set at a fixed pick-up value and
a particular tripping time characteristic based on load and fault
current analysis with one source supplying. The introduction of
DG results in the changes of currents in both magnitudes and
directions. These may cause the miscoordination of protective
devices [2] [5]. The sympathetic tripping or false tripping is
concerned when a protective device of unfaulted feeder, at
which DG is installed, trips to isolate that healthy feeder from a
faulted one. Changes of fault current flowing through a relay or
a fuse may also make that current become less than the pick-up
value which is set before adding DG. This was reported at [2]
as the reduced reaching problem.
Overcurrent protections, as the main protection scheme in
distribution systems, comprise phase overcurrent protection
and ground overcurrent protection. The former uses phase
currents to operate while the latter uses the zero-sequence
currents. This will be mentioned in more details in Section II.
Both phase and ground overcurrent functions may be found in
one relay. However, previous researches are mostly based on
three phase fault current analysis [1]-[5]. Therefore, those
researches do not cover all cases of faults and DG impacts on
the operation of ground overcurrent protection.
This paper analyzes currents flowing through protective
devices under both phase and ground fault conditions. To
analyze ground fault currents, system grounding methods and
connection types of transformer must be considered. These are
discussed thoroughly in Section III. The reduced reaching
problem is then evaluated in Section IV. In addition, some
B mnThe
H 8th
thng
in - i
hc Bch Khoa
H Ni Computer,
Electrical
Engineering/
Electronics,
Page 696
292
Ground
Faults
''
I12
=
V3' ( 0 )
(2)
( 0)
Z1 + Z12 + Z 23 + Z f + Z f ( Z1 + Z12 + Z 23 ) Z new
V3''
(3)
2
3
'
I12
=
V3'' ( 0 )
Z1 + Z12 + Z 23 + Z f
B mnThe
H 8th
thng
in - i
hc Bch Khoa
H Ni Computer,
Electrical
Engineering/
Electronics,
Page 697
293
B mnThe
H 8th
thng
in - i
hc Bch Khoa
H Ni Computer,
Electrical
Engineering/
Electronics,
Page 698
294
[2]
[3]
[5]
[6]
[7]
[8]
IX. BIOGRAPHIES
Dao Van Tu received his Engineer degree and Master degree in power
systems at Ha Noi University of Science and Technology in
2007 and 2009 respectively. Now he is pursuing a Ph.D.
degree in electrical engineering under the supervisions of Dr.
Surachai Chaitusaney at Chulalongkorn University, Thailand
and Professor Akihiko Yokoyama at The University of
Tokyo, Japan. He is interested in distributed generation and
power system protection and control.
Surachai Chaitusaney received the B.Eng. and M.Eng. in
Electrical Engineering, Chulalongkorn University, Thailand,
in 2000 and 2002 respectively. From 2002 to 2004, he
worked at Energy Research Institute and Center of Excellent
in Power Technology, Chulalongkorn University. In 2007,
he obtained the Ph.D. degree from the University of Tokyo,
Japan, with JICA scholarship. At present, he is a lecturer at Department of
Electrical Engineering, Chulalongkorn University. His interests include
distributed generation and system reliability.
VI. APPENDIX
Parameters of system in Fig. 1:
Transmission system: Xshort circuit = 0.
B mnThe
H 8th
thng
in - i
hc Bch Khoa
H Ni Computer,
Electrical
Engineering/
Electronics,
Page 699
295
VIII. REFERENCES
[1]