Escolar Documentos
Profissional Documentos
Cultura Documentos
College of Energy and Electrical Engineering, Hohai University, Nanjing 210098, China
State Grid Electric Power Research Institute, Nanjing 210003, China
a r t i c l e
i n f o
Article history:
Received 17 March 2015
Accepted 9 May 2015
Keywords:
Dynamic optimal reactive power ow
Voltage stability
Reactive voltage network partition
VQ curves
Dynamic reactive power reserves
Branch-bound and primaldual interior
point method
a b s t r a c t
A day-ahead voltage stability constrained dynamic optimal reactive power ow (VSC-DORPF) model is
proposed in this paper. The amount of dynamic reactive power reserves (DRPR) is used as a measure
of voltage stability of power system. The effective dynamic reactive power reserves (EDRPR) of reactive
power sources are calculated to obtain DRPR of each area and the maximum variations in reactive power
generation under contingency are taken as the required minimal DRPR for each area. Then the DRPR are
introduced into the VSC-DORPF model as one of multiple objective functions and constraints in order to
enhance the voltage stability of power system. A hybrid method, integrated by branch-bound method
and primaldual interior point (PDIP) method, is proposed to solve this VSC-DORPF problem. The discrete
control variables and the time coupled constraints are handled by the proposed branching and pruning
principles. As a result, the VSC-DORPF problem is decomposed into a series of optimal reactive power
ow (ORPF) problems with continuous control variables only. Numerical tests with IEEE 30-bus system
and IEEE 118-bus system show that the proposed model and method are effective.
2015 Elsevier Ltd. All rights reserved.
Introduction
Dynamic optimal reactive power ow (DORPF) determines the
proper settings of reactive power control devices in next day based
on the day-ahead load forecast and active power scheduling plan in
order to reduce the daily network losses, enhance voltage prole
and avoid excessive operation.
As voltage stability has not been taken into account in the general DORPF model, the scheduling results cannot respond to the
impact which acute load uctuation bring to power system. A
fuzzy membership function of bus voltage was taken as one of
optimization objectives to increase voltage quality in [1].
However, keeping bus voltages within qualied ranges simply
cannot maintain voltage stability. Thus it is necessary to carry
out further researches on DORPF considering voltage stability.
Dynamic reactive power reserves (DRPR) have always been linked
with voltage stability as they have a signicant effect on the reliable operation of power system [2]. In [3] an optimal reactive
power ow (ORPF) model with DRPR of power system being one
of objective functions was proposed. It is worth noting that since
each reactive power source gives a different impact on the entire
system, DRPR of large system cannot be obtained by merely summing up individual reserves. Thus the network was partitioned
into several areas and the reactive power sources were assigned
weighting factors based on the reactive power load margin of each
area in [3]. But it is unreasonable to give the same factors to the
reactive power sources in an area. Moreover, it cannot be guaranteed that there are sufcient DRPR in each area to maintain voltage
stability merely by the weighted sum of individual reserves in
objective functions without any explicit constraints.
On the other hand, DORPF problem is essentially a large scale
mixed integer nonlinear programming problem. The presence of
a large number of discrete control variables and time coupled constraints makes it difcult to solve. Different methods have been
proposed and they can be classied basically into four categories.
(1) Simultaneous solution method [4]. The operation limits of control devices are described by the analytic mathematic expressions
of their control variables. The DORPF problem is solved as a whole
and the discrete control variables achieve their discrete values by
an embedded algorithm. Although this method usually shows good
performance on small test systems, its application on larger power
system will be hard. (2) Modern intelligent algorithm [1,5]. The
control variables in the whole day are encoded into an individual
and a modern intelligent algorithm is adopted to solve the problem. This kind of algorithm cannot be put into practical application
because of its stochastic nature. (3) Decomposition coordination
602
2. Operation constraints
V i;min 6 V ti 6 V i;max
i 1; . . . ; NB ; t 1; . . . ; N T
Q tg;j;eff Q tg;j P Q trs;k;min
k 1; . . . ; Narea t 1; . . . ; NT
j1
4
4. The constraints of control variables
i 1; . . . ; NG ; t 1; . . . ; NT
i 1; . . . ; NK ; t 1; . . . ; NT
i 1; . . . ; Nc ; t 1; . . . ; NT
3
2
PN G t
2
t
PNB t
NT
t
X
V
V
j1 Q g;j;eff Q g;j
P
i;set
i
i1
loss
4x1
5
min
x3
t x2
t
t
f1
f2
f3
t1
1
where NT is the number of intervals, N B and NG are the number of
buses and generators, x1 , x2 and x3 are the weighting factors of
optimization objectives, the rst objective component is daily network losses, P tloss is the active power losses at interval t, the second
component is voltage deviation, V ti is the voltage magnitude of bus i
at interval t, V i;set is the expected voltage magnitude of bus i, the
jK ti K t1
j 6 Sk;i;D K i;step
i
i 1; . . . ; NK ; t 1; . . . ; NT
jQ tc;i Q t1
c;i j 6 SQc;i;D Q c;i;step
i 1; . . . ; Nc ; t 1; . . . ; NT
8
9
K ti K it1 6 Sk;i;max
i 1; . . . ; NK
NT
X
Q tc;i Q t1
6 SQc;i;max
c;i
10
i 1; . . . ; Nc
where V i;max and V i;min are voltage limits of bus i, Q g;i;max and
Q g;i;min are reactive power limits of generator i, N K and N C are
the number of transformers and compensators, Q tc;i , Q c;i;max ,
Q c;i;min and Q c;i;step are the reactive power compensation, its
limits and step size of compensator i at interval t, K ti , K i;max ,
K i;min and K i;step are the ratio, its limits and step size of transformer i at interval t, xt is the vector of control variables and
state variables at interval t, Narea is the number of areas, NG;k
is the number of generators in area k, Q trs;k;min is the required
DRPR for area k at interval t, Sk;i;D , SQc;i;D , Sk;i;max and SQc;i;max are
maximum allowable action range between successive intervals and maximum allowable action number in a day of
transformer i and compensator i respectively.
Constraints
1. Power ow equations
11
t1
g t xt 0 t 1; . . . ; NT
third component is DRPR of power system, Q tg;j and Q tg;j;eff are the
reactive power output and its effective limit of generator j at
t
t1
Fig. 1. VQ curve.
603
where m is the number of controlled buses, n is the number of reactive power source buses.
Thus the spatial electrical distance between bus i and bus j can
be calculated using (15).
Mij
aij
DV j DV j DV i 1
DV i DQ i DQ i
12
dij lg jaij j
13
d11
d12
d1n
6d
6 21
D6
6 ..
4 .
d22
..
.
..
.
d2n
..
.
7
7
7
7
5
dm1
dm2
dmn
q
2
2
di1 dj1 din djn
15
Based on the electrical distance, a hierarchical clustering algorithm [15] is adopted to obtain the optimal number of areas and
the areas of reactive power source buses. In order to make sure that
controlled buses are partitioned into the same area with their
corresponding reactive power source buses, a mapping partition
algorithm is used to determine the areas of controlled buses. The
electrical distance between the controlled bus and each reactive
power source bus is calculated and the controlled bus is classied
into the same area with its electrically closest reactive power
source.
The computation of DRPR for each area based on VQ-curve method
Reactive power reserves can be classied into two types: static
reactive power reserves and DRPR. Static reactive power reserves
cannot respond to unforeseen events effectively as they do not
have the ability to assist voltage control continuously while
dynamic reactive power sources can increase their production
quickly to maintain the controlled voltages at constant values in
a contingency. Thus DRPR are the major study objects in this paper.
The most straightforward denition of DRPR is the difference
between the maximum reactive power generation that determined
by the physical factors and its current value. However, the above
denition may not represent the useful quantity of DRPR since at
the voltage collapse point not all the amount of DRPR can be
utilized. Thus the concept of effective dynamic reactive power
reserves (EDRPR) of reactive power sources is introduced in this
paper which is dened as the difference between the reactive
power output at the voltage collapse point and current operation
point [16,17].
In order to make the process of voltage collapse insensitive to
load increasing modes, VQ-curve method is used to compute the
voltage collapse point [18,19]. First, the pilot bus of each area is
determined based on the electrical distance obtained in partition
process. The average electrical distance from controlled bus to all
the reactive power source buses of the area is given by formula
(16). Choose the bus with closest average electrical distance to
be the pilot bus. The obtained pilot bus is the electrical central
bus of the area with moderate electrical distance to each reactive
power source and it can give a comprehensive consideration to
all the reactive power sources in the area.
j
D
X
2
d
i2G ij
16
14
604
Q rs;k
N G;k
X
Q g;j;eff Q g;j k 1; . . . ; Narea
17
j1
Q rs;k;min
N G;k
X
Q g;j;eff Q g;j
k 1; . . . ; Narea
18
Branching process
Considering the inuences of different control variables on
objective function are different, a proper ranking of discrete control
variables before branching can improve the calculation efciency.
As compensators change the quantity of reactive power directly
while transformers only effect its distribution, the compensators
are given a high priority. Besides, another priority list according
to the difference between the non-discrete value of discrete control
variable and its nearest discrete value is also used here.
The branching method used in this paper is called dichotomy.
If the discrete control variable does not achieve its discrete
value, bound constraints are introduced to construct two new
sub-problems. As constraints (8) and (9) should be taken into
account during the solution process, it is need to judge whether
the sub-problem is at the same interval with its parent problem.
If so, the bound constraints are introduced by the method shown
in Fig. 3. In the gure, ~
xi is the continuous solution of the discrete
control variable xi between [xi;min ; xi;max ], xi;step is its step size, Ii is a
xi 6 Ii xi;step .
discrete value of discrete control variable xi and Ii 6 ~
If not, the bound constraints should be calculated using formula
(19) and (20).
T1
19
Ii xi;step ; ~xT1
Si;D xi;step \ Ii xi;step ; xi;max
i
20
where ~xT1
is the value of control device xi at the previous interval,
i
Si;D is the maximum allowable action range between successive
intervals.
Solution of sub-problems by PDIP method
As the discrete control variables are considered as continuous
variables within their bounds in sub-problems, each sub-problem
becomes essentially an ORPF problem with only continuous
control variables as follows. It can be solved by PDIP method [11]
directly.
j1
Table 1
The data of compensators in IEEE 30-bus system (MVar).
Bus
Lower limits
Upper limits
Hierarchy capacity
10
24
0
0
20.0
10.0
4.0
2.0
Table 2
The network losses of IEEE 30-bus system (MW).
Interval Load rate
(p.u.)
Initial
values
Continuous
method
Upper
bound
This
Paper
Ref.
[1]
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
3.1731
3.1542
3.0841
3.0249
3.0394
3.2190
3.6636
3.8800
4.2756
4.8910
5.0724
5.0930
4.2588
4.2264
4.7725
4.8531
4.8901
4.9512
4.8099
5.3058
5.4403
4.9945
4.0193
3.3708
2.6732
2.6561
2.5922
2.5382
2.5515
2.7149
3.1167
3.3109
3.6968
4.1018
4.2727
4.2920
3.6815
3.6519
3.9894
4.0659
4.1009
4.1587
4.0249
4.4906
4.6151
4.1995
3.4619
2.8525
2.6735
2.6563
2.5926
2.5386
2.5519
2.7152
3.1175
3.3124
3.7002
4.1081
4.2804
4.2999
3.6848
3.6550
3.9947
4.0719
4.1072
4.1654
4.0305
4.5003
4.6262
4.2066
3.4640
2.8529
2.6735
2.6563
2.5925
2.5386
2.5518
2.7153
3.1173
3.3111
3.6973
4.1042
4.2764
4.2959
3.6820
3.6523
3.9911
4.0681
4.1033
4.1615
4.0268
4.4949
4.6192
4.2026
3.4621
2.8556
2.6736
2.6563
2.5926
2.5386
2.5518
2.7157
3.1175
3.3120
3.6975
4.1073
4.2766
4.2958
3.6824
3.6529
3.9919
4.0680
4.1063
4.1647
4.0298
4.4977
4.6251
4.2066
3.4624
2.8534
0.7677
0.7653
0.7563
0.7486
0.7505
0.7735
0.8271
0.8517
0.8984
0.9447
0.9635
0.9656
0.8966
0.8931
0.9321
0.9407
0.9446
0.9510
0.9361
0.9869
1.0000
0.9555
0.8703
0.7923
This paper
Ref. [1]
Bc10
Bc24
T11
T12
T15
T36
2
2
1
1
1
1
4
2
1
1
1
1
Ploss
min x1
x2
f1
s:t: gx 0
PN B
i1 V i
V i;set 2
f2
V i;min 6 V i 6 V i;max
PN G
x3
j1 Q g;j;eff
Q g;j
f3
i 1; . . . ; NB
i 1; . . . ; N G
K 0i;min 6 K i 6
Q 0c;i;min 6 Q c;i
1; . . . ; N K
K 0i;max i
6 Q 0c;i;max
21
i 1; . . . ; Nc
NG;k
k 1; . . . ; N area
j1
where K 0i;min , K 0i;max , Q 0c;i;min and Q 0c;i;max are the ratio limits of transformer i and compensation limits of bus i after branching
respectively.
605
22
t1
Solution procedure
Pruning process
Once the sub-problem meets pruning principles in the solution
process, it should be removed. Four pruning principles of
VSC-DORPF problem are listed. (1). The sub-problem is infeasible;
(2). All the discrete control variables have achieved discrete values
in the sub-problem; (3). The discrete control variables in the
sub-problem do not meet constraints (10) and (11); (4) The objective function value of sub-problem is greater than or equal to the
upper bound.
Brief instructions of pruning principles (4) are given as follows.
As the VSC-DORPF problem is decomposed into a series of static
ORPF problems, the objective function value of the entire problem
cannot be obtained until all the intervals in a day have been solved.
Thus, a heuristic strategy is adopted in pruning principle (4). First a
xed upper bound is determined in advance. Based on the periodicity between interval 0 and interval 24, the values of discrete control variables at interval 24 can be taken as the initial values at
interval 0 in the next day. Give discrete control variables at each
interval the initial values and solve the ORPF problems by PDIP
606
Table 6
Dynamic reactive power reserves of IEEE 118-bus system (MVar).
Interval
Area
Before
optimization
Network losses
minimized only
The proposed
model
Total
2
3
4300.26
749.84
301.74
4684.22
767.13
192.86
4894.38
872.56
309.78
10
Total
2
3
4260.79
790.26
303.27
4688.82
824.96
174.01
4989.62
919.76
306.09
20
Total
2
3
4214.23
783.45
303.05
4652.01
812.08
170.21
4979.11
921.61
304.73
Simulations
The proposed VSC-DORPF is applied to IEEE 30 and 118-bus
system to test its effectiveness. The algorithm has been coded in
Matlab R2010b on a computer with Pentium R 3.0 GHz CPU and
3 GB memory.
IEEE 30-bus system
Conclusion
As voltage stability is not taken into account in general DORPF
model, the amount of DRPR is taken as a measure of voltage stability and a VSC-DORPF model is presented in this paper. It can reduce
daily network losses, improve voltage quality and enhance voltage
stability of power system. A hybrid method combined by
branch-bound method and PDIP method is adopted to solve the
VSC-DORPF problem. The discrete control variables and time coupled constraints can be handled accurately by the proposed
branching and pruning principles. Numerical tests with IEEE 30
and 118-bus system show that the proposed model and method
are effective.
Acknowledgements
Table 4
The data of compensators in IEEE 118-bus system (MVar).
Bus
Lower limits
Upper limits
Hierarchy capacity
5
17, 107, 110
34, 4446, 48, 74, 83
37
79, 82, 105
50.0
0
0
30.0
0
0
10.0
20.0
5.0
30.0
10.0
2.0
5.0
5.0
5.0
Table 5
The optimization results of IEEE 118-bus system.
Interval
Before
optimization
Network losses
minimized
only
Time (s)
The
proposed
model
10894.39
12603.44
Network
losses
(MW)
5
10
20
76.01
118.64
129.23
58.96
94.22
103.10
64.55
99.88
109.10
Voltage
deviation
(p.u.)
5
10
20
0.03
0.04
0.04
0.38
0.36
0.35
0.02
0.03
0.03
References
[1] Zhang A, Yang H. Pareto-set based multi-objective dynamic reactive power and
voltage control. In: International conference on sustainable power generation
and supply; 2009. p. 16.
[2] Leonardi B, Ajjarapu V. Investigation of various generator reactive power
reserve (GRPR) denitions for online voltage stability/security assessment. In:
IEEE power and energy society general meeting, 2008. p. 17.
[3] Dong F, Chowdhury BH, Crow ML, et al. Improving voltage stability by reactive
power reserve management. IEEE Trans Power Syst 2005;20:33845.
[4] Liu M, Canizares C, Huang W. Reactive power and voltage control in
distribution systems with limited switching operations. IEEE Trans Power
Syst 2009;24:88999.
607
[13] Estevam CRN, Rider MJ, Amorim E, et al. Reactive power dispatch and planning
using a non-linear branch-and-bound algorithm. IET Gener Transm Distrib
2010;4:96373.
[14] Xie Y, Chiang H. A novel solution methodology for solving large-scale
thermal unit commitment problems. Electr Power Compon Syst 2010;38:
161534.
[15] Lagonotte P, Sabonnadiere JC, Leost JY, et al. Structural analysis of the electrical
system: application to secondary voltage control in France. IEEE Trans Power
Syst 1989;4:47986.
[16] Capitanescu F, Van Cutsem T. Evaluation of reactive power reserves with
respect to contingencies. In: Bulk power system dynamics and control V; 2001.
p. 19.
[17] Song H, Lee B, Kwon SH, et al. Reactive reserved-based contingency
constrained optimal power ow (RCCOPF) for enhancement of voltage
stability margins. IEEE Trans Power Syst 2003;18:153846.
[18] Chowdhury BH, Taylor CW. Voltage stability analysis: VQ power ow
simulation versus dynamic simulation. IEEE Trans Power Syst 2000;15:
13549.
[19] Mousavi OA, Bozorg M, Ahmadi-Khatir A, et al. Reactive power reserve
management: preventive countermeasure for improving voltage stability
margin. In: IEEE power and energy society general meeting; 2012. p. 17.