Professional Documents
Culture Documents
I. I NTRODUCTION
RANSEINT
stability-constrained
optimal
power
flow (TSCOPF) is an effective tool for balancing
the economic concerns and the stability requirements in
power system operations. Given an unstable contingency, the
TSCOPF aims to optimize the system operating states while
retaining the ability to withstand the contingency should it
really occur [1].
Manuscript received August 19, 2015; revised October 28, 2015 and
November 30, 2015; accepted December 13, 2015. This work was supported in part by the Faculty of Engineering and Information Technologies,
University of Sydney through the Faculty Research Cluster Program and the
Early Career Researcher Development Scheme, and in part by China Southern
Power Grid Company under Project WYKJ00000027. The work of Y. Xu was
supported by the University of Sydney Postdoctoral Research Fellowship.
Paper no. TSG-00969-2015.
Y. Xu and Z. Y. Dong are with the School of Electrical and Information
Engineering, University of Sydney, Sydney, NSW 2006, Australia, and also
with the China Southern Power Grid Research Institute, Guangzhou 510000,
China (e-mail: eeyanxu@gmail.com; zydong@ieee.org).
J. Ma is with the School of Electrical and Information Engineering,
University of Sydney, Sydney, NSW 2006, Australia (e-mail:
j.ma@sydney.edu.au).
D. J. Hill is with the Department of Electrical and Electronic Engineering,
University of Hong Kong, Hong Kong, and also with the School of Electrical
and Information Engineering, University of Sydney, Sydney, NSW 2006,
Australia (e-mail: dhill@eee.hku.hk).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSG.2015.2510447
From a mathematics perspective, the TSCOPF is a largescale non-linear and non-convex programming problem containing differential-algebraic equations (DAEs), which is difficult to be directly solved. A variety of solution methodologies
have been proposed with varying pros and cons. In general, the methodologies can be classified into the following
categories: 1) numerical discretizing method [1][4], which
discretizes the differential equations into numerically equivalent algebraic ones, making the model tractable by classic
programming algorithms. However, the major drawback of
this approach lies in the heavy computational burden which
is proportional to the product of the number of generators
and the simulation steps; 2) direct method [5][7], which
analytically derives the required generation shifting amounts
for stabilizing the system and incorporates it into the ordinary OPF model. This approach is advantageous for its high
solution efficiency and transparency. However, it is usually
argued that it can only obtain a near-optimal solution; 3) model
reduction method [8][10], which approximates the original,
multi-machine model with a reduced, much smaller one; therefore the solution efficiency can be increased drastically but the
accuracy may be sacrificed; 4) evolutionary algorithm (EA)based method [11], [12], which drives an EA to stochastically
search the optima of the problem. Its advantage lies in higher
chance to obtain high-quality solutions, however it usually suffers from slow convergence and inconsistence issues. In this
regard, Xu et al. [13] developed a hybrid approach combining
an EA and classic programming algorithm to overcome these
drawbacks; 5) statistical method [14], [15], which acquires the
stable/unstable operating regions or rules through statistical
learning from a transient stability database and incorporates
the regions or rules as explicit constraints into the ordinary
OPF model. The stable/unstable operating region or rules are
transparent and interpretable which can also be used for stability monitoring. However, this method is sometimes sensitive
to the change of stability database. A systematical review,
discussion, and comparison on these methods can be found
in [16].
While the generator dynamics are well represented in
the TSCOPF models, the load dynamics has not been
properly treated yet. In the literature, almost all of the
reported methods assume the loads to be static (e.g., constant impedance [1][15]). But in real world, an actual load
bus is a mix of various static and dynamic components
such as induction motors. Such dynamic components have
a substantial impact on power systems short-term stability
c 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
1949-3053
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2
Fig. 1.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
XU et al.: ROBUST TSCOPF WITH UNCERTAIN DYNAMIC LOADS
TABLE I
T ESTING S CENARIOS D ETERMINED OF OA L8 (27 )
3/2 (m ),
and if m is symmetrically
m
(1)
s.t. g(x, u) = 0
h(x, u) 0
TSI(x, u, )
f (x, u)
(2)
(3)
(4)
NG
ai P2Gi + bi PGi + ci
(5)
i=1
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4
B. Equality Constraints
Eq. (2) corresponds to the power flow equations:
NB
V
Vj Gij cos ij + Bij sin ij = 0
Gi
Di
i
j=1
NB
Gi
(6)
j=1
N (t) = (MN )
iC
jN
(8)
Mj j (t)
jN
(9)
C. Inequality Constraints
Eq. (3) corresponds to the steady-state lower and upper
limits of the operational variables:
min
PGi PGi Pmax
i = 1, 2, . . . , NG
Gi ,
Qmin Q Qmax , i = 1, 2, . . . , N
Gi
G
Gi
Gi
(7)
min V V max ,
V
i
=
1,
2,
.
.
.
,
N
i
B
i
i
min
Li Li Limax ,
i = 1, 2, . . . , NL
where Li is the apparent power across the i-th branch.
D. Transient Stability Constraints
Eq. (4) is the transient stability constraint which corresponds
to a large set of DAEs characterizing the systems dynamic
behavior. In most of the literature, the transient stability criterion is expressed as the maximum post-disturbance rotor angle
deviation bounded by a pre-defined threshold, e.g., 180 [1][4].
However, this threshold is usually system and/or operating
state dependent, which is not accurate in reflecting the transient stability: if this type of thresholds are set smaller, the
resulting system tends to be conservative, which hence leads
to higher operating costs; if too relaxed, the stability may not
be retained even if the threshold is not exceeded [9].
In this paper, the TSI is implicitly expressed in terms of
the transient stability margin defined by the extended equalarea criterion (EEAC) [28]. The motivation is twofold: first,
it provides a way for quantitative and fast assessment of transient stability, which can measure the degree of stability while
greatly increasing the computation efficiency; second, it can
derive the required generation shifting to stabilize the system, based on which equivalent linearized stability constraints
can be formulated. In this way, it is possible to decompose
the original large-scale problem into a series of smaller and
tractable problems, diminishing significantly the complexity
of the problem and enabling parallel computing to improve
solution efficiency.
IV. E XTENDED E QUAL -A REA C RITERION
A. Principle
The EEAC was proposed by Xue et al. in [28]. Its state-ofthe-art version IEEAC [29], also known as single machine
where subscripts C and N denote the CMs and NMs, respectively; MC and MN are respectively the inertia coefficient of
CMs and NMs, calculated as:
Mi ; MN =
Mj
(10)
MC =
jN
iC
Pmi (t) (MN )1
Pmj (t)
Pm (t) = M (MC )1
jN
iC
Pe (t) = M (MC )1
(12)
Pej (t)
jN
iC
(13)
M = (MC MN ) (MC + MN )
(14)
where and here denote the rotor angle and angular speed
of the OMIB, respectively; Pm and Pe denote the mechanical
and electrical power of the OMIB, respectively.
Finally, the EAC is applied to the OMIB P plane
for quantifying the transient stability degree and extracting
stability information of the original multi-machine system.
B. Transient Stability Assessment
IEEAC or SIME provides the following TSA outputs [30]:
1) CMs and NMs, which determines an instability mode.
2) The time to instability Tu indicating the time that system loss synchronism. At this time, the curve of Pe
crosses Pm :
Pa (Tu ) = Pm (Tu ) Pe (Tu ) = 0, P a (Tu ) > 0 (15)
3) The time to first-swing stability Tr indicating the time
that the system can be declared as first-swing stable. At
this time, the curve Pe stops its excursion and return
back before crossing Pm :
Pa (Tr ) = Pm (Tr ) Pe (Tr ) < 0, (Tr ) = 0
(16)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
XU et al.: ROBUST TSCOPF WITH UNCERTAIN DYNAMIC LOADS
Eq. (18)-(21) reveal that by shifting real power output of CMs to NMs, the transient stability can be
restored [6], [28][30].
Numerous simulations have shown an approximate linear
relationship between changes of stability margin and OMIB
mechanical power at pre-contingency state [6], [28][30], i.e.,
= Pm (t0 )
(22)
(n2) (n1)
Pm (t0 )(n2) Pm (t0 )(n1)
(23)
PC
+
(24)
n
MC
MN
Note that the power loss is temporarily neglected in (20)
since it is unknown before the generation shifting. The power
loss will be implicitly compensated by NMs through (24)
which will be added as a constraint to the OPF model.
iC
(20)
(25)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6
Fig. 2.
Based on the concept of the trajectory sensitivity, a sensitivity index (SI) can be calculated to rank the relative influence
of the load model parameters on system transient stability:
SI =
TSI(x, u, + ) TSI(x, u, )
TSI
=
(26)
The sub-problems evaluate the stability of the master problem solution under each contingency and scenario and generate
stabilization constraints if unstable.
For each testing scenario determined by the TOAT, the
EEAC-based TSA is performed and if (4) is not satisfied, the
stabilization constraint, i.e., generation shifting between CMs
and NMs, is generated. To simultaneously stabilize multiple
contingencies/scenarios with least number of additional constraints, the contingencies/scenarios can be grouped according
to their resulting instability modes. Those having a common
instability mode are divided into one group. Each group is then
represented by the severest mode which is the one with the
smallest stability margin in that group. For each representative
instability mode, the stabilization constraint is:
M
M 1
us +
Pmi (t0 )
Pmi (t0 )
+
n
MC
MN
iC
iC
(27)
where P mi (t0 ) denotes the active power output of generator i
obtained from the master problem.
The stabilization constraint (27) conveys information about
how the generation dispatch should be modified to retain the
transient stability. It is important to note that (27) is a linear
constraint. Hence, it can be directly attached to (3) in the
master problem. In such a way, the original problem size can
be reduced to one similar to an ordinary OPF. Note that the
power loss neglected in (20) will be implicitly compensated
after adding (27) to the master problem.
C. Computation Steps
1) Prepare the computation data, including the system
data (network data and dynamic models), contingency set,
and the mean value and the standard deviation of
the load model parameters. The load model parameters can
be obtained from load composition statistics, load modeling practices [33], and/or advanced smart grid functions (e.g.,
smart meter data-based home appliance load modeling [34]).
2) For each load model parameter, estimate their impact
(around their mean value) on the transient stability of the system. This corresponds to calculating the SI based on Eq. (26).
Only those with high SIs are included in the uncertainty
parameter set whose dimension is denoted as M and the
remaining load model parameters are fixed at their mean
values.
3) For the uncertainty parameters, select an appropriate representative level (i.e., OA level), denoted as B. It will be shown
later that B=2 can be sufficient as the load model parameters
exhibit a quasi-linear effect on the stability margin.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
XU et al.: ROBUST TSCOPF WITH UNCERTAIN DYNAMIC LOADS
TABLE II
C ONTINGENCY S ET
TABLE III
BASE C ASE -I NITIAL G ENERATION D ISPATCH (MW) [13]
TABLE IV
M EAN VALUE () OF THE L OAD M ODEL PARAMETERS
Fig. 3.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8
TABLE VI
M ULTI -C ONTINGENCY S OLUTION R ESULTS (MW)
Fig. 4.
D. Multi-Contingency Case
All of the three contingencies are included simultaneously
in the TSCOPF model. The solution results are given in
Table VI and the system trajectories under each contingency
are shown in Fig. 5. The CMs are: G2, G3, G4, and G5 for C2;
and G9 for C3. The increased cost over the base case is
1815.7 $/Hr or 2.98%.
The obtained solution is stable for all of the three contingencies under all of the eight testing scenarios. To illustrate, the
system trajectories under the eighth testing scenario are shown
in Fig. 5. It is important to note that the time to first swing
stability Tr allows earlier-termination of the TDS, which can
significantly save the CPU time. Namely, when the TDS process till Tr , it can be stopped immediately without progressing
ahead.
E. Result Validation
To validate the robustness of the solution to the load model
uncertainty, a series of random validation scenarios are generated by Monte Carlo sampling given and of the load
model parameters, and each scenario is assessed for the three
contingencies. If the system is stable, the validation scenario
is defined as feasible, otherwise defined as infeasible. The
robustness degree is defined as follows:
=
Fi
100%
F
(28)
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
XU et al.: ROBUST TSCOPF WITH UNCERTAIN DYNAMIC LOADS
TABLE VII
ROBUSTNESS D EGREE
NIt
= (TOPF + TTSA K H) NIt +
(TTSA Nui )
i=1
(29)
where TOPF , TTSA respectively denote the CPU time for OPF
solution, and TSA for a contingency; K and H respectively
denote the number of contingencies and the testing scenarios
determined by TOAT; NIt denotes the total iteration number,
and Nui denotes the number of unstable scenarios in the i-th
iteration. The second term in (29) represents that an unstable
scenario requires an additional TDS to calculate the stability
margin sensitivity for deriving the stabilization constraint.
For the test in this paper, the total computation time is 34.3s
and 110.4s for single- and multi-contingency cases, respectively, which are much faster than conventional methods such
as discretizing and EA-based methods. Further discussions are
given below.
For contingency number K, it is 3 in this paper. In practice,
although a larger system has more contingencies, only unstable ones are to be included in the TSCOPF. In fact, since
the transient stability is considered at the planning stage, for
a well-planned system, the number of unstable contingencies
can be limited. Besides, since different contingencies may have
different probabilities of occurrence, it is sensible to include
only those more likely (or higher impact) ones in TSCOPF.
For testing scenario H, it is 8 in this paper. According to
OA tables [27], H can be very limited even for a large model.
For iteration number NIt , given the effectiveness of EEAC
stabilization constraint, this number is generally small. It is
2 in this paper which is consistent with other literature [6][9].
VIII. C ONCLUSION
Load dynamics have a substantial impact on system stability. While conventional TSCOPF models neglect this concern,
this paper shows that the dispatch results from such models
may not be able to sustain the stability if the dynamic load is
considered. A robust TSCOPF model is proposed to consider
the dynamic load and its parameter uncertainty. In terms of the
ability to handle dynamic load model uncertainties, the proposed model reports a robustness degree of 98.7% compared
against 16% for the conventional model. In terms of the economic operation, for single contingency case, the increased
cost by the proposed model over the conventional model is
0.89%, and for the multi-contingency case, the cost increment
is 2.98%, which are quite limited in the context of stability
enhancement. The computational efficiency of the proposed
approach is very high due to the use of only a small number of testing scenarios and the EEAC-based decomposition
structure.
On the other hand, existing TSCOPF methods are deterministic, i.e., the system is either stable or unstable. Besides,
the probabilistic nature of the contingencies is not modelled
which may result in a costly solution. In the future, it would be
interesting to investigate this problem based on a probabilistic
basis or risk-based criterion.
R EFERENCES
[1] D. Gan, R. J. Thomas, and R. D. Zimmerman, Stability-constrained
optimal power flow, IEEE Trans. Power Syst., vol. 15, no. 2,
pp. 535540, May 2000.
[2] L. Chen, Y. Taka, H. Okamoto, R. Tanabe, and A. Ono, Optimal operation solutions of power systems with transient stability constraints,
IEEE Trans. Circuits Syst. I, Fundam. Theory Appl., vol. 48, no. 3,
pp. 327339, Mar. 2001.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
10
Yan Xu (S10M13) received the B.E. and M.E. degrees from the South
China University of Technology, Guangzhou, China, in 2008 and 2011, respectively, and the Ph.D. degree from the University of Newcastle, Australia, in
2013. He is currently a University Postdoctoral Fellow with the School of
Electrical and Information Engineering, University of Sydney, Australia. He
was a Research Fellow at the Center for Intelligent Electricity Networks,
University of Newcastle, from 2013 to 2014. His research interests include
power system stability and control, power system planning, smart grid, and
intelligent system applications to power engineering.
Jin Ma (M06) received the B.S. and M.S. degrees from Zhejiang University,
Hangzhou, China, in 1997 and 2000, respectively, and the Ph.D. degree from
Tsinghua University, Beijing, China, in 2004, all in electrical engineering.
From 2004 to 2013, he was a Faculty Member of North China Electric
Power University. Since 2013, he has been with the School of Electrical
and Information Engineering, University of Sydney. His major research interests are load modeling, nonlinear control system, dynamic power system, and
power system economics. He is a Member of CIGRE W.G. C4.605 Modeling
and Aggregation of Loads in Flexible Power Networks, and a Corresponding
Member of CIGRE Joint Workgroup C4-C6/CIRED Modeling and Dynamic
Performance of Inverter Based Generation in Power System Transmission and
Distribution Studies. He is a registered Chartered Engineer in the U.K.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
XU et al.: ROBUST TSCOPF WITH UNCERTAIN DYNAMIC LOADS
Zhao Yang Dong (M99SM06) received the Ph.D. degree from the
University of Sydney, Australia, in 1999. He is currently a Professor
and the Head of the School of Electrical and Information Engineering,
University of Sydney. He was the Ausgrid Chair and the Director of
the Centre for Intelligent Electricity Networks, University of Newcastle,
Australia. He also held academic and industrial positions with the Hong Kong
Polytechnic University and Transend Networks (now TasNetworks), Australia.
His research interest includes smart grid, power system planning, power
system security, renewable energy systems, electricity market, load modeling, and computational intelligence and its application in power engineering.
He is an Editor of the IEEE T RANSACTIONS ON S MART G RID, the IEEE
T RANSACTIONS ON S USTAINABLE E NERGY, IEEE P OWER E NGINEERING
L ETTERS, and IET Renewable Power Generation.
11
David J. Hill (S72M76F93LF14) received the B.E. degree in electrical engineering and the B.Sc. degree in mathematics from the University of
Queensland, Australia, in 1972 and 1974, respectively, and the Ph.D. degree
in electrical engineering from the University of Newcastle, Australia, in 1976.
He is the Chair of Electrical Engineering with the Department of Electrical
and Electronic Engineering, University of Hong Kong. He is the Director
of the Centre for Electrical Energy Systems and the Program Coordinator
for the multiuniversity RGC Theme-Based Research Scheme Project on
Sustainable Power Delivery Structures for High Renewables. He is also a
part-time Professor with the Centre for Future Energy Networks, University
of Sydney, Australia. From 2005 to 2010, he was an Australian Research
Council Federation Fellow at Australian National University. Since 2006, he
has been the Theme Leader and the Deputy Director with the ARC Centre of
Excellence for Mathematics and Statistics of Complex Systems. He has held
various positions at the University of Sydney since 1994, including as the
Chair of Electrical Engineering until 2002 and again from 2010 to 2013, along
with an ARC Professorial Fellowship. He has also held academic and substantial visiting positions at the universities of Melbourne, California (Berkeley),
Newcastle (Australia), Lund (Sweden), Munich, and Hong Kong (City and
Polytechnic). He currently holds Honorary Professorships with the City
University of Hong Kong, South China University of Technology, Wuhan
University, and Northeastern University, China. From 1996 to 1999 and from
2001 to 2004, he served as the Head of the respective departments in Sydney
and Hong Kong. His general research interests are in control systems, complex networks, power systems and stability analysis, control and planning of
future energy networks, and basic stability and control questions for dynamic
networks.
Prof. Hill is a Fellow of the Society for Industrial and Applied
Mathematics, USA; the Australian Academy of Science; and the Australian
Academy of Technological Sciences and Engineering. He is also a Foreign
Member of the Royal Swedish Academy of Engineering Sciences.