Mixed Transient Stability Analysis Using AC and DC Models

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON POWER SYSTEMS 1

Mixed Transient Stability Analysis


Using AC and DC Models
Soobae Kim, Student Member, IEEE, and Thomas J. Overbye, Fellow, IEEE

Abstract—This paper presents a modified method for power is therefore an open challenge to develop faster algorithms for
system transient stability analysis, allowing the reduction of the transient stability analysis.
computational requirements while retaining important system Traditionally, dynamic equivalents have been used to reduce
dynamics. The approach formulates the power balance equations
depending on the area of interest. The detailed ac model, including
the computational burden. This approach retains an area of
full real and reactive power equations, is used for the area in interest for study, called the internal system and replaces the
which high accuracy is required, while the simpler dc model, in- external system—neighboring areas connected to the internal
cluding linear real power equation, is used for more remote areas system—with an electrical equivalent. The power system
where real power dynamics dominate. A high level of accuracy model is thus reduced in size. There are three general ap-
can be achieved using the ac model in the area of interest, and proaches used to develop these equivalents: modal, coherency
computational complexity can be reduced with the less detailed dc
model in the remainder of the system. In order to prevent the loss and measurement-based [4], [5]. While certainly useful in some
of simulation accuracy, this paper proposes a way to compensate situations, the equivalent approach has weaknesses in that the
line losses neglected using the dc model. Case studies with the equivalent must be updated whenever the operating point or
IEEE 118-bus system are provided to validate the performance of system topology is substantially changed [6] and the simulation
the proposed method. accuracy of the traditional equivalent methods is dependent on
Index Terms—Ac power flow, dc power flow, mixed approach, the type of disturbance (fault) [7].
transient stability simulation. One example of a disturbance that can usually be addressed
quite well by the use of a substantially reduced equivalent is
a bus fault and subsequent line opening actions. The effects
I. INTRODUCTION will be associated mostly with the nearby voltage magnitudes.

W ITH power systems being operated closer to their se- The frequency impact is small because such disturbances do
curity limits, it is critically important that utilities have not usually create a substantial imbalance between the electric
access to powerful computation tools that can perform rigorous load and generation. The reduced system with the buffer zone
analysis of transient behavior [1]. However, due to interconnec- around the fault and with an appropriate dynamic equivalent at
tions, modern power grids have become increasingly complex. the boundary buses can give accurate responses [4]–[7].
They are comprised of millions of loads and generators, and tied On the other hand, when generator or load outages happen,
together by hundreds of thousands of miles of transmission and the situation is different in that the frequency perturbation will
distribution wires with a myriad of control devices. As a result, affect the entire interconnect. The use of the traditional equiv-
a power system transient stability simulation for dynamic secu- alent could be inappropriate to represent system responses cor-
rity assessment is required to solve a huge number of nonlinear rectly. One issue would be if the dynamic equivalent does not
differential-algebraic equations (DAE) and needs to be run for have enough reserve generator governor response to make up
many contingency cases [2]. Computational demands in terms for the generator loss or the load increase. The reduced system
of storage and simulation time are significant. cannot maintain additional real power injections from the al-
Dramatic advancements in microprocessor technology have ready removed external system. Therefore, an enhanced tran-
made substantial improvements in transient stability compu- sient stability simulation method is required both to reduce the
tational speed. Nevertheless, when simulating larger systems computational requirements and to achieve better simulation ac-
(i.e., with tens of thousands of buses), computation time with curacy than the pure equivalent approach.
commercial transient stability packages is still limiting [3]. It The authors presented a hybrid power-flow approach in [8]
that combines full ac models with a less detailed dc power flow
models. Here this approach is extended to transient stability
Manuscript received May 14, 2014; revised August 02, 2014, November 09,
2014, and February 20, 2015; accepted March 22, 2015. This work was support analysis. The idea originates from the fact that although reactive
in part by Korea Electric Power Corporation (KEPCO), Seoul, Korea and in power does not travel well in power systems because the trans-
part by the U.S. Department of Energy through award number DE-OE0000097. mission network is mostly inductive, the real power does. Both
Paper no. TPWRS-00655-2014.
S. Kim is with Korea Electric Power Research Institute (KEPRI), Daejeon,
real and reactive power dynamics are dominant in the internal
Korea (e-mail: soobkim@kepco.co.kr). system where any type of fault occurs. For the external system
T. J. Overbye is with the University of Illinois Urbana-Champaign, Urbana, which is impacted by a fault in the internal, real power dy-
IL 61801 USA (e-mail: overbye@illinois.edu).
Color versions of one or more of the figures in this paper are available online
namics play a more important role. The significant dynamics of
at http://ieeexplore.ieee.org. both areas can be preserved by formulating different power-flow
Digital Object Identifier 10.1109/TPWRS.2015.2416355 equations at each region. The detailed ac model, including full

0885-8950 © 2015 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/
redistribution requires IEEE permission. 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 IEEE TRANSACTIONS ON POWER SYSTEMS

with the variables showing algebraic variables such as the net-


work bus voltage and angle. The dimension of vector equals
twice the number of buses.
Fig. 1 shows a flow chart of a typical transient program which
solves the above set of mathematical equations. The first step is
to find the initial conditions. Those values are determined from
power-flow solutions and by setting all the differential operators
to zero. Then the time-domain solutions of the equation sets are
obtained through either explicit or implicit numerical integra-
tion formulas applied to each differential operator. This gives a
set of algebraic equations which is then solved by an iterative
method, such as the Newton method. When a disturbance occurs
at a certain time, dynamic or algebraic equations are changed ac-
cordingly. This process repeats until the simulation time reaches
its end time.

Notations
is the bus voltage magnitude at bus in per-unit and is
the voltage phase angle at bus . and are the net injected
real and reactive powers at bus . and are real and
imaginary part of the admittance matrix elements corresponding
to a line between bus and bus .
Fig. 1. Flowchart of a conventional transient simulation.
B. AC Power-Flow Model
In the ac model approach the power balance equations are
real and reactive power equations, is used for the area of in- represented by the (3) and (4). These are associated with the
terest, while the simpler dc model, including only linear real algebraic (2) for transient stability analysis:
power equation, is used for more remote areas. In order to pre-
vent the loss of simulation accuracy, this paper presents a way
to compensate for the line losses neglected using the dc model.
Therefore, the approach can reduce the computational require- (3)
ments, and still achieve a high level of simulation accuracy in
the area of interest.
This paper is organized as follows. Section II presents a
brief analytic basis for the power system transient simulation (4)
and power-flow models. The proposed approach is presented
in Section III. Section IV illustrates simulation results with the The ac model is formulated with nonlinear equations. An iter-
IEEE 118-bus system. The conclusion is presented in Section V. ative algorithm is required to solve those equations. Thus the
ac model is computationally expensive, especially for transient
II. POWER SYSTEM TRANSIENT SIMULATION stability analysis which must find solution for every time step.

The basics of power system transient stability simulation and C. DC Power-Flow Model
power-flow models are briefly explained in this section.
The dc model simplifies the ac model by making several as-
A. Transient Stability Simulation sumptions. These are 1) the reactive power balance equations
are completely ignored, 2) all voltage magnitudes are one per-
A transient stability simulation involves solving a set of dif- unit, 3) voltage angle differences are small enough and 4) line
ferential equations and an accompanying set of algebraic equa- losses are ignored. Hence the dc model reduces the size of the
tions shown in (1) and (2) [9], [10]. These are called differential power-flow problem and changes nonlinear problems to a set of
algebraic equations and are of the following form: linear equations. It is represented with the following equation:

(1)
(5)
(2)

Equation (1) represents the power system dynamics, with the Even though the assumptions reduce simulation accuracy, the dc
variables showing the dynamic state variables such as generator model has been widely used because of its computational advan-
rotor angle and speed. The dimension of vector is dependent tages over the ac model. It can run many times faster because
on the modeling detail and the size of the power systems. Equa- of following reasons [11]. The dc model reduces the size of the
tion (2) represents the stator and network algebraic equations, equation set to about half of the ac model, because the reactive
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

KIM AND OVERBYE: MIXED TRANSIENT STABILITY ANALYSIS USING AC AND DC MODELS 3

power equations are ignored. The model is linear. Thus no iter- for larger power systems the cumulative sum of these individ-
ations are required and only one factorization of the matrix is ually small errors might not be negligible and hence total real
needed no matter what operation conditions are considered. power balances might not be preserved [12]. Therefore, the dc
model cannot capture the wide spread real power impact accu-
III. PROPOSED METHOD rately from the internal system. The proposed approach expects
that the dimension of the external system would be much larger
The proposed method formulates the algebraic equations
than the internal system. It is thus necessary to compensate for
for transient stability simulation by combining the ac with
these losses to maintain the real power balances in the external
the dc power-flow models, depending on the area of interest.
system. The proposed approach converts the losses into shunt
The detailed ac model is used for the internal system where
resistances at each external bus. The loss can be obtained from
both real and reactive power dynamics are dominant. The
(3) and has the following form:
simpler dc model is used for the external system where real
power dynamics dominate. The proposed method can then
represent both the localized effects related to real and reactive (8)
power-flows using the ac model for the internal system, and the
entire effects mostly related to real power-flows using the dc The loss modeling is based on initial ac power-flow solutions
model for the external system. Thus, the proposed method aims and it is performed before the transient simulation loop. The
to achieve a high level of simulation accuracy with the full ac power balance equations for the external system have the fol-
model in the internal system and to speed up transient stability lowing form:
simulation with the linearized dc model.

A. Boundary Bus Modeling (9)


In the proposed approach, the more approximate dc model
is used in the external system. The external system makes an
impact on the internal system through the boundary buses that C. Proposed Approach
comprise the boundary between the two systems. Careful con- In the proposed method, the set of algebraic equations de-
siderations should be made at the boundary buses to minimize pending on system division can be summarized as follows.
the error propagation from the external system. A high level of • Internal system
accuracy can then be achieved in the internal system where ac- — Nonlinear and equations using (3) and (4)
curate solutions are required. Thus, the detailed ac model is used • Boundary buses
at the boundary buses and algebraic formulations are shown — Nonlinear and equations using (6) and (7)
in (6) and (7). Both bus voltage magnitude and angle at the • External system
boundary buses can then be calculated for every time step, while — Linearized equation with loss addition using (9)
bus voltage magnitudes in the external system are not updated Fig. 2 shows a flow chart of the proposed transient stability
since reactive power equations in the external system are ne- approach. Compared to the conventional method shown in
glected. Instead, the initial values obtained from the power-flow Fig. 1, most of procedures are same, but the presented approach
solutions are used: adds the loss equation for the external system and builds new
power-flow equations, depending on the area of interest.

D. Computational Benefits
As shown in Figs. 1 and 2, transient simulations need to solve
a set of algebraic equations for every time step. The use of the
dc model in the external system reduces the dimension of the
(6) equations and thus the computational requirements can be re-
duced greatly. The benefits are dependent on the ratio of the
number of internal buses to external buses. More computational
reductions can be achieved when the dc area is much larger than
the ac area. Analytic details about computational benefits de-
pending on the ratio can be found in [8]. In addition, the dc
model does not consider voltage magnitude variation and reac-
(7) tive power-flows. Therefore, additional speed up can be made
by neglecting the dynamic equations related to voltage magni-
where is the initial voltage magnitude at bus obtained tude or reactive powers in the external system. These dynamics
from power-flow solutions. are associated mostly with the exciter models.

B. Loss Compensation in the External System IV. CASE STUDY


The dc model applied to the external system does not take into The proposed approach is implemented with Matlab. The per-
account the losses, which has the potential to introduce a usu- formance of the proposed method is validated with the IEEE
ally relatively small error in the branch MW flows. However, 118-bus system shown in Fig. 6 [13]. The case contains 186
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 IEEE TRANSACTIONS ON POWER SYSTEMS

Fig. 3. Block diagram of TGOV1 [15].

Fig. 4. Relative rotor angle of Generator 31 w.r.t. Generator 12.

Fig. 2. Flowchart of the proposed transient stability analysis.

branches, 19 generators, and 99 loads. System dynamics are rep-


resented with classical machine model in (10) [14] and simple
TGOV1 (turbine-governor) model of which block diagram is
depicted in Fig. 3 [15]. Constant impedance loads are modeled.
Detailed model parameters for the case study are presented in Fig. 5. Simulation comparisons of bus voltage magnitude.
the Appendix:

(10a) examine a more localized disturbance. The steady-state oper-


ating points are then disturbed and dynamic responses are com-
(10b) pared.
Simulation comparisons are made between the full system
where model, a standard equivalent model, and the proposed mixed
ac-dc approach. Dynamic equivalent circuit including dynamic
rotor angle position of machine ;
parameters is obtained from PowerWorld simulator [16]. And
rotor angle velocity of machine ; the difference is measured using RMSE (Root Mean Square
inertia constant of machine ; Error) over the simulation period [17]:
mechanical torque of machine ;
electrical torque of machine ; (11)
damping coefficient of machine .
The mixed transient method is tested with the system divi- where is the number of simulation time steps and is time-
sion provided in Table I. Overall there are 45 buses are in the series data compared.
internal system, four boundary buses, and 69 buses in the ex-
ternal system. In this example two different types of disturbance A. Load Addition(100 MW) at Bus 3
are applied: 1) a load addition to cause a system-wide frequency For the first disturbance, the load (100 MW) at bus 3 is of-
variation, and 2) a balanced three phase bus fault to ground to fline initially and is changed to online at 0.1 s. The load connec-
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

KIM AND OVERBYE: MIXED TRANSIENT STABILITY ANALYSIS USING AC AND DC MODELS 5

Fig. 6. One-line diagram of the IEEE 118-bus system [13].

TABLE I
DETAILS OF THE SYSTEM DIVISION

Fig. 8. RMSE of bus voltage magnitude.

model method than the dynamic equivalent. And the angle


differences in the external system from the mixed method are
similar with those in the internal system. It can be understood
that the dc model shows a quite good performance to represent
Fig. 7. RMSE of bus voltage angle. the real power dynamics.

B. Bus to Ground Fault at Bus 3


tion introduces real power imbalance in the system and the in- The second comparison is with the simulation of a bus to
dividual generators via governor control supply additional real ground fault. A three-phase bus to ground fault is applied at bus
power to restore the system frequency. 3 at 0.1 s and it is cleared at 0.15 s. Unlike the previous load con-
Fig. 4 shows the relative rotor angle of Gen 31 with respect nection, the fault does not introduce any significant real power
to Gen 12, while the bus voltage magnitudes for the internal imbalance between the net generation and load, causing real and
system are presented in Fig. 5. The responses from the proposed reactive power variations mostly just in the buses close to the
method show overall good agreement with those from the full fault location. Figs. 9 and 10 show the simulation comparisons
transient stability approach. The RMSE in (11) over the period with Gen 10's rotor angle and bus voltage magnitudes, respec-
from 0 to 5 s are shown in Figs. 7 and 8. The reduced system tively.
from the equivalent method does not produce simulation results With the bus to ground fault, the proposed method shows vir-
in the external area and thus the RMSE for the external buses tually identical responses to the full model method. The fault
are expressed with zero. makes an impact locally, mostly on the internal system. This is
As shown in the Fig. 7, the mixed method shows a smaller because the impact of the bus to ground fault is dependent on the
RMSE compared to the dynamic equivalent and it means that electrical distance between the bus fault location and the neigh-
the proposed method provides better matching with the full boring buses. When the fault occurs far from a bus, its impact
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 IEEE TRANSACTIONS ON POWER SYSTEMS

Fig. 9. Relative rotor angle of Generator 10 w.r.t. Generator 12.


Fig. 12. Bus voltage magnitude RMSE of all buses.

TABLE II
COMPUTATION TIME OF 118-BUS CASE

Fig. 10. Simulation comparisons of bus voltage magnitude.

Fig. 13. Average of bus angle RMSE with varying load amount at bus 3.

equivalent and this is not included in the time computation. To


makes worse is in that a new equivalent must be derived when-
ever operating point or system topology is altered. The mixed
Fig. 11. Bus voltage angle RMSE of all buses. approach needs no extra time. For the given system configura-
tion, the ratio of the internal to external buses is about 1 to 1.4.
More computational benefits can be attained with the case where
on the bus is lessened. Therefore, the deviation of the system the external system is much bigger than the internal system.
states in the external system is very small and the use of the Fig. 13 shows accuracy comparison between the mixed and
more approximate dc model does not introduce a big difference the equivalent methods when the amount of the load addition
in the internal system. Furthermore, as shown in Fig. 12, the at bus 3 is varied from 50 MW to 500 MW. Average of RMSE
voltage magnitude errors in the external system are quite small angle values for the buses in the internal system is computed.
even though the proposed method deals with the fixed voltage The equivalent approach shows bigger discrepancy in responses
magnitude. As shown in Figs. 11 and 12, the simulation results as the load amount increases. Conversely, the mixed approach
of both angle and voltage magnitude validate that the proposed provides quite good accuracy whatever loads are added. The
method accomplishes better accuracy than the dynamic equiva- equivalent system does not maintain the generators in the ex-
lent method. ternal area and thus additional injections from those generators
cannot be represented with the equivalent approach.
C. Computational Benefits and Accuracy There is a speed and accuracy tradeoff in the proposed
The computation time with different simulation approaches method, similar to what occurs between the ac and dc power
and the 118-bus system is provided in Table II. The equiva- flow models. Therefore, the proposed method can provide
lent method gives faster solution than the proposed method. quite accurate responses in the area of interest, less so for the
However, additional computational time is needed to obtain the external system. The simplification in the external system leads
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

KIM AND OVERBYE: MIXED TRANSIENT STABILITY ANALYSIS USING AC AND DC MODELS 7

TABLE III TABLE IV


MACHINE PARAMETERS FOR THE IEEE 118-BUS SYSTEM. TGOV1 MODEL PARAMETERS. (ALL GENERATORS HAVE SAME PARAMETERS)
(MACHINE BASE : 100 MVA)

REFERENCES
[1] K. Morison, L. Wang, and P. Kundur, “Power system security assess-
ment,” IEEE Power Energy Mag., vol. 2, pp. 30–39, 2004.
[2] P. Kundur, N. J. Balu, and M. G. Lauby, Power System Stability and
Control. New York, NY, USA: McGraw-Hill, 1994.
[3] K. Birman, Computational Needs for the Next Generation Electric Grid
Proceedings, Lawrence Berkley Nat, Lab,, 2012, LBNL-5105E.
[4] U. D. Annakkage, N. K. C. Nair, Y. Liang, A. M. Gole, V. Dinavahi, B.
Gustavsen, T. Noda, H. Ghasemi, A. Monti, M. Matar, R. Iravani, and
J. A. Martinez, “Dynamic system equivalents: A survey of available
techniques,” IEEE Trans. Power Del., vol. 27, pp. 411–420, 2012.
[5] J. H. Chow, Power System Coherency and Model Reduction. New
to reduced computational demands. In addition, the accuracy of York, NY, USA: Springer, 2013.
the proposed method can be dependent on particular case, and [6] F. Ma and V. Vittal, “Right-sized power system dynamic equivalents
the relative size of the system modeled with the dc approxima- for power system operation,” IEEE Trans. Power Syst., vol. 26, pp.
1998–2005, 2011.
tion. For example, cases with high values of transmission line [7] D. Fabozzi, A. S. Chieh, B. Haut, and T. Van Cutsem, “Accelerated
resistance will incur more error since the dc model neglects the and localized Newton schemes for faster dynamic simulation of large
resistance. Also, converter connected dynamics, such as those power systems,” IEEE Trans. Power Syst., to be published.
associated with wind turbines, require accurate voltage mag- [8] S. Kim and T. J. Overbye, “Mixed power flow analysis using AC and
DC models,” IET Gener., Transm., Distrib., vol. 6, pp. 1053–1059,
nitude and angle solutions to maintain the correct simulation 2012.
responses. Hence special consideration may be needed for their [9] B. Stott, “Power system dynamic response calculations,” Proc. IEEE,
use in the proposed mixed approach, something that was not vol. 67, pp. 219–241, 1979.
[10] J. M. Undrill, “Structure in the computation of power-system nonlinear
explored in this paper. Finally, an excessively large portion of dynamical response,” IEEE Trans. Power App. Syst., vol. PAS-88, pp.
the dc approximation may also degrade the accuracy. 1–6, 1969.
[11] T. J. Overbye, X. Cheng, and Y. Sun, “A comparison of the AC and DC
power flow models for LMP calculations,” in Proc. 37th Annu. Hawaii
V. CONCLUSION Int. Conf. System Sciences, 2004, p. 9.
[12] B. Stott, J. Jardim, and O. Alsac, “DC power flow revisited,” IEEE
An alternative method that speeds up transient stability sim- Trans. Power Syst., vol. 24, pp. 1290–1300, 2009.
[13] Power Systems Test Case Archive, University of Washington [Online].
ulations while still maintaining a high level of simulation accu- Available: http://www.ee.washington.edu/research/pstca/
racy has been proposed. It is based on the fact that the reactive [14] P. W. Sauer and M. A. Pai, Power System Dynamics and Stability.
powers tend to be more localized compared to the real powers. Upper Saddle River, NJ, USA: Prentice Hall, 1998.
[15] “Dynamic Models for Turbine-Governors in Power System Studies,”
In the area of interest, where the disturbances are applied and
IEEE PES Technical Report, Power System Dynamic Performance
accurate simulation outcomes are required, both real and reac- Committee, 2013.
tive power equations should be considered to represent power [16] Powerworld Corporation [Online]. Available: http://www.power-
system dynamics correctly, while real power equations would be world.com/
[17] S.-K. Joo, C.-C. Liu, L. E. Jones, and J.-W. Choe, “Coherency and ag-
enough in the more remote areas where real power dynamics are gregation techniques incorporating rotor and voltage dynamics,” IEEE
dominant. The proposed approach thus formulates power-flow Trans. Power Syst., vol. 19, pp. 1068–1075, 2004.
equations for transient stability simulation by combining the full
ac and dc models. A detailed ac model, including nonlinear real
and reactive power equations, is used in the internal system, Soobae Kim (S'10) received the B.S degree in electrical and computer engi-
while a simpler dc model is used in the external system. The real neering from Kyungpook National University, Daegu, Korea, in 2002, the M.S.
power losses ignored by the standard dc model are compensated degree from Seoul National University, Seoul, Korea, in 2004, and the Ph.D.
degree from the University of Illinois at Urbana-Champaign, Urbana, IL, USA,
for in the external system by using the initial power-flow solu- in 2014.
tions. The test simulations performed with IEEE 118-bus system Since 2004, he has been with Korea Electric Power Research Institute
have confirmed that the proposed method achieves faster and (KEPRI), Daejeon, Korea. His special fields of interest are power system
analysis, dynamic system equivalent, and model reduction.
accurate solution. More computational benefits can be achieved
with larger power system cases which have higher dimension-
ality of the external system compared to the internal system. It
is expected that the proposed method would be a promising so- Thomas J. Overbye (S'87–M'92–SM'96–F'05) was born in Milwaukee, WI,
USA. He received the B.S., M.S., and Ph.D. degrees in electrical engineering
lution for advanced transient stability study. from the University of Wisconsin, Madison, Madison, WI, USA.
He was with Madison Gas and Electric Company, Madison, WI, USA, from
1983 to 1991. He is currently the Fox Family Professor of Electrical and Com-
APPENDIX puter Engineering at the University of Illinois at Urbana-Champaign, Urbana,
IL, USA. His current research interests include power system visualization,
Table III lists the machine parameters for the IEEE 118-bus power system dynamics, power system cyber security, and power system ge-
system. Table IV lists the TGOV1 model parameters. omagnetic disturbance analysis.

You might also like