Professional Documents
Culture Documents
On The Accuracy of The Online Static Security Assessment Under Different Models: Assessment and Basis
On The Accuracy of The Online Static Security Assessment Under Different Models: Assessment and Basis
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
1
This work was partially supported by the National Science Foundation of Meng-yu Ruan is with the School of Electrical and Information Engineering,
China (51577092) and the National Science Foundation (USA) under award Tianjin University, Tianjin, China (e-mail: alice_ruanmy@tju.edu.cn). Hsiao-
#1508986. Dong Chiang is with the School of Electrical and Computer Engineering,
Cornell University, Ithaca, NY, USA (e-mail: hc63@cornell.edu).
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
2
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
3
a Missed Alarm (MA) (i.e., an insecure contingency is classified exponential recovery loads, over-excitation limiters (OXLs),
as secure). Conversely, a False Alarm (FA) indicates that a and turbine governors (TGs), and (4) describes the long-term
secure contingency is classified as insecure. We point out that discrete events like load tap changers (LTCs). zc and zd are the
there is another situation in which both the PF model and the continuous and discrete long-term state variables, respectively.
TD model classify a contingency as insecure, while the PF τ=εt and 1/ε are the maximum time constants among devices.
model may miss some violations or falsely give out some The QSS model is derived from time-scale decomposition
violations. We term this situation as a Missed/False Violation and aims to offer an appropriate tradeoff between speed and
(MFV), the third type of misclassification. To provide detailed accuracy in simulating trajectories. If the short-term dynamics
insight into misclassification, Table II presents detailed results are stable and settle down infinitely fast in the long-term time
on the 14-bus system during light and heavy loading conditions. scale, then the corresponding equilibrium equations take the
place of (2) and we have the following QSS model [21]:
TABLE II
DETAILED ASSESSMENT RESULTS IN THE IEEE 14-BUS SYSTEM 0 = g ( x, y, z c , z d ) (5)
DURING LIGHT AND HEAVY LOADING CONDITIONS 0 = f ( x, y, z c , z d ) (6)
dz c
Line
Light loading Heavy loading = h c ( x, y, z c , z d ) (7)
outage PF TD PF TD d
model model model model z d (k + 1) = h d ( x, y, z c , z d (k )) . (8)
V12 V11-14 V11-14
S9-10 On the other hand, the PF model only focuses on the power
violation violation violation
5-6 overload
S9-10 S9-10/ S9-14 S9-10/ S9-14 flow balance under static network constraints, described by the
(MFV)
overload overload overload
power flow equation:
V14 violation
9-14 secure secure
(FA)
secure 0 = g ( y) . (9)
V4/V5 The PF model captures the power flow balance, ignoring
secure violation
1-5 secure secure dynamic behaviors and the internal constraints of short-term
(MA) S2-5
overload and long-term state variables. Overlooking the countereffects
secure S1-5 between dynamic devices makes the PF model too simplified to
2-4 secure secure
(MA) overload
S1-5/S2-4/S3-4
reflect the real power system steady-state after a contingency.
S1-5/S2-4 Therefore, considerable differences between the solutions of
2-3 secure secure Overload
overload
(MFV) the PF model and those of the TD model can exist.
secure S2-4
4-5 secure secure
(MA) overload B. Analytical Basis for QSS-based SSA
It was shown in [16] that the QSS model provides an
III. QUASI STEADY-STATE MODEL-BASED SSA accurate approximation of the TD model in terms of the ω-limit
In this section, the QSS model and the TD model are set under a minor set of assumptions described below:
reviewed. We then present theoretical foundations for the S1. Neither the TD model nor the QSS model meets the
accuracy of the QSS model in performing SSA. singularity points.
S2. The trajectories of the TD model, the QSS model, and
A. Power System Models
the transient stability models with specified initial conditions
The general dynamic power system model for stability exist and are unique. Additionally, Dzc is compact.
analysis can be described as a set of nonlinear differential S3. The equilibrium point of the transient stability model is
algebraic equations (DAE) with continuous and discreet continuous in zc when zd are fixed as parameters.
variables:
0 = g ( x, y, z c , z d ) (1) Definition: ω-Limit Set: (t , x) is the system trajectory
dx starting at initial state x at time t=0. A point s is said to be the
= f ( x, y, z c , z d ) (2)
dt ω-limit point of x if, corresponding to each ε>0 and T>0, there
dz c
= h c ( x, y, z c , z d ) (3) is a t>T with the property that (t , x) − s . Equivalently,
d
there is a sequence ti in , for ti →+∞, with the property that
z d (k + 1) = h d ( x, y, z c , z d (k )) (4)
s = lim (t i , x) . The set of all ω-limit points for x is defined as
where g, f, and hc are continuous functions and hd is a discrete t i →+
function; g in (1) represents system network algebraic functions its ω-limit set [16].
and describes the static behaviors of passive devices; f in (2)
describes the short-term dynamic functions of synchronous Theorem 1 (ω-limit set relation) [16]:
machines, their automatic voltage regulators (AVRs), and the If positive constants r and ε0 exist such that assumptions S1-
interconnecting transmission network, together with induction S3 and the following conditions are satisfied for all [τ,zc,zd,x,y,ε,]
and synchronous motor loads as well as other devices such as ∈[τ0,+∞]×Ur×[0,ε0]:
the HVDC converter and SVC; and vectors x,y are the
a) The trajectory of the QSS model (5)-(8) moves along the
corresponding short-term variables and the algebraic variables.
stable component of the constraint manifold Γs;
Equation (3) describes long-term dynamics, including
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
4
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
5
already stated the assumptions that must be satisfied, so the 14-bus system: The IEEE 14-bus system has 20 lines with
QSS model cannot reproduce the transient instability or short- 1 LTC placed between bus4 and bus9. We increase the real
term and long-term dynamic interaction instability. While the power demand at buses 9, 12, and 13 uniformly in the base case.
SSA is focused on the post-contingency steady-state problem For the TD and the QSS models, each of the 5 generators is
and does not involve the above-mentioned problems, this paper equipped with an AVR and an OXL. Generators 1 and 2 are
suggests that the QSS model be used for Online-SSA in energy controlled by TGs. Three exponential recovery loads are
control centers because of its solid theoretical foundation, its located at bus9, bus10, and bus14, respectively. The LTC has
high accuracy, and its computational efficiency. an initial time delay of 30 seconds and a fixed tapping time
delay of 10 seconds.
IV. DETAILED ANALYSIS
30-bus system: The IEEE 30-bus system has 41 lines.
To evaluate the accuracy of both PF-based SSA and QSS-
Voltages of PV node 11 and 13 are set as 1.045p.u. We place 2
based SSA, we define two indices—MaxV_diffPF and
LTCs between bus4-bus12 and bus28-bus27. For the TD and
MaxS_diffPF— to capture the differences in voltage and apparent
the QSS models, each of the 6 generators is controlled by an
power between the PF model and the TD model for each
AVR and an OXL. Generators 1 and 2 are controlled by TGs.
contingency:
The LTCs have the initial time delay as 30s and the fixed
Vi PF − ViTD tapping delay as 10s.
MaxV _ diff PF = max (14)
bus num i =1,2,... Vi base To classify the operating conditions into light and heavy
S PF − S TD loading conditions, we employ the metric of a PF-based load
MaxS _ diff PF =
j j
max . (15) margin to steady-state stability limit to classify the degree of
line num j =1,2,... S themal
j
limit
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
6
0%
Line 9-14 Line 2-3 Line 1-5 Line 4-5 Line 2-4 Line 5-6
It is observed that over 85% of the voltages, the voltage
(b) MaxS_diffPF (%) differences (dv) are usually less than 1e-4. Regarding the
Fig. 4. Histogram of maximal voltage and apparent power differences apparent power differences (ds), they are slightly larger, but all
between the PF model and the benchmark with different loading of them are still less than 1%. It is important to point out that,
conditions in the IEEE 14-bus system. from Table III, the QSS-based SSA has the same assessment
results as the benchmark; confirming the accuracy of QSS-based
We take the 14-bus system as an example for detailed
SSA at different loading conditions. These assessment results of
analysis. Fig. 4 presents a histogram of maximal voltage and
high accuracy agree with the theoretical prediction.
apparent power differences between the PF model and the TD
model of several critical contingencies at several loading B. The IEEE 145-bus System
conditions. The PF-based SSA is acceptable during light loading In this modified test system, each generator is equipped with
conditions; however, during heavy loading conditions, there are an AVR, a TG, and a power system stabilizer. Generators 1-6
large differences in both voltage and apparent power leading to are controlled by OXLs. We add 3 LTCs to the system with an
severe misclassifications made by the PF-based SSA (as shown initial time delay of 30 seconds and the fixed tapping delay of
in Table V). For instance, with Line 1-5 outage, the PF model 10 seconds.
classifies it as secure (during heavy loading conditions) while
this contingency is classified as insecure by the TD model in
which V4 and V5 are already over limit and branch 2-5 is
overloaded (as shown in Table II).
(a)
(a) Voltage differences
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
7
V. PHYSICAL EXPLANATION
In this section, we present a physical explanation for the
different behaviors of the PF model and the benchmark during
heavy loading conditions.
Since misclassifications become more frequent and severe
with increasing loading conditions, we take the IEEE 14-bus
system under a heavy loading condition λ=1.1 and the base case
Fig. 7. A 3-D cone bar chart of apparent power differences for most with a line 1-5 outage as an example. Numerical simulation
critical contingencies by using the PF model and the QSS model during results show that PF-based post-contingency voltages are
different loading conditions in the IEEE 145-bus system. higher than that of the benchmark, while QSS-based post-
The voltage differences and apparent power differences, as contingency voltages are the same as the benchmark. The PF
compared with the steady-state of the TD model, for most model assesses this contingency as secure, while in fact it is
critical contingencies are shown in Figs. 6 and 7. From these insecure because bus4 and bus5 are all under-voltage.
results, we have the following observations:
(i) The steady-state calculated according to the PF model is
quite different from the benchmark. The differences show up in
both the voltage and apparent power, while the steady-state
calculated according to the QSS model is very close to the
benchmark.
(ii) For the PF model, voltage differences and apparent
power differences grow rapidly as loading conditions increase,
which contributes to more misclassifications during heavy
loading conditions; especially for the unacceptable MA-type
misclassification.
(iii) For the QSS model, though voltage differences and
apparent power differences slowly grow as loading conditions
increase, they stay in a small range. This is why the QSS model Fig. 8. Voltage profiles under three different models (the time-domain
provides the same SSA results as the benchmark. model, QSS model, and power flow model) of the IEEE 14-bus system
under a heavy loading condition λ=1.1 with line 1-5 outage. (The red
C. CPU Time dashed lines define the upper and lower voltage limits, respectively.)
The computation time and calculation ratio, calculated by First, the assumption of PV nodes may contribute to the
the PF model and the QSS model and compared with the differences. In the PF model, a generator bus maintains a
benchmark, are presented in Table VI and were implemented constant voltage magnitude until it reaches its reactive power
on an Inter Core i7-4790 @3.60GHz processor with 16GB limit. However, in reality, there are AVRs connected to
RAM. We observe that the QSS model takes only about 5%~10% generators for primary voltage regulation. Due to the difference
of the computation load of the TD model. Moreover, with regulation characteristics of the AVRs’ model, the voltages of
recent progress in accessible high-performance computing [13], generator buses decrease as the loading conditions increase. In
[14], the separate contingencies can be assigned to more than the PF model, bus3 is set as a PV node without a Q limit. With
one processor to reduce the required computation time. sufficient reactive power capability, bus3 maintains its constant
TABLE VI voltage at 1. 01p.u after the contingency. In the TD model, after
CPU TIME IN SECONDS AND COMPUTATION LOAD OF THE FOUR EXAMPLES the line outage, part of the load supported by the generator at
Ratio of calculation bus1 is transferred to the generator at bus2 and bus3, which is
Processing time(s)
Case Loading (%) equivalent to increasing the load current of the generator at bus3.
PF QSS TD PF QSS TD And, with the increased load current, the external characteristic
Light 1.5 5.8 115.3 1.30 5.03 100 of bus3 decreases (V3↓), caused by the degaussing of the
6-bus
system
Base case 1.6 6.1 122.1 1.31 5.00 100 armature reaction and the leakage impedance voltage drop.
Heavy 1.8 6.6 122.7 1.47 5.38 100 Then the field current If3 increases, trying to maintain the
Light 3.0 21.8 248.1 1.21 8.79 100 voltage, and without Q limits in bus3 (syn3), the OXL3 is
14-bus
system
Base case 3.3 26.4 252.4 1.31 10.46 100 inactive and the reference AVR3 voltage stays constant (see Fig.
Heavy 4.6 35.6 258.6 1.78 13.77 100 9). Therefore, due to the armature reaction of the synchronous
Light 5.6 32.0 473.1 1.18 6.76 100
30-bus generator and the differential regulation characteristic of the
Base case 6.3 41.1 477.1 1.17 8.61 100
system AVR, its voltage decreases and settles at 0. 986p.u in the post-
Heavy 9.2 61.6 488.5 1.88 12.62 100
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
8
contingency steady state. As a result, the voltages of the TD original level, while bus4 is still locked to its limit. However, in
model and the QSS model are lower than those of the PF model the TD and QSS models, after the transient period, when the
(see Fig. 8). slow dynamic device, the OXL, reaches its delay time and If is
0.07 vOxl 2 vOxl 3 vOxl 4 vOxl 5
larger than its upper limit, OXLs are activated, which limits the
0.06
reactive power output of synchronous generators in the system,
0.05
while the LTC tries to restore the secondary voltage (bus9) by
continuously lowering its tap ratio. This process will increase
Voxl (p.u)
0.04
0.03 the demand for reactive power in the system. Because of the
0.02 countereffect between the LTC and the OXL, the LTC fails to
0.01 restore bus9 to its original voltage level and the primary voltage
0
0 50 100 150 200
(bus4) is over limit, thus resulting in an increase in the
time (s) differences.
1.2 0.9
vrefExc 2 vrefExc 3 vrefExc 4 vrefExc 5 qSyn 2 qSyn 3 qSyn 4 qSyn 5
0.8
1.18 VI. CONCLUSION
Reactive Power Output (p.u)
0.7
1.16 Using the steady-state of time-domain (TD) simulation as
0.6
the benchmark, this paper examines the accuracy of SSA using
Vref (p.u)
1.14
0.5 the power flow (PF) model and the Quasi-static state (QSS)
1.12
0.4 model, respectively. As the loading conditions increase, it is
1.1
0.3 observed (from our numerical evaluation) that
1.08 0.2
misclassifications in SSA become more pronounced and can be
0 50 100 150 200 0 50 100 150 200
time (s) time (s) serious in the PF-based SSA, while misclassifications in QSS-
Fig. 9. Changes of dynamic variables If, Voxl, Vref and reactive power based SSA do not occur in the four test systems under both light
outputs of synchronous generators in the TD model in the IEEE 14- and heavy loading conditions. This observation persists when
bus system.
load variations and generation re-dispatch patterns are taken
Second, the treatment of reactive power limits is too into account during increases in loading conditions. Physical
simplified in the PF model. In practice, the maximum reactive explanations for misclassifications made by the PF-based SSA
power output of a generator is related to its field current If, are provided. First, the assumption of PV nodes in the PF model
which is proportional to the field voltage Vf. The OXL contributes to the differences. Second, the treatment of reactive
connected to the corresponding AVR is modeled as a pure power limits in the PF model is too simplified. Third, ignoring
integrator with anti-windup hard limits and provides an the interaction between LTC regulation (discrete characteristics)
additional signal Voxl to modify the reference voltage Vref of the and other dynamic components may further increase the
AVR [23]. As Fig. 9 shows, after a time delay, when If of the differences.
generator is greater than its limit, the OXL is activated and It is a common practice to perform PF-based SSA in modern
outputs the Voxl to modify the reference AVR voltage Vref, which energy control centers worldwide. This study raises a serious
limits field voltage Vf of the generator. Thus, at the same time, issue regarding the inaccuracy of PF-based SSA under heavy
the generator limits its reactive power output. So, the standard loading conditions. This paper suggests using the QSS model in
box constraints for the reactive power limit set in the PF model performing SSA and presents a theoretical foundation for the
may be too simplified to catch the actual situation, which also QSS-based SSA to show that the QSS model offers an accurate
brings about different voltage profiles and apparent power approximation of the TD model in terms of the ω-limit set.
outputs in this case. Numerical studies show the accuracy of QSS-based SSA under
both light loading conditions and heavy loading conditions,
confirming the theoretical prediction. It is worth mentioning that
the accuracy of PF-based SSA and QSS-based SSA was also
examined under the following different load models: (i) a
constant PQ load, (ii) the ZIP load, and (iii) the ZIP load with
an induction motor load model. From our extensive numerical
results, we observe that misclassifications in SSA produced by
the PF model also occur in all three load models. On the other
Fig. 10. The LTC regulating process between bus4 and bus9 of the hand, the accuracy of QSS-based SSA under both light and
time-domain dynamic, and the QSS simulation compared with the heavy loading conditions with these three load models is
power flow solution in the IEEE 14-bus system. confirmed in the numerical studies. Due to space limitations,
the examination results were not presented in this paper.
Third, ignoring the interaction between LTC regulation
We have observed that simulating the QSS model is about
(discrete characteristics) and other dynamic components may
one magnitude faster than the TD model. Due to the
further increase the differences. In the PF model, due to the
parallelizable nature of contingencies and recent progress in
rough approximation of the reactive power limitation, within
high-performance computing and commercial availability of
the normal regulating range of the LTC, bus9 can go back to its
GPUs, the applicability of QSS-based SSA in the online
0885-8950 (c) 2018 IEEE. Personal use is 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRS.2019.2914955, IEEE
Transactions on Power Systems
9
environment is feasible. On the other hand, we feel that it is and Applications. Cambridge, UK: Cambridge University Press, 2015, pp.
357-386.
possible to develop a detailed power flow model by [20] R. Fischl, T. F. Halpin, and A. Guvenis, "The Application of Decision
incorporating reactive power/voltage control aspects. For Theory to Contingency Selection," IEEE Trans. Circuits and Syst., vol.
example, the voltage source behind a reactance to approximate 29, pp. 712-723, Nov. 1982
[21] T. V. Cutsem, M. E. Grenier, and D. Lefebvre, "Combined Detailed and
generators’ behavior can be added to the model. Further
Quasi Steady-State Time Simulations for Large disturbance Analysis,"
research and development along these directions are desired. International Journal of Electrical Power and Energy Systems, vol. 28,
One focus of our future work is to include corrective control no. 9, pp. 634–642, 2006.
actions such as SPS/RAS in the model for performing SSA. [22] H.-D. Chiang, A. J. Flueck, K. S. Shah, and N. Balu, "CPFLOW: a
practical tool for tracing power system steady-state stationary behavior
Another is to evaluate the QSS-based SSA on large-scale, due to load and generation variations," IEEE Transactions on Power
realistic test systems whose dynamic parameters are publicly Systems, vol. 10, no. 2, pp. 623-634, May 1995.
available. [23] F. Milano, Power System Modelling and Scripting. London, UK: Springer,
2010, pp. 215-217, 373.
VII. REFERENCES
VIII. BIOGRAPHIES
[1] N. Balu, T. Bertram, A. Bose, V. Brandwajn, G. Cauley, D. Curtice, A.
Fouad, L. Fink, M. G. Lauby, B. F. Wollenberg, and J. N. Wrubel, "On- Meng-yu Ruan received the B.S. degree (June 2016)
in Electrical Engineering from Hebei University of
line power system security analysis," Proceedings of the IEEE, vol. 80,
Technology, Tianjin, China, and she is currently
pp. 262-282, Feb. 1992.
[2] Allen J. Wood, Bruce F. Wollenberg, and Gerald B. Sheble, Power pursuing the M.S. degree in Electrical Engineering at
Tianjin University, Tianjin, China. Her current research
Generation, Operation, and Control. Hoboken, NJ, USA: John Wiley &
interests include power system modelling, power
Sons, 2014.
[3] B. Stott, O. Alsac, and A. J. Monticelli, "Security Analysis and system static security assessment, and nonlinear
computation and applications.
Optimization," Proceedings of the IEEE, vol. 75, pp. 1623-1644, Dec.
1987.
[4] A. Ozdemir, Jae Yun Lim, and C. Singh, "Branch outage simulation for
MVAR flows: bounded network solution," IEEE Trans. Power Syst., vol.
18, no. 4, pp. 1523-1528, 2003.
[5] P. Ruiz and P. Sauer, "Voltage and Reactive Power Estimation for Hsiao-Dong Chiang (M'87–SM'91–F'97) received the
Ph.D. degree in electrical engineering and computer
Contingency Analysis Using Sensitivities," IEEE Trans. Power Syst., vol.
sciences from the University of California at Berkeley,
22, no. 2, pp. 639-647, 2007.
[6] Z. Wu, Z. Hao, and D. Yang, "A new MVA sensitivity method for fast Berkeley, CA, USA. He is a professor in the School of
Electrical and Computer Engineering at Cornell
accurate contingency evaluation", Int. J. Elec. Power, vol. 38, no. 1, pp.1-
8, 2012. University, Ithaca, NY, USA. He and his research team
[7] R. Salgado and A. Berizzi, "A new second-order method for branch have published more than 400 papers in refereed
journals and conference proceedings, obtaining an h-
contingency analysis and static voltage security," Electric Power Systems
Research, vol. 123, pp. 137-146, 2015. index of 56. His current research interests include
[8] G. C. Ejebe and B. F. Wollenberg, "Automatic contingency selection," nonlinear system theory, nonlinear computation,
nonlinear optimization, and their practical applications. He holds 18 U.S. and
IEEE Trans. Power App. Syst., vol. PAS-98, no. 1, pp. 97-109, Jan. 1979.
[9] S. Vemuri and R. E. Usher, "On-line automatic contingency selection overseas patents and several consultant positions. He is the author of two books:
algorithms," IEEE Trans. Power App. Syst., vol. PAS-102, no. 2, pp. 346- Direct Methods for Stability Analysis of Electric Power Systems: Theoretical
Foundations, BCU Methodologies, and Applications (Hoboken, NJ, USA:
354, Feb. 1983.
[10] V. Brandwajn and M. G. Lauby, “Complete bounding method for AC Wiley, 2011) and (with Luis F. C. Alberto) Stability Regions of Nonlinear
contingency screening,” IEEE Trans. Power Syst., vol. 4, no. 2, pp. 724– Dynamical Systems: Theory, Estimation, and Applications (Cambridge, UK:
Cambridge Univ. Press, 2015). Dr. Chiang has served as an associate editor for
729, May 1989.
[11] P. Mitra, V. Vittal, B. Keel, and J. Mistry, "A Systematic Approach to n- several IEEE transactions and journals, and on the editorial board of IEEJ
1-1 Analysis or Power System Security Assessment," IEEE Power and (Japan) and the journal Electric Power Components and Systems. He is the
founder of Bigwood Systems, Inc. and Global Optimal Technology, Inc. in
Energy Technology Systems Journal, vol. 3, pp. 71-80, Jun. 2016.
[12] E. Polymeneas and A. P. Sakis Meliopoulos, "Margin-Based Framework Ithaca, NY, USA.
for Online Contingency Selection in Unbalanced Networks," IEEE Trans.
Power Syst., vol. 32, pp. 30-38, Jan. 2017.
[13] D. Chen, H. Jiang, Y. Li, and D. Xu, "A Two-Layered Parallel Static
Security Assessment for Large-Scale Grids Based on GPU," IEEE
Transactions on Smart Grid, vol. 8, no. 3, pp. 1396-1405, May 2017.
[14] Gan Zhou, Yanjun Feng, Rui Bo, Lungsheng Chien, Xu Zhang, Yansheng
Lang, Yupei Jia, and Zhengping Chen, "GPU-Accelerated Batch-ACPF
Solution for N-1 Static Security Analysis," IEEE Trans. on Smart Grid,
vol. 8, pp. 1406-1416, May 2017.
[15] X. Li, P. Balasubramanian, M. Sahraei-Ardakani, M. Abdi-Khorsand, K.
W. Hedman, and R. Podmore, "Real-Time Contingency Analysis with
Corrective Transmission Switching," IEEE Transactions on Power
Systems, vol. 32, no. 4, pp. 2604-2617, Jul. 2017.
[16] X. Wang and H. D. Chiang, “Analytical Studies of Quasi Steady-State
Model in Power System Long-Term Stability Analysis,” IEEE Trans. on
Circuits and Systems I: Regular Papers,” vol. 61, pp. 943-956, Mar. 2014.
[17] X. Wang and H. D. Chiang, "Numerical investigations on quasi steady-
state model for voltage stability," International Transactions on
Electrical Energy Systems, vol. 24, no. 11, pp. 1586-1599, 2014.
[18] P. Kundur. Power System Stability and Control. New York, NY, USA:
McGraw-Hill, 1994.
[19] H.-D. Chiang and L. F. C. Alberto, "Bifurcations of stability regions,” in
Stability Regions of Nonlinear Dynamical Systems: Theory, Estimation,
0885-8950 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.