Professional Documents
Culture Documents
Abstract- Power system islanding contingency has always been of island forming and sent alerts/commands to protection
one of the most severe wide area system failures. The recent devices in several severe contingencies during the 2008 Gustav
Hurricane Gustav landed in west coast in 2008 was one exact case to Hurricane.
illuminate the importance and severity of this kind of contingencies. The PMUs offer several types of measurements including
Longtime the researches of islanding situations focus on studying voltage phasors from PTs, current phasors from CTs and the
the coherence groups of the generations under an islanding
contingency. In this paper, a new method in detecting power system
breakers status. These data could all be used for islanding
islanding contingencies using both the systems topological identification. TEPCO have developed a scheme detecting
structure and real-time system dynamic state variables is presented power system islanding using the voltage angle differences
and tested. An islanding severity index concept is given for ranking between PMUs [2, 4]. According to their research, the
the severity of the islanding cases. frequency scheme is not as sensitive as the angle scheme in
their tests [4]. While in the 2008 Hurricane Gustav case, the
Index Terms Phasor measurements, PMU placement protection judgments were made upon frequency
strategy, Islanding contingency, Decision tree algorithm, islanding measurements [1, 3].
severity index The study of power system islanding in the power system has
started for years. N. Senroy and V. Vittal has identified and
I. INTRODUCTION analyzed the behaviors of controlled islands based on forming
program optimizes the results and generates the best decision model is used. The islanding area is limited within the Virginia
tree. This method is strict and has a very good accuracy. states region and a total number of 190 islanding contingency
Through the decision tree generation, three questions are cases are generated and recorded, with 400 non-islanding
expected to be answered: contingency and normal operating cases served as the control
1) The critical locations where the primary splitting parameters group (300 normal operation cases; 100 short circuit, rapid load
are located; changing and line tripping cases). The channels selected are the
2) How much difference of the same system variables between voltage and current measurements at the 50kV level buses
islanding cases and non-islanding cases; inside Virginia area. Table 2 shows the decision tree prediction
3) How many system variables are needed to create an efficient success percentages at different tree sizes using both 10% test
decision tree in detecting islanding contingencies. sample method and 10-fold cross validation method.
For small system, the modified depth first method is no other Terminal
Node 1
Terminal
Node 2
than enumeration method. Both the methods use the same Class = 0
Class Cases %
Class = 1
Class Cases %
0 298 100.0 0 2 22.2
computing concept and speed. When the system is getting 1 0 0.0
W = 298.00
1 7 77.8
W = 9.00
complex, more unreasonable islanding combinations are N = 298 N=9
Category 3:
1) Machines in the islands are angle coherent;
2) The islanded area remains stable after the contingency and
the islanded machine angles go along with the main grid
although it has already been physically separated;
3) Some machines near the islanding boundary are influenced
more by the contingency. Sometimes the generators related to
the islanded area lost stability.
The Islanding Severity Index:
The islanding severity index, an alternative of the PMU
location determination is described in this section.
Voltage stability index (VSI) is widely used in power system
stability analysis. However, theres no specific severity index
Fig. 4: Machine angles oscillation of category 1 cases for judging islanding cases yet. The islanding cases, which vary
from the normal contingencies, show the diversity of the
Category 1: structure of the system. In some separated region, the system
1) Machines in the islands are angle coherent; may stay balanced power generation and load requests. This
2) The islanded area remains stable although it is physically does not mean the region is stable, on the contrast; the system
separated from the main grid; loses its interchange ability between each area and becomes
3) The islanded machine angles have a relatively larger extremely vulnerable for further power oscillation. This may
oscillation in a different direction compared with the main not be reflected from common system state variables for normal
grids; VSIs. While deciding the islanding severity, the power
mismatches and the machines inertia are selected as the state
variables in the proposed method.
And the Islanding Severity Index is defined as:
1) Long term: the summation of the power mismatch over the
machines total inertia time constant in each isolated area.
N
M Ni
mislanding ( ISI ) = i H i = H ij
i =1 H i j =1
related to the generators inertia and have effect to their power [2] Hashiesh, F.; Mostafa, H.E.; Mansour, M.M.; Khatib, A.-R.; Helal, I.;
Wide area transient stability prediction using on-line Artificial Neural
re-dispatch characteristic. Those Island cases with huge Networks, Electric Power Conference, 2008. EPEC 2008. IEEE
generation have correspondingly much larger longterm ISI Canada, 6-7 Oct. 2008 Page(s):1 7
values however a large ISI value does not indicate the machines [3] Kolluri, S.; Mandal, S.; Galvan, F.; Island formation in entergy power
in the island will lose stability. There are also some meter grid during Hurricane Gustav, Thomas, M.; Power & Energy Society
General Meeting, 2009. PES '09. IEEE, 26-30 July 2009 Page(s):1 5
islanding cases in the system; case 7-10 all have interarea [4] T.; Yasuda, T.; Takahashi, O.; Kaminaga, M.; Imai, S.; Islanding
tielines connected with the islanded area. Although in these protection system based on synchronized phasor measurements and its
cases they have relatively high instant and longterm ISI values, operational experiences, Ohno, Power and Energy Society General
the islanded areas are still stable, however the influences to the Meeting - Conversion and Delivery of Electrical Energy in the 21st
Century, 2008 IEEE, 20-24 July 2008 Page(s):1 5
system is more severe than category 1 cases, which are also [5] Chakrabarti, S.; Kyriakides, E.; Optimal Placement of Phasor
shown as the more severe angle oscillations of the boundary Measurement Units for Power System Observability, Power Systems,
generators. IEEE Transactions on, Volume 23, Issue 3, Aug. 2008 Page(s):1433
1440
TABLE 3: ISLANDING SEVERITY INDEX ANALYSIS [6] Ming Zhou; Centeno, V.A.; Phadke, A.G.; Yi Hu; Novosel, D.; Volskis,
Mismatches H.A.R.; A preprocessing method for effective PMU placement studies,
Island
Case (Initial) Inertia Cate ISI_ ISI_lo Electric Utility Deregulation and Restructuring and Power Technologies,
Generation
# Combo gory Instant ngterm 2008. DRPT 2008. Third International Conference on, 6-9 April 2008
(Initial) P Q
1 165+j18.3 -63.42 8.469 15.3 1 0.384 4.145 Page(s):2862 2867
2 210+j14.8 169.98 6.196 12.28 2 0.809 13.842 [7] Ishibashi, A.; Imai, M.; Omata, K.; Sato, S.; Takagi, T.; Nakachi, Y.;
3 1744-j254.2 -429.2 -385.8 20.22 1 0.246 21.226 Ogawa, S.; New type of islanding detection system for distributed
4 606+j74.4 556.23 52.487 27.54 2 0.918 20.197 generation based on voltage angle difference between utility network and
5 348+j84.8 316.02 73.682 26 2 0.908 12.155 distributed generation site, Developments in Power System Protection,
6 275+j9.9 46.582 0.069 15.3 1 0.169 3.0446 2004. Eighth IEE International Conference on, Volume 2, 5-8 April 2004
7 227-j12.6 92.111 -32.85 6.82 3 0.406 13.506 Page(s):542 - 545 Vol.2
8 227-j12.6 84.76 -36.06 6.82 3 0.373 12.428 [8] Wang, H. ; Thorp, J.S.; Optimal Locations for Protection System
9 147.5-j11.9 95.738 -53.60 6.232 3 0.649 15.362 Enhancement: A Simulation of Cascading Outages, Power Delivery,
10 147.5-j11.9 118.67 -47.22 6.232 3 0.805 19.042
IEEE Transactions on, Volume 16 , Issue 4 Oct 2001 Pages(s): 528 533
[9] Senroy, N. ; Heydt, G.T. ; Vittal, V. ; Decision Tree Assisted Controlled
Islanding, Power systems, IEEE Transaction on, Volume 21, Issue 4,
IV. CONCLUSIONS Nov. 2006 Page(s):1709 - 1797
[10] Ruisheng Diao ; Vittal, V. ; Kai Sun ; Kolluri, S. ; Mandal, S. ; Galvan, F.
In this paper, the power system islanding contingency ; Decision Tree Assisted Controlled Islanding for Preventing Cascading
analysis method is presented and tested. The islanding Events Power Systems Conference and Exposition, 2009. PSCE '09.
IEEE/PES, 15-18 March 2009, Page(s): 1 - 8
detecting strategy is using the systems topological structure [11] Hashiesh, F.; Mostafa, H.E.; Mansour, M.M.; Khatib, A.-R.; Helal, I.;
and real-time breaker status for the judge of whether islands Wide area transient stability prediction using on-line Artificial Neural
occur or not. In the islanding contingency simulation Networks, Electric Power Conference, 2008. EPEC 2008. IEEE
procedure, a modified enumeration method is introduced in Canada, 6-7 Oct. 2008 Page(s):1 - 7
[12] Jerel Alan Culliss, A Method for PMU-Based Reconfigurable
selecting reasonable potential islanding locations. The Monitoring, VT graduate thesis, 2009
geographical and electrical distances are converted into the [13] Power Systems Engineering Research Center, Optimal Placement of
observation depth and unrealistic islanding cases are Phasor Measurement Units for State Estimation, Texas T&M
successfully eliminated. The decision tree algorithm is used to University, 2005
[14] Jian Chen; Abur, A.; Improved bad data processing via strategic
analyze the distinction between islanding contingencies and placement of PMUs, Power Engineering Society General Meeting,
other operating conditions. Simulations are carried out on east 2005. IEEE, 12-16 June 2005 Page(s):509 - 513 Vol. 1
coast full-loop system and an average predication success of
98% is achieved for a full scale tree; the simulation results VII. BIOGRAPHIES
demonstrate the decision tree algorithm is effective in the Rui Sun (S10) received his B.S. degree from Southeast University, Nanjing,
islanding judgment for large scare power system models. China, in 2007 and M.S. degree from Clemson University, in 2009,
Furthermore, the important variables and primary splitters respectively. He is currently pursuing the Ph.D. degree from the Department of
could help in deciding the PMU location. The concept of the Electrical and Computer Engineering, Virginia Polytechnic Institute and State
University, Blacksburg. His main research interests are power system state
islanding severity index, an alternative of the PMU location estimation and power system wide area measurement.
determination is also introduced and tested using simulated
islanding contingencies. Zhongyu Wu (S10) received her B.S. degree from Southeast University,
Nanjing, China, in 2008 and M.S. degree from Virginia Tech, in 2009. She is
currently pursuing the Ph.D. degree from the Department of Electrical and
V. ACKNOWLEDGEMENTS Computer Engineering, Virginia Tech, Blacksburg. Her main research interests
The authors gratefully acknowledge the contributions of are power system wide area measurement, power system state estimation and
power system protection.
Arun G. Phadke, James S. Thorp, Craig Crider and Matt
Gardner from Dominion Virginia Power for their work on this Virgilio A. Centeno (M92--SM06) received the M.S. and Ph.D. degrees in
document. electrical engineering from Virginia Polytechnic Institute and State University
(Virginia Tech), Blacksburg, in 1988 and 1995, respectively. He worked as a
Project Engineer at Macrodyne, Inc., Clifton Park, NY, in the development of
VI. REFERENCES phasor measurement units from 1991 to 1997. He joined the faculty of Virginia
[1] Galvan, F.; Mandal, S.; Thomas, M.; Phasor Measurement Units (PMU) Tech as a Visiting Professor in the fall of 1997 and became an Associate
instrumental in detecting and managing the electrical island created in the Professor in 2007. His area of interest is wide area measurement and its
aftermath of Hurricane Gustav, Power Systems Conference and applications.
Exposition, 2009. PES '09. IEEE/PES, 15-18 March 2009 Page(s):1 4