Académique Documents
Professionnel Documents
Culture Documents
111.3 Selection of the relevant disturbed machine In order to obtain the desired decomposition of equation
groupings (3) (refer to Reference 6 for detailed derivation) the system
The actual selection of the relevant disturbed machine is divided into r groups, where each group contains at
groupings can now be accomplished by performing the least one machine. The COIs of these r groups form the
following algorithm. slow subsystem. The difference between the C O I angle
For each machine k advanced at the controlling UEP, of each of these r groups and each machine rotor angle
perform steps (i) to (iv) below. Note that machine k within each group constitutes the fast subsystem. For the
corresponds to the kth row of the state vector as given subsequent discussion, two typical groups, e and fl, are
in equation (2) with respect to the rotor angle. denoted, and have centre of inertia angles y~ and y~
(i) Calculate the participation factor for every mode. respectively. As in Reference 6, the rotor angles with
This consists of calculating Pk, = Vk~Uk,,i = 1,3 . . . . . m. respect to the local COls are defined as
(ii) Select the modes in which the participation factor Pk,, X,=Oi--ya, ie~, ~ = 1,2 . . . . . r
i ----1,3 . . . . . m are significant. This consists of selecting
the modes where Pk~, i = 1,3,..., m are greater than and those of the local COIs with respect to the global
a specified minimum value. For the power system COl, 30, as
investigated, typically a value of 0.01 has been used.
~=y~-6 o, :~= 1,2 . . . . . r
Let o,//be the set of modes selected for this advanced
machine k. In other words, Also define
o.#={i:Pki>O.Ol, i = 1,3 . . . . . m}. MT~ = ~ M,
(iii) Rank the modes of .J# in descending order of
F. For example let d be the number of modes The transient energy function (3) can be expressed in
selected in step (ii), then ~# = {[`~, F 2. . . . . Va} where terms of the slow variables y~, and the fast variables
F 1 > F2 > --. > Fa. The advanced machine k may not 2,. Note that the symbol ~ is used to imply a sum-
always be placed in the most dominating mode of mation over all machines in the group e. The transient
,/l. The identification of the mode in which the energy function, in terms of the variables 2,, y~,
advanced machine is to be placed, depends on the ice, c~= 1,2 . . . . . r, is given by
dominance measure of the mode and also on the V = v K E - ] - V PE
where where.~ is the set of relevant disturbed group> and PARI
refers to the partial energy. It should be noted that ~he
Vlqt!= I mi-(:~ ~- M T ~ . ~?;2
, number of relevant disturbed groups in .# is equal t~
2~= 1 = t.... 1. The rth group consists of the machines not
contained in ~
KE
= v,~,,~+ ~ ~,KE
....
~'--1 ~--1
V. Transient stability assessment using
= vf.V, + K,% 141 the M T E F method
The potential energy is divided into three parts so that In order to determine the relevant disturbed machine
vPE= vPE ~_ PE PE groupings by modal analysis and to assess transient
slow-- Vfast-}- [/fast slow (5)
stability using the M T E F method, it is essential that the
The potential energy components are defined by controlling U E P is known. The procedure used for
determining the controlling U E P is based on the method
K,o. = ~X E (PiMg-- P eMI)(y~#- Y:t~) proposed in Reference 12. A brief outline of this
1 //=~+ 1 j procedure is given below.
(5a)
(1) Integrate the disturbed system.
PF
V<,,,= ~f MI ( ~~ ~'~ (P,M~-P~M,)(Sc,~ 2[0 Save conditions at clearing.
:e= 1 k T\ i j=i+ 1 Check for exit point, x e, this is encountered when the
summation, ~ 7 - l - j i & i , goes through a zero (i.e.,
+~ ~[ (PiMj -- PjMi)(2ij -- 5i5) goes from a positive value to a negative value).
(2) Integrate the associated gradient system, with x e as
~s the initial condition, to a point, x c, where, ~7= 1[~(0)1
+ y, Cu(cos 2 u - cos xo-) - I~,,
i /=i+1 i j=i+l is a minimum. For a stressed power network, this
(5b) gradient system is typically stiff, hence a variable
PE
Vfast s l w = 0 t = l
i{ii' -
i O=~t+l
y~ C ~ ; [ c o s ( ~ j - S'=#)
j
step-size integration technique is utilized for its
numerical solution.
(3) Solve the n non-linear equations, Ji(0)=0, for the
controlling U E P with x c as the starting point.
- cos(.%- y~#)]
Transient stability is assessed by calculating the partial
(5c) energy margin given by
i j=i+l i /1==+1
where
A VPART =V:..TI%
PE PARTIO (7)
I ;1
c4.p . 9514
q98B ROGERS I
"C -T~P- " -- - ./
~538
7-16B'~ S 0
F~
S-75=FI
DICKINSON . NORTHTOWN PINES
f 1505-1520 COON 9508
CREEK "" -_ - .
;7 ~
,~6 3-17~
LAN
OSSEO I PARK
1738 , 1774 -1710 '7846
' 16~4 / P A-[-" ]
,'NO. CORCOR~ / LONG LAKE
6~7 ) 9691 i LEXINGTON / 7-1730
/.~GREENFIELD O-17B0
7826 )'BOSS LK ITWIN 8-7861 llll~73" 16~5
~1713-7849 ~AKES(~ -1697
CROW RIVER / ~769 KING
572
M~qINA ( 1.723
~1 ]7"~30q INOIANA~
PARKERS 1J733
~ . . ~ E MEOI"C~'E LK PLACE
1757,B 1747 Obz~D~.E
$NERN~NNCL8 . . HUDS
-OIm2~ ALDRICI 1761 1714 ] , 85C
TED~ / 1717 (.~laL e~X
t BLUFF
-MERRIANMV 7-1922
LE ' ASSUMPM 9268
ORCHARDLK :CEGAN
7626 i ]C~I~ 9244 LAKE 09293
1 CREDIT R]VERj~j.. HILLS F~INGTI~N
' BELPLAIN . v ~ N MARK T HASTING DC
JORDI SPRING AIR LAKE783~''~
7623 9ZRI PI~E
I N.O. BE~ID
MV
ST THOMAS/ N.O.~.~ ~
NEW MARKET. - - 9274
NO i CASTLE R [ 3-1GI
'~027 / 9290 5-16I
]N I 5/B9 - _ - ~ EW PRAGUE I MARIOIW CASTLE R ,PRAIRIE
~F" - - ~ - " 7627 N.O. VESELI 7614 ISLANO
LESUEUR 77B69, B 6 ~ 7622 5-1602 92L:~
7-1603 SPRING
17673 ) I~IRCLE LAKE 8-7642 L CREEK RE
~ ~15~sEDEN. . . . .~ . . . ) ~ PRAGLE_
9259 I~/. . . _. ~., I_~ _ _ . _
L~-.L%Q__.
8-=!.216
ILAKE EMILY
ER' 7676
)MONTGOMERY
k 087
!
I ~
9210 j ~e I\
~IELO CANN~4~AL
)1 ~" f';:klTtCt~ I va, , v DIST
Figure 1. Map of the 345kV loop around the Twin Cities area
Note that both equations (3) and (6) are evaluated groups. The energy absorbing capacity between the
between the conditions at clearing and the controlling non-disturbed machines is neglected.
UEP. By summing V ~ T only over the relevant disturbed
groups, the energy associated with the inter-machine
oscillations of the rest of the system (i.e., non-disturbed V I I I . N u m e r i c a l results of p r e d i c t e d
machines) is neglected. By doing so, V ~ T better v o l t a g e dip
represents the energy of separation associated with the The same five contingency cases previously analysed are
critical group of machines. A similar statement can be used to test the accuracy of the procedure for determining
made for summing V~RT only over the relevant disturbed the post-disturbance transient voltage dip. In all five
300 Notice that V~fw, and VfaKE~,pass through a minimum
Clearing time
=O.~ / and V~AERTpasses through a maximum simultaneously at
about 0.52 and 1.96 s. However, the difference between
200 VpPAERTand V p ~ is larger at 1.96 s. Hence it is this point
that is used to solve for the Prairie Island voltage. It is
interesting to note that the plots of Vslow
, K E and Vfast
KE,.~ in
I00 Figure 3 illustrate the oscillatory features of both the fast
and slow subsystems.
Boundary Dam #70 The bottom graph of Figure 3 illustrates the COI angle
i i
of a group of thirteen generators that are located in the
I
Time (seconds)
2 Twin Cities area. Note here that the COI angle peaks of
the group correspond to the Prairie Island voltage dips.
Figure 2. Unstable swing curves for case 5 The data for the voltage magnitude at the Prairie Island
bus and the post-disturbance system trajectory were
obtained by time simulation (ETMSP). The post-
Table 1. Conlparison of critical clearing times disturbance system trajectory associated with the second
and third graphs was obtained by time simulation
Critical clearing incorporated into the TEF computer program. This time
time (s) simulation consists of a modified trapezoidal method 15.
Time simulation Critical Critical In Table 3, results are shown that compare the voltage
(ETMSP) clearing clearing dip at the Prairie Island 345 kV bus obtained by time
time (s) time (s)
Stable Unstable MTEF TEF*
Table 2. Relevant disturbed machine groupings
Fault at 0.1200 0.1250 0.1124 0.131
Coon Creek Advanced machines
Clear Terminal not in
Case 1 Relevant disturbed relevant disturbed
machine groupings machine groupings
Fault at King 0.1460 0.1465 0.1469 0.168
Clear Red Rock Case 1 (74,77,78)(79)(12,13,35) 8,19,20,24,25,26
Case 2 (31,32)(15)(80)(9,10,11 )
Fault at 0.1175 0.1180 0.1178 0.132 (27)(29,76)(16,17)(65,66)
Monticello (67, 68,69,70,73)
Clear Parkers Lake (21,23,30,33,34,36,37,
Case 3 38,39,40,41,75)(72)(71)
Fault at 0.1250 0.1255 0.1217 0.149
Monticello Case 2 (74,77,78,79)(12,13) 8,15,19,20,24,25,26
Clear Sherco (31,32)(80)(9,10,11,73)
Case 4 (27)(29)(16, 17)
(65,66,67,68,69,70)
Fault at Sherco 0.1115 0.1120 0.1091 0.144
(21,23,30,33,38,39,40,
Clear Coon Creek
41,75)(72)(71 )(76)
Case 5
Case 3 (49)(69)(72) 8,24,26,35,67,68
* This m e t h o d used kinetic e n e r g y c o r r e c t i o n 1.
(65,74, 77,78,80)(22)(28)
cases, the clearing time is set at 0.1 s (6 cycles) so that (48)(44)(12,13,14)(31,32)
each case is stable. (15)(80)(9,10,11)(27)
Through extensive analysis of the system using the (29,30,50,76)(66,70,71,73)
ETMSP time simulation program, it was observed that (19,20,21,23,25,33,34,
there are only two instants during the post-disturbance 42,75)(36,37,45)(16,17)
period in which the voltage magnitude at the Prairie (38,39,40,41,46,47)
Island bus seems to experience a substantial dip. The
first dip usually occurs at about 0.4 s, whereas the second Case 4 (74,77,78,79,80)(76)(71)
dip occurs at about 1.9 s. (9,10)(29,30)(19,20)
The three graphs in Figure 3 illustrate the procedure (65,66,67,68,69,70)
described earlier for case 5. The top graph of Figure 3 (21,23,33,34,36, 37, 38,39,
is comprised of the voltage magnitude at the Prairie 40, 41,75 )(48 )(24)(49)(42)
Island bus and the partial potential energy (with reference (11,25,35,45)
to the conditions at clearing) evaluated along the
post-disturbance system trajectory. Note the two sub- Case 5 (74, 80)(79)(77,78)
stantial voltage dips. This top graph is presented to verify (35)(71)(9,10,11)
the energy based criteria associated with the middle graph (29,30, 76)(19,20, 25)
of Figure 3. (65,66,67,68,69,70)
The middle graph of Figure 3 has the values of V ~ T, (23,33,75)(36,37,42,45)
KE
V~ow, and VfKE (38,39,40,41)(48)(24)
asr ,.~, as evaluated along the system trajectory.
Reference 5 which were formed using slow cohcrcnc~
analysis and are independent of fault location.
While the proposed procedure is computationall~
,4
intensive, it should bc noted that for on-line applications
only certain disturbances may be of interest. F'or each
post-disturbance system operating at a certain nominal
point, modal analysis can be used only once to calculate
the relevant disturbed machine groupings. The M T E F
/11 ~ i ~ i ! i x- i ! i VPE part
- - PI v o l t
method can then be used for transient stability assessmem.
However, the validity of these groupings will be in
question when the system operating point is substantially
changed. The range of operating conditions for a
generator grouping is a subject for future investigation.
In the second part of this paper a procedure to predict
the maximum post-disturbance transient voltage dip is
', ,:L[ 1 1 1 / VKE part
presented. In actuality it is not only the magnitude of
................ VKE slow the dip that is important but also the time duration of
- - ~ fast
this voltage dip. This procedure can be helpful in an
on-line or fast screening environment because it can be
used to test for the violation of a voltage dip limit. If the
voltage does not dip below this limit, the time duration
is irrelevant. Otherwise, further investigation by time
simulation may be required. In the United States the
} ...................
/I ........... ...........
i........... ..........
current regulations state that in a nuclear plant, fi)r the
I:::: coolant motors to be in jeopardy, the voltage must dip
below 0.7 pu for 1 s. None of the contingencies studied
:::::::::::::::::::::::::::::: here gave results that would require further study with
time simulation. However, if such a disturbance did
, ' I I
occur, it is felt that the proposed technique would perform
Time/seconds) satisfactorily.
The procedure seemed to sort out carefully the instants
Figure 3. Voltage-energy, energy and angle plots for where the given criterion almost held up but where there
case 5 was no substantial dip in the voltage. It also accurately
predicted the post-disturbance voltage dip as can be seen
when compared with time simulation results. The
simulation and by the procedure incorporating the procedure demonstrated (as is illustrated by Figure 3)
MTEF. The times at which the voltage dips do occur that the maximum voltage dip did not always coincide
are also shown for both methods. with the largest peak of partial potential energy.
In cases 3, 4, and 5 the maximum voltage dip occurred
at about l.Os. A smaller voltage dip also occurred at
IX. Discussion and conclusions about 0.4 0.5 s in these three cases. It should be noticed
The technique developed for transient stability assess- that the actual difference between the corresponding
ment effectively combines two independent ideas: (1) voltage dips at about 1.9 s and about 0.4 0.5 s is relatively
dominant mode interaction and (2) the modal-based
transient energy function. These are used to capture the
Table 3. Comparison of voltage dips
energy exchange mechanism which occurs in the inter-
area mode separation, even if it occurs several seconds
Voltage dip (in pu) from
into the transient. This is evident in the critical clearing
times obtained.
Time Error
As illustrated in Figure 2, the system seems to maintain
simulation MTEF percent
stability in its first swing; however, the system eventually
becomes unstable. Whereas the T E F method may
Case l 0.927 0.926 0.108 %
incorrectly categorize this case as stable, the M T E F
(a 0.41 s ,~ 0.44s 7.32%
method results agree with the time simulation results.
Therefore, it is evident that the M T E F method has the
Case 2 0.952 0.9522 ....0.021%
capability to assess transient stability beyond the first
(a 0.38 s (, 0.42 s - 10.53%
swing.
The results also reinforce the concept of dominant mode
Case 3 0.922 0.914 0.868%
interactions causing the inter-area mode of separation.
ii~ 1.95s (is 1.94s 0.52%
The relevant disturbed machine groupings are depen-
dent on the disturbance location as well as the system
Case 4 0.934 0.929 0.538%
operating point. The machines in these groupings are
(o 1.88s (a 1.88s 0.00%
placed there because of their strong participation in
certain modes. There is no constraint requiring them to
Case 5 0.904 0.897 0.78 %
be electrically close to each other. These groupings are
,~ 1.92s ,~ 1.96s .... 2.08%
in contrast to the relevant disturbed areas described in
small. Hence if the first instant at which criterion (1), Trans. on Power Systems Vol PWRS-3 (February 1988) pp
which was defined in section V.2, holds is picked and the 239-244
4 V i t t a l , V, Bhatia, N and Fouad, A A 'Analysis of the
voltage solved for, the accuracy of the voltage dip results inter-area mode phenomenon in power systems following large
may be adequate. disturbances" IEEE Winter Power Meetings New York Paper,
The main goal of this work was to demonstrate the no 91WM228-7 PWRS (Feb. 1991)
ability of a criterion based on the M T E F to predict the 6 Verghese, G C, Perez-Arriaga, I J, S c h w e p p e , F C and
Tsai, K W-K "Selective modal analysis in power systems" EPRI
voltage dip during the transient. The method developed
Report EL-2830 (January 1 983)
is general and could be applied without loss of generality 6 Khorasani, K, Pal, M A and Sauer, P W Modal-based
to any load model. However, in this work the constant stability analysis of power systems using energy functions'
impedance load model was used as suggested by the Electric Power andEnergy Systems Vol 8 No 1 (January 1986)
utility engineers on whose system the technique was pp11-16
7 Ray, B "Incorporation of the modal interactions in stressed
tested. power systems using the transient energy function method"
Master's Thesis Iowa State University, Ames, Iowa (1991 )
8 Debs, A 'Voltage dip at maximum angular swing in the context
X. A c k n o w l e d g m e n t s of direct stability analysis' Proc. 1989 PICA Conference,
Seattle, Washington (1989) pp366 371
This work was supported by Iowa State University and 9 Sreedhara, R 'Transient voltage dip analysis using the
by a grant from the Northern States Power Co. (NSP) transient energy function method' Master's Thesis Iowa State
to whom grateful acknowledgement is made. University, Ames, Iowa, (1990)
The authors wish to thank Messrs Dan Nordell, Mike 10 Anderson, P U and Fouad, A A Power System Controland
Stability, Vol 1 The Iowa State University Press, Ames, Iowa
McMullen and Craig Turner of NSP for providing the (1977)
data for the study and for their helpful comments and 11 Treinen, R T 'Application of the transient energy function
suggestions method to the Northern States Power System' Master's Thesis
Iowa State University, Ames, Iowa (1 991 )
12 Chiang, H D, Wu, F F and Varaiya, P P 'Foundations of
the potential energy boundary surface method for power
XI. References system transient stability analysis' IEEE Trans. on Circuits and
1 Fouad, A A and V i t t a l , V Power system transient stability Systems Vol CAS-35 (June 1988) pp 712 728
analysis using the transient energy function method Prentice- 13 Electric Power Research I n s t i t u t e Extended transient
Hall, Englewood Cliffs, New Jersey (1992) mid-term stability package version 2.p EPRI Research Project
2 Carvalho, V F, EI-Kady, M A, Vaahedi, E, Kundur, P, 1208-9, Palo Alto, CA (1 991 )
Tang, C K, Rogers, G, Libaque, J, Wong, D, Fouad, A A, 14 Oh, T K 'Correlation of the transient energy margin to
V i t t a l , V and Rajagopal, S 'Demonstration of large scale out-of-step impedance relay operation' PhD dissertation Iowa
direct analysis of power system transient stability" EPRI Report State University, Ames, Iowa (1986)
EL-4980 (December 1986) 15 Gear, C W Numerical initial value problems in ordinary
3. V i t t a l , V, e t a / . 'Transient stability analysis of stressed power differentia/ equations Prentice-Hall, Englewood Cliffs, New
systems using the transient energy function method' IEEE Jersey (1 971 )