Professional Documents
Culture Documents
Losses Epsr 1998
Losses Epsr 1998
Abstract
The search for a reliable and computationally efficient index for voltage security analysis has been the subject of concern of
many researchers. Recently, a method based on the tangent vector technique has been shown to be accurate as a voltage stability
index and also as an indicator of the system weak area. This technique is tested in this paper with different load scenarios. The
tangent vector technique is then used as a tool for determination of system loss sensitivity. This new application for the tangent
vector technique is tested in the IEEE-14 bus system test. © 1998 Elsevier Science S.A. All rights reserved.
duces the quadratic extrapolation method used to de- bifurcation point. This drawback may be overcome
termine the bifurcation point. Section 4 proposes the through the methodology proposed in the next section.
loss sensitivity technique based on tangent vector. Sec-
tion 5 shows the results obtained by the technique when
the IEEE-14 bus system test is employed and Section 6 3. Bifurcation point determination by quadratic
gives the conclusions of this work. extrapolation
ÆDu gÇ ÆP g0Ç
1
à Du l à =[J] à P l0 Ã
−1
(6) 4. Loss sensitivity based on tangent vector
Dl
ÈDV lÉ ÈQ l0É
The tangent vector calculated in Eq. (6) is used here
The investigation of tangent vector behavior as a as a new tool for system loss sensitivity. This novel
function of load increase is exploited by Zambroni de approach is based on the information provided by
Souza et al. [8], where a comparison between a tangent tangent vector, i.e. how state variables vary as a func-
vector and a right Eigenvector is shown for the purpose tion of system parameter. The total system active power
of critical bus identification. It is concluded that the losses are given by:
tangent vector allows early identification of the critical nl
bus, which is not obtained by any other index. Zam- Psystem = % VikVjk (Gk (cos (d(ij )k )+ cos (d( ji )k ))
k=1
broni de Souza et al. however [8], show that tangent
vector behavior does not allow one to predict the − Gk (V 2ik + V 2jk) (8)
A.C. Zambroni de Souza / Electric Power Systems Research 47 (1998) 65–70 67
5. Results
then
flow is used. As can be seen, the quadratic extrapola-
nl
dPsystem dVik dV dA tion method proposed by Zambroni de Souza et al. [8]
= % Gk Vjk + jk Vik A +VikVjk
dl k=1 dl dl dl provides good results, also emphasizing the shorter
dVik dVjk computational time involved in the quadratic extrapo-
−2Gk Vik + V (9) lation process. The voltage collapse point shown above
dl dl jk
has been evaluated assuming a load increase direction
where, A= Gk (cos (d(ij )k ) + cos (d( ji )k ) and dA/dl = as a function of the initial system loading. All load
Gk (2 sin (d( ji )k )(dd(i )k /dl − dd( j )k /dl)). buses have experimented load increase, up to the stage
Eq. (9) shows how active power losses vary as a where the voltage collapse point is encountered. This
function of system parameter. Notice that all the partial typical scenario has been widely exploited in the litera-
derivatives of Eq. (9) consist of tangent vector compo- ture. For this load increase direction, the critical bus is
nents, known from Eq. (6). Therefore, computing Eq. evaluated as Bus 14. It is known that different load
(9) is not time consuming. increase directions may provide different bifurcation
points [9]. This situation has been exhaustively studied
Table 1
Load margin for little bus load variation
in the literature. However, in order to check whether
the technique proposed in Section 3 works properly for
Load increase at buses Critical bus Load margin different load scenarios, some situations are considered.
(l)
14 14 8.1
9, 13, 14 (Level 1) 14 4.8
1
9, 13, 14, 12, 6, 10, 7, 4 (Level 2) 14 4.1 Test system data are available by anonymous ftp at wa-
hoo.ee.washington.edu
68 A.C. Zambroni de Souza / Electric Power Systems Research 47 (1998) 65–70
5.1.1. Load increase in part of the system tion does not allow one to obtain good results if load
The idea of analyzing load increase in part of the increase is not considered for all the buses contained
system arises for two reasons: (1) simulating a real in the critical area. In contrast, some buses may expe-
situation, such as a sudden load increase in a particu- rience a sudden load variation and this consideration
lar system bus; and (2) obtaining the system response may be analyzed here; Table 1 summarizes this
analyzing a reduced part of the system only. scenario.
Item (2) above refers to network partitioning analy- From Table 1, one can see that as the level is
sis. Vargas and Quintana [10] propose a network par- increased, the subsystem load margin becomes closer
titioning technique based on Eigenvalue to the result obtained for the whole system. The result
decomposition, whereas Cañizares et al. [9] show many obtained for Level 2 indicates that a good subsystem
different network partitioning techniques. The re- (partition) has been obtained. Notice however, that
ported results [9] have been improved by Zambroni de most of the system has been considered, leading one
Souza [11], where the knowledge of the critical bus is to conclude that the network partitioning problem is
incorporated to the techniques proposed by Cañizares not trivial [11].
et al. [9] in the following way: (a) from the critical If a load increase is experienced by the whole sys-
bus, create an initial area formed by this bus and its tem except the critical bus, the following result is
first neighbors. (Level 1); and (b) create a new level obtained:
formed by the buses considered in level 1 and its l= 4.6
neighbor buses.
Network partitioning technique is beyond the goal The result above stresses the importance of critical
of this work. The interested reader should refer to area identification. Notice that the load margin ob-
tained in this case is larger than in the base case, when
Cañizares et al. [9] and Zambroni de Souza [11]. These
load is increased in the whole system. Another impor-
references show that a good partitioning from the
tant characteristic regarding this load increase scenario
critical bus identifies the critical area, i.e. that where
relates to critical bus identification; this time, Bus 12
voltage collapse starts. However, critical bus identifica-
overtakes Bus 14 as the critical one.
Table 2 Table 3
Active loss variation (base case) Active loss variation (bifurcation point)
4 0.0922 4 2.955
5 0.0921 5 2.9537
7 0.0916 7 2.9586
9 0.0915 9 2.9503
10 0.0914 10 2.9510
11 0.0917 11 2.9642
12 0.0917 12 2.9819
13 0.0912 13 2.9736
14 0.0911 14 2.9583
A.C. Zambroni de Souza / Electric Power Systems Research 47 (1998) 65–70 69
5.1.2. Load ZIP model 12). On the contrary, it is the worst candidate for this
Many works in the literature have shown that a static purpose. Hence, loss reduction must be analyzed for
system model may suffice for voltage collapse analysis, each operating point separately from voltage collapse
although the phenomenon is recognized as a dynamic analysis. Table 3 shows the active power losses when a
event. Xu and Mansour [12] propose a dynamic load capacitor of 0.1 p.u. is installed at each load bus. The
model for voltage collapse analysis. In contrast, Dob- results shown in this table confirm the load buses rank
son [13] shows that under certain considerations, the easily obtained via Eq. (9).
dynamic load model may not be so relevant. However, Comparing with Table 2, one can see that system
even when dealing with static load models, the bifurca- active losses are much larger in this case. This is
tion point may vary significantly [14]. Fig. 2 shows the expected, since Table 3 has been evaluated at the
results obtained when the following polynomial model bifurcation point, where transmission lines are heavily
is employed: the amount of constant impedance is fixed loaded.
at 20%. The remaining 80% is varied between constant
power and constant current. The x-label shows the load
margin obtained when the constant power amount 6. Conclusions
shown in the y-label is considered. Notice that the
smaller the amount of constant power, the larger the A new and effective methodology for bifurcation
load margin. point is tested for different operating conditions. It is
shown that the method produces results depending on
5.2. loss sensiti6ity load increase direction and load model. For any operat-
ing scenario, the bifurcation point is accurately deter-
This section employs the technique proposed in Sec- mined. A novel methodology for system loss sensitivity
tion 4. The idea is that for a known operating point, the is also proposed. It is shown that the critical bus may
load bus which reduces the overall system losses when a not provide the most effective results for loss reduction.
capacitor is installed is evaluated. This technique has Regardless of the purpose of the analysis (voltage col-
been implemented in MATLAB [15]. As mentioned in lapse or loss reduction), tangent vector is shown to be
Section 4, the evaluation of Eq. (9) does not require a very effective, even though it does not allow the incor-
large computational effort. For the initial operating poration of a dynamic system model.
point, Eq. (9) provides the following rank:
14, 9, 10, 5, 4, 7, 11, 13, 12
Acknowledgements
In order to prove the result stated above, Table 2 shows
the system active power losses when a capacitor of 0.1 The author would like to thank the Brazilian agen-
p.u. is installed at each load bus. cies CNPq, FAPEMIG and Finep/Recope (profect
Therefore, Bus 14 (the critical one) is the bus whose 0626/98-SAGE) for financial support and Professor
capacitor installation most reduces the system losses. Armando M.L. da Silva, for his valuable comments.
Next, the system is driven to the bifurcation point and
loss sensitivity is calculated again, yielding:
14, 9, 10, 5, 4, 7, 11, 13, 12 References
Once more, the critical bus is indicated as the best one [1] CIGRE WG 38.02 Task Force No. 10, Modelling of voltage
for reduction of the overall system losses. Therefore, for collapse including dynamic phenomena, Technical report of task
this case evaluating the critical bus at the initial point is force 38-02-10, draft 3, CIGRE, June 1992.
[2] IEEE Work Group on Voltage stability, System Dynamic Per-
sufficient to identify the bus whose capacitor installa- formance Subcommittee, Voltage Stability of Power Systems,
tion is more effective for system loss reduction at the Technical report 90THO358-2-PWR, IEEE, 1990.
bifurcation point. However, there are situations where [3] P.W. Sauer, M.A. Pai, Power system steady-state stability and
the critical bus may change during the voltage collapse the load-flow Jacobian, IEEE Trans. Power Syst. 5 (4) (1990)
process. This situation must be considered when analyz- 1374 – 1381.
[4] Claudio A. Cañizares, A.C. Zambroni de Souza, V.H. Quintana,
ing system loss reduction. In this work, such a situation Comparison of performance indices for detection of proximity to
occurs when Bus 14 does not experience load increase. voltage collapse, IEEE Trans. Power Syst. 11 (3) (1996) 1441–
Taking the system to the bifurcation point in this 1447.
scenario yields the ranking: [5] C.A. Cañizares, F.L. Alvarado, Point of Collapse and continua-
tion methods for large ac/dc systems, IEEE Trans. Power Syst. 8
9, 10, 5, 4, 14, 7, 11, 13, 12 (1) (1993) 1 – 8.
[6] A. Venkataramana, C. Collin, The continuation power flow: a
In this case, system losses are not maximally reduced tool for steady-state voltage stability analysis, IEEE PICA Conf.
when the capacitor is installed at the critical bus (Bus Proc., Baltimore, MD, May 1991, pp. 304 – 311.
70 A.C. Zambroni de Souza / Electric Power Systems Research 47 (1998) 65–70
[7] R. Seydel, From Equilibrium to Chaos—Practical Bifurcation mine proximity to static voltage collapse, PhD thesis, Univer-
and Stability Analysis, Elsevier, Amsterdam, 1988. sity of Waterloo, Ont., 1995.
[8] A.C. Zambroni de Souza, C.A. Cañizares, V.H. Quintana, [12] W. Xu, Y. Mansour, Voltage stability using generic dynamic
Critical bus and point of collapse determination using tangent load models, IEEE/PES Winter Meeting, Columbus, OH, Jan-
vectors, 28th North American Power Symp., Cambridge, MA, uary 31 – February 5, 1993.
November 10 – 12, 1996, pp. 329–333. [13] I. Dobson, The irrelevance of load dynamics for the loading
[9] C.A. Cañizares, A.C. Zambroni de Souza, V.H. Quintana, Im- margin to voltage collapse and its sensitivities, in: L.H. Fink
proving continuation methods for tracing bifurcation diagrams (Ed.), Proc. Bulk Power System Voltage Phenomena III—
in power systems, in: L.H. Fink (Ed.), Proc. Bulk Power Sys- Voltage Stability and Security, ECC Inc., Davos, Switzerland,
tem Voltage Phenomena III—Voltage Stability and Security, 1994, pp. 509 – 518.
ECC Inc., Davos, Switzerland, 1994, pp. 349–358. [14] F.L. Alvarado, Voltage stability including detailed load mod-
[10] L. Vargas, V.H. Quintana, Clustering techniques for els, in: L.H. Fink (Ed.), Proc. Bulk Power System Voltage
voltage collapse detection, Electr. Power Syst. Res. 26 (1993) Phenomena III — Voltage Stability and Security, ECC Inc.,
53 – 59. Davos, Switzerland, 1994, pp. 471 – 475.
[11] A.C. Zambroni de Souza, New methods to efficiently deter- [15] MATLAB, The Math Works Inc., Natick, MA, 1993.
.
.