Você está na página 1de 6

Elimination of trannsmission system overloaads through

bbranch switching
D.C. Pinhheiro, and C.A Castro, Senior Member, IEEE
University of Campinas (UNICAMP)
Campinas, Brazil
dpinhheir@dsee.fee.unicamp.br, ccastro@ieee.org

Abstract This paper aims to contribute to the development of best control variables to be used. Another
A approach is based
methods for alleviating overloads in the traansmission system. on the application of linear or nonlinear programming, as in
Among the several possible alternatives to achieve this goal, [3]-[4]. In [5], generation redispattch and/or load shedding
branch (transmission lines and transformers)) switching will be were used to decrease line currents based on sensitivity
the focus in this work. Even though this p possibility is well-
factors. In [3] the idea of local optimization (LO) was
known, a very interesting branch switch hing method was
proposed recently in the literature to identtify the candidate introduced. A fixed sized region (n named working region) in
branches to be switched. In this work the iidea is to propose the neighborhood of the overload ded branch is previously
improvements to this method in such a way itt becomes efficient defined and corrective control action ns are defined within this
enough to be used in real time monitorring and control region. Recently, the utilization of diistribution factors (derived
environments. Those improvements are b based on specific from sensitivity analysis) along witth a very simple and fast
calculation techniques and information aboout the nature of technique to identify the candidate branches to be switched
power systems operation. was proposed [6].
Index Terms -- Power system security analyysis, power system
operation, violations elimination, DC Power fl
flow, overload, line
outage.

I. INTRODUCTION

T he power system operation conddition undergoes


frequent changes as a result of loaad and generation
variations and even contingencies. Thhese changes may
occasionally lead to violations of its state vvariables (such as
voltage magnitudes) and/or dependent vaariables (such as
power flows). The occurrence of such violatiions are related to Fig. 1. Estimate of investments in new transmission lines in Brazil (adapted
power systems security [1], which alsso includes the from [7]). The amounts are in billions of Brazilian Reais (1USD is
determination of procedures to keep the systems secure approximately equal to 3BRR as of March, 20 015).
operation even in equipment failure situationns.
Environmental constraints for the generation and This paper presents contribution
ns to improve the method
transmission expansion, growing penetratioon of alternative of [6], by adding heuristic and d adaptive local search
generation, postponement of network reinfforcement due to techniques, improving the analysis efficiency and providing
adverse economic scenarios (Fig. 1), and improvement of more precise information to the conntrol centers operator, as
power systems analysis and control techniquues are among the described ahead.
factors that have been leading power system ms equipment to
operate closer to their limits. Operational limmit violations have II. PROBLEM FOR
RMULATION
become more frequent, especially after contingencies or The selection of a transmission line
l to be switched off for
unforeseen load variations. In order to keeep safe operation eliminating an overload is a complex x process [8]. This control
conditions, efficient corrective control measuures must be taken action should not cause additional overloads. The selection
to eliminate such violations. Several viollation elimination process must include some procedurre to identify combination
methods have been proposed in the literrature. In [1]-[2] of lines to be switched off as well as to rank them according to
sensitivity analysis methods, particularly nnamed generation their efficacies. The evaluation of these situations through
shift factors, distribution factors and participaation factors were running several load flows is not a practical
p solution, since it
proposed. Heuristics were added for the dettermination of the is computationally expensive.

The authors acknowledge the financial support from Brazilian funding


agencies FAPESP and CNPq.
In [6] a switching method was proposed specifically to be | = | (8)
used in real-time operation environmennts, that is, the
computational aspect was a crucial aspect took into Therefore, it follows a corollary [6]-[10].
[
consideration. The linearized (or DC) loadd flow model was Corollary I: Given a power system modeled
m by the linearized
adopted. load flow, the following relationship
p are valid:
Consider a power system with 1 buses and
branches. The linearized load flow model is ggiven by If > 0, then | a|b |<| | , and (9)
If < 0, then | a|b | > | |. (10)
, (1)
where one of the buses is considered as ann angle reference, Taking into account the sensitivitty between branches a and
is the nodal active power vectorr, is the b, the corollary above can be illusstrated by Fig. 2, that is,
susceptance matrix and is the nodal pphase angle vector reciprocity is valid within the sensitiv
vity analysis procedure.
[9].
Assume that line a is overloaded. In case line a is switched
off either intentionally or by the protection ddevices, the active
power flows through other branches w will change. The
particular flow change in branch b will be
.
| , (2)

where | is the distribution factor (senssitivity factor that Fig. 2. Physical interpretation of corollary I:I reciprocity in power flow ( )
relates the power flow change in b and the ooutage of a), is variation in branches verified by sensitivity an
nalysis.
the power flow through a before its outage, aand , , are the Based on this corollary, the prooposed algorithm will be
reactances of branches a and b respectively. AAlso described ahead, along with the swittching strategies in case of
, (3) line overloads.

where is a n-element vector with alll zeroes but the III. ADAPTIVE LOCAL SEARCH AND HEURISTIC RULES
terminal buses of branch a, which are equual to 1 and 1.
Finally, In general, the consequences of a topological change in a
large system are restricted to a relativ
vely small area around the
, (4) change itself. It has already been observed in practice that the
further the control variable is from the overloaded branch, the
where is a n-element vector with alll zeroes but the
smaller the sensitivity between them m will be. Fig. 3 shows this
terminal buses of branch b, which are equal to 1 and 1. Eqs.
situation, where the sensitivity of a control
c variable from area
[2]-[4] are very well-known [1].
A1 to overloaded branch (OV) iss larger than of control
In [6], rather than considering the effectss of the outage of
variables in area A2, and so on. Theerefore, the inclusion of a
branch a over the remaining branches, the iidea is to evaluate
local search procedure in an allgorithm for eliminating
the possibility of eliminating the overload aat a by switching
overloads results in precise and more efficient results,
off other branches. Let b|a the power flow tthrough branch b
especially for large systems.
after the outage of a, and a|b the power floow through a after
the outage of b. These power flows are givenn by
b|a = + | (5)
and

a|b = + | . (6)
The following theorem [6]-[10] suggestss an algorithm for
eliminating the overload at branch a.
Theorem: Given a power system modeled by the linearized
load flow, | | > 0 always hold. Fig. 3. Sensitivity levels of control variables to overloaded branch OV: large
When a is outaged, the power flow variaation in b is given (A1); moderate (A2); small (A3).
by
Besides, it is important that thee new operating point be
| | (7) obtained by the least amount forr control changes in the
neighborhood of the overloaded branch. In this sense, the
Similarly, when b is outaged, the power flow in a is given concept of local optimization intro oduced here is extremely
by relevant, with the objective of findin ng close control variables
that result in smaller topological chan
nges as possible.
In addition to the local search procedure, heuristic rules E C EC
(11)
have also been applied seeking the selection of control O R P %
variable candidates that provide the largest overload relief,
with the least changes in the systems operating state, where
avoiding undesirable side effects such as new overloads. R (cost-benefit relationship): cost associated to power flow
increase in neighbor branches of the candidate branch (to be
IV. PROPOSED ALGORITHM switched off) and benefit associated to the relief of
Fig. 4 shows a flowchart with the basic steps of the overloaded branch OV.
proposed switching algorithm. EC (estimated changes in the operating condition): refers to
the undesired power flow increase in the neighbor branches of
the candidate branch. These power flow increases are
estimated by equations similar to (8), applied to the neighbor
branches. The values of EC are actually weighted averages
resulting from the estimated power flow variations in the
region around the candidate branch. Table I shows the criteria
and respective weights adopted in this work.

TABLE I: CRITERIA AND WEIGHTS FOR EC


Adopted
Description Weight
criterion
Number of new overloads after the switching of
c1 w1 = 6
the candidate branch
c2 Number of branches with 1 > 40% w2 = 5
c3 Number of branches with 30% < < 40% w3 = 4
c4 Number of branches with 20% < < 30% w4 = 3
c5 Number of branches with 10% < < 20% w5 = 2
c6 Number of branches with 10% < w6 = 1
1
Note: is the sum of the power flow increases through neighbor branches
(N) of the candidate branches.

Therefore, EC can be calculated by the weighted average as



(12)

% (overload reduction) corresponds to the percentual


overload estimate after the control action (branch switching).
It is given by

% (13)

where is obtained from (8).


In (F) the same idea of (C) holds and is applied to the
Fig. 4. Flowchart of the proposed line switching algorithm.
candidate branches.
Regarding this algorithm, the following must be observed.
A. Some remarks about the proposed algorithm
Assume that in (B) there is an ordered list of overloaded
branches, with the worst overloads on the top. In addition to the process described above, the following
In (C) one branch is selected at each iteration following the remarks must be observed during the execution of the
priority defined in (B). Also, the overloaded branch is algorithm:
flagged to avoid control actions that could lead the
The proposed approach is intended for high voltage,
algorithm to cycle.
meshed Bulk Electric Systems (BES) [6]-[10].
In (D) the candidate branches satisfy . > 0. Starting
from the terminal buses of the overloaded branches, the Radial branches connected to generation plants or load
work region is expanded tier by tier in order to find substations are unlikely to be selected in step (D), since the
branches for which their switching would effectively original goal of the proposed algorithm is to avoid the loss
alleviate the overloads. of loads and generations [6].
In (E) the candidate branches are ranked according to the
It has been assumed in the proposed algorithm that at most
following relationship.
two transmission branches can be switched
simultaneously, since too many switchhing actions may
reduce the security of the system to an uunacceptable level
and it may be also a logistically diff fficult task to be
executed in practice.
The adaptive local search is carried out in (D). The
stopping criterion is based on (11) to aavoid unnecessary
calculations involving candidate branchees with very low
sensitivities. The search for candiddate branches is
interrupted when 10. This criterrion was defined
empirically, after numerous simulations uunder a variety of
networks and operating conditions.
If it is impossible to eliminate the overrload at a certain
branch, the process goes on by trying to eliminate the
overload at the next branch. The proceess is interrupted
only when all overloaded branches have bbeen checked [2].

V. SIMULATION RESULTTS
The algorithm was implemented with MA
ATLAB using the
load flow from MATPOWER [11]. The simuulations presented Fig. 5. One line diagram of the 30-bus sysstem: (OV) overloaded branch;
here involve: (BC) candidate branches for switching. BC*
* is the branch recommended to
be switched off. (adapted from [12]).
The application of sensitivity analyysis to the 30-bus
system [12], and The process is initiated by runn ning a load flow and all
The application of the proposed alggorithm to the full power flows are determined. Then, thet overload of branch 6-8
interconnected Brazilian power syystem with 4,526 is detected ( ). Sensitivity analysis is used to determine
buses [13]. for all other branches of the system m. Finally, the condition
. 0 defines the candidate brranches to be switched off
A. 30-bus system - Sensitivity analysis to alleviate overload OV.
By applying this procedure to the 30-bus system, six
The simulation presented here aims to cclarify the process branches were identified as candid dates to be switched off.
of searching for candidate branches, which is the basis for the Table III shows the results of thiis search. The rightmost
algorithm proposed in this paper. column shows the estimated pow wer flow change at the
The data from this system can be oobtained in [12]. overloaded branch 6-8 for each switcching.
Overload at branch 6-8 was defined as shownn in Table II.
TABLE III: SEARCH FOR CANDIDATE BRAN
NCHES (POWER FLOWS IN MW)

TABLE II: OVERLOAD AT BRANCH (6-8) FROM THEE 30-BUS SYSTEM


Branch Overload
Overloaded From To Branch Raated Amount of .
from-to (estimated) reduced?
branch bus bus flow lim
mit Overload
6-8 6 8 24.75MW 24MMW 0.75MW
1-2 9.17 0.022 >0 yes -0.01
6-9 4.8 -0.467 <0 no 0.19
The one line diagram of the 30-bus system
m highlighting the
9-10 4.8 -0.467 <0 no 0.19
overloaded branch 6-8 (OV) is shown in Fig. 5. 4-12 -1.27 -0.416 >0 yes -0.05
12-15 9.76 -0.279 <0 no 0.16
10-21 -2.92 -0.486 >0 yes -0.02
22-24 -2.91 -0.778 >0 yes -0.16
23-24 7.47 -0.372 <0 no 0.34
24-25 -4.14 -1.151 >0 yes -0.83
28-27 -6.27 1.151 <0 no 1.21
8-28 -5.25 24.74 <0 no 5.38
6-28 -1.02 -23.594 >0 yes -0.85
Note1: Branches for which P . P 0 were not
n included in this table.
Note2: The values in column P were obtained by (8).
The switchings of branches (6-28) and (24-25) are the ones strong predominance of hydroelectric plants from different
which result in the largest overload relieves (respectively ownerships. Its operation is very complex, as compared to
0.85MW and 0.83MW). In the case of branch (6-28), its most countries [13]. The SIN is basically divided into four
switching correctly eliminates the overload at branch (6-8), as interconnected areas, namely North, Northeast,
shown in Table IV. Southeast/Midwest and South, as illlustrated in Fig. 7.

TABLE IV: POWER FLOW RESULTS AFTER THE SWITCHING OF CANDIDATE


BRANCH (6-28)

Overloaded From To Branch Rated Security


branch bus bus flow limit Margin
6-8 6 8 23.9MW 24MW 0.1MW

Positive results were also obtained by considering the


switching of branch (24-25), with the resulting elimination of
the overload (Table V).

TABLE V: POWER FLOW RESULTS AFTER THE SWITCHING OF CANDIDATE


BRANCH (24-25)

Overloaded From To Branch Rated Security


branch bus bus flow limit Margin
6-8 6 8 23.92MW 24MW 0.08MW
Fig. 7. Regional Interconnections: North, Northeast, Southeast/Midwest and
South.
However, a careful analysis of the systems operating
conditions after the switching of branch (24-25) reveals that The largest demand occurs in the Southeast/Midwest
the overload elimination is followed by the appearance of new (SE/MW) area. This demand is supplied by a large
overloads. Fig. 6 shows the branches that were affected the transmission grid which corresponds to approximately 60%
most the control action. of SIN. This features makes the SE/MW area very suitable
for testing the proposed method, as mentioned in Sec. IV-A.
25
Base case power flow Table VI shows information about an overload in this area.
Power flow after switching
Power flow limit

20
TABLE VI: OVERLOAD AT THE SIN
Overloaded From To Branch Rated Amount of
branch bus bus flow limit Overload
Pow er flow [M W ]

15
Branch 150 134 396 251MW 239MW 12 MW

10
The one-line diagram shown in Fig. 8 illustrates the details
of the overloaded branch (OV) as well as its neighborhood.
The figure also shows the candidate branches (BC) to be
5
switched off.

0
6- 8 6- 9 9-10 28-27 6-28
Branch
Fig. 6. Switching of branch (24-25) eliminates the overload of (6-8) but
results in new overloads at (6-9), (9-10), (28-27), and (6-28).

Taking into account the side effects (new overloads) of


the switching of (24-25), it becomes clear that the best option
in this case is to switch off branch (6-28).
These results show that the sensitivity-based procedure is
excellent for the search of candidate branches, however, it
does not offer complete information from the decision
making point of view. Therefore, heuristic rules based on (11)
were introduced to refine the decision.

B. Interconnected Brazilian power system with 4,526 buses


The Interconnected Brazilian power system (Portuguese
for SIN - Sistema Interconectado Nacional) is a large system Fig. 8. Overloaded branch and its neighborhood: (OV) overloaded branch;
with some unique features. It is a hydrothermal system with (BC) candidate branches.
After applying the adaptive local search and heuristic generation redispatch and load shedding, with a minimal
rules, branch (213-216) is indicated as the best option for deviation from the initial operating condition.
eliminating the overload. The corresponding results are shown Tests showed that the proposed method is efficient for real
in Table VII. time applications, since its computational burden is very light,
with good precision. The combination of the linearized load
TABLE VII: RANKING OF CANDIDATE BRANCHES ACCORDING TO THE flow, sensitivity analysis, adaptive local search, and
VALUES OF R
appropriate heuristic rules allows the fast and accurate
Candidate
Sensitivity
determination of control actions to manage transmission
Rank branch EC % R overloads.
level
from-to
Another important issue about branch switching is that in
1st 213-216 A2 2.52 5.2 0.5
2nd 367-368 A3 2.61 0.67 3.9 an emergency situation, finding a feasible solution is more
3rd 210-217 A3 2.28 0.33 6.9 important than finding the optimal solution when an online
4th 323-138 A4 2.47 0.039 63.3 decision is needed. Therefore, the branch switching algorithm
(eliminated) does not necessarily guarantee an optimal solution, since a
feasible and approximate solution is sufficient at that point
Table VII also shows that candidate branches with 10 [14].
are considered only. This heuristic criterion was found to be Finally, it has to be emphasized that only MW flow
adequate, eliminating candidates with very small sensitivities violations are considered in this paper. The effect of branch
or with a small potential of solving the overload problem. switching on the voltage profile has not been taken into
The switching of candidate branch (213-216) is indeed a consideration at this point. Of course, its implementation must
viable solution for eliminating the overload without resulting take into account the particular nonlinear characteristics of the
in new overloads, as shown in Fig. 9. voltage problem. The authors are currently working on this
problem.

VII. REFERENCES

[1] A.J. Wood, B.F. Wollenberg, Power generation, operation and control,
John Willey and Sons, 1984.
[2] C.A. Castro, A. Bose, Correctability of voltage violations in on-line
contingency analisys, IEEE Transactions on Power Systems, vol.9, n.3,
pp. 16511657, 1994.
[3] A. Shandilya, H. Gupta, J. Sharma, Method for generation rescheduling
and load shedding to alleviate line overloads using local optimization,
IEE Proceedings C, vol.140, n.5, pp. 337342, 1993.
[4] H.D. Abrantes, C.A. Castro, A new efficient nonlinear programming
based method for branch overload elimination, Controle & Automao,
vol.14, n.1, pp. 7683, 2003.
[5] T.K.P. Medicherla, R. Billinton, M.S. Sachadev, Generation
rescheduling and load shedding to alleviate line overloads, IEEE
Transactions on Power Apparatus and Systems, vol.98, n.6, pp. 1876
1884, 1979.
[6] H. Zhang, B. Gou, A new algorithm of corrective control to relieve
overloads, IEEE PES General Meeting, San Diego, CA, USA, 2012.
Fig. 9. Power flows through branch (134-396) before and after the switching [7] Energy Expansion Decennial Plan 2023 / Ministry of Mines and
of candidate branch (213-216). Energy. EPE, Braslia: MME/EPE, 2013 (in Portuguese).
[8] E.B. Makram, K.P. Thornton, H.E. Brown, Selection of lines to be
Additionally to Fig. 9, Table VIII shows the main switched to eliminate overloaded lines using a Z-matrix method, IEEE
information on the operation of branch (134-396) after the Transactions on Power Systems, vol.4, pp. 653661, 1989.
[9] A.J. Monticelli, Load flow in electric energy networks, Edgard Blcher,
control action. 1983. (in Portuguese)
[10] H. Zhang, B. Gou, A Fast Online Corrective Control Strategy for
TABLE VIII: POWER FLOW RESULTS AFTER THE SWITCHING OF Relieving Overloads, published in IET Generation, Transmission &
CANDIDATE BRANCH (213-216) Distribution, 2013.
[11] R.D. Zimmerman, C.E. Murillo-Snchez, R.J. Thomas, MATPOWER:
Overloaded From To Branch Rated Security Steady-State Operations, Planning and Analysis Tools for Power
branch bus bus flow limit Margin Systems Research and Education, IEEE Transactions on Power
150 134 396 238MW 239MW 1MW Systems, vol. 26, no. 1, pp. 12-19, Feb. 2011.
[12] O. Alsac, B. Stott, Optimal Load Flow with Steady State Security,
VI. CONCLUSIONS IEEE Transactions on Power Apparatus and Systems, Vol. PAS 93, No.
3,1974, pp. 745-751.
This paper presented a procedure for the elimination of [13] Brazilian ISO (ONS - Operador Nacional do Sistema Eltrico),
available on http://www.ons.org.br/.
overloads in transmission systems based on branch switching [14] Wei Shao, Vijay Vittal, A New Algorithm for Relieving Overloads and
defined by sensitivity analysis. The overload elimination is Voltage Violations by Transmission Line and Bus-bar Switching,
achieved without the need of expensive control actions as Proceedings of Power System Conference & Exposition, NY, 2004.

Você também pode gostar