Вы находитесь на странице: 1из 6

776

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004

Quasi-Coupled Three-Phase Radial Load Flow


Esther Romero Ramos, Antonio Gmez Expsito, and Gabriel lvarez Cordero

AbstractIgnoring mutual coupling in three-phase load-flow


computations saves a lot of computational effort but provides
inaccurate results for many applications. In this paper, the effect
of mutual coupling is taken into account through equivalent
branch voltage sources or bus current injections. This simple idea
allows accurate enough solutions to be obtained, while much of
the computational saving associated with decoupled load flows
is retained. Experiments on several unbalanced networks are
reported showing the performance of the proposed idea when
applied to distribution load flows based on the forward/backward
tree sweeps.
Index TermsDistribution networks, three-phase load flow, unbalanced loads.

I. INTRODUCTION

OTH at operational and planning stages, the load-flow


tool constitutes the cornerstone of many important applications. Efficient and reliable load-flow solutions, such as
the NewtonRaphson [1] and the fast decoupled load flow
[2], have been widely used in transmission systems. It has
long been known, however, that these techniques may present
convergence problems when applied to certain distribution netratios,
works, owing to their distinctive features (higher
radial or weakly meshed topology, high ratio of long-to-short
line reactance for lines incident to the same bus).
Consequently, load-flow techniques specially designed for
distribution networks have been developed. References [3][15]
constitute examples of such load-flow methods intended for radial networks. Most of them take full advantage of the radial
topology to save computation time. Some of these techniques
are also suitable for weakly meshed cases [4][9], [12][15],
and/or voltage-dependent loads [4], [5], [9], [10], [12][14].
Most published methods are suitable to perform three-phase
simulations. This is of interest particularly in the U.S., where
many single-phase laterals and service transformers exist, but
also in those cases where balanced networks feed significantly
unbalanced loads. In addition to the increased data and model
complexity, three-phase load flows involve much more computational effort than three single-phase load flows. This computational effort becomes critical when the load flow is run
many times, as happens, for instance, in service restoration [16],
feeder reconfiguration [17], [18], phase balancing [19], [20],
volt/var control [21], optimal location of capacitors [22], etc.
The application of such complex optimization tools to unbal-

Manuscript received June 12, 2003.


E. R. Ramos and A. G. Expsito are with the Electrical Engineering Department, University of Sevilla, Sevilla 41092, Spain (e-mail: eromero@us.es;
age@us.es).
G. . Cordero is with REE, Madrid, Spain (e-mail: galvarez@ree.es).
Digital Object Identifier 10.1109/TPWRS.2003.821624

anced networks demands for as fast as possible load-flow subroutines.


As discussed in [23], most voltage and current imbalance is
due to unbalanced loading, the network asymmetry being much
less important. Partly because of this, ignoring mutual coupling
and running three separate single-phase load flows usually provides satisfactory results. This strategy is, however, unacceptable when exact solutions are sought.
In this paper, a methodology is presented from which several phase-decoupled distribution load-flow schemes can be developed. The basic idea lies in modeling the effect of mutual
coupling as branch voltage sources or additional bus injections,
rather than neglecting it. This makes phase decoupling compatible with accurate results but still much less expensive than full
three-phase computations.
The paper is organized as follows. Section II briefly reviews
electrical models associated with distribution feeders and the
two major categories of load-flow methods. Section III presents
the proposed idea and discusses possible ways of applying it
to existing load flows. Finally, test results are provided in Section IV.
II. BACKGROUND
Distribution systems consist of unbalanced three-phase
feeders from which two- and single-phase laterals are usually
tapped off. In this situation, one-line equivalent models do not
constitute accurate representations.
The impedances for overhead and underground lines can be
calculated in a very accurate form using Carsons equations.
Carsons original equations can be reduced to a much simpler
and easy to use form, known as modified Carsons equations,
with no significant loss of accuracy [24]. When modified
4 primitive impedance
Carsons equations are used, a 4
matrix results for an overhead four-wire grounded-wye line
section. In the same way, for an underground grounded-wye
line segment consisting of three concentric neutral cables, a
6 6 matrix will result [24].
For most applications, the primitive impedance matrix needs
to be reduced to a 3 3 phase-frame matrix consisting of the
self- and mutual-equivalent impedances for the three phases (see
Fig. 1). A standard method to do that is the Krons reduction
[25]. In this case, the assumption is made that the line has a
multigrounded neutral. So, the final form of the Krons reduction technique allows relating bus voltages with branch currents
by means of the following equation:

0885-8950/04$20.00 2004 IEEE

(1)

RAMOS et al.: QUASI-COUPLED THREE-PHASE RADIAL LOAD FLOW

777

Fig. 1. Three-phase branch section between buses i and j .

The above coupled model constitutes the building block


of any computational method that simultaneously analyzes
the three phases and their interactions (mutual coupling).
Broadly speaking, there are two kinds of three-phase load-flow
techniques, namely
1) methods formally applying the NewtonRaphson iterative scheme which, at each iteration, solves a system of
the form

Fig. 2. Modeling of branch coupling by a voltage source.

(2)
Usually, but not necessarily,
comprises the bus
reprevoltage corrections in polar coordinates and
sents some form of the power [26] or current mismatches
[9], [27];
2) methods taking advantage of the radial structure to avoid
explicitly storing and factorizing any matrix. A backward sweep first estimates branch currents (or powers)
followed by a forward-voltage-updating procedure that
applies (1) [3], [15].
Among the main benefits of three-phase analysis, by far the
largest in terms of general accuracy is representation of unbalanced loading and configuration. However, some models of
electrical behavior used in distribution planning and operation
neglect mutual coupling by assuming all off-diagonal terms in
(1) as zero
(3)

In this approach (see [23]), a one-line load-flow algorithm is


used to compute voltages and currents on each of the three
phases individually, as if the other two did not exist. This greatly
reduces complexity of data and computation, but introduces noticeable inaccuracies, particularly when representing rural distribution systems.
III. PROPOSED METHODOLOGY
Equation (1) can be rewritten as follows:

(4)
where

Fig. 3. Voltage-source to current-source transformation followed by


current-source shift.

are the additional voltage drops originated by the mutual cou, in series
pling. These voltage sources
, are represented in Fig. 2.
with the impedances
in series with
It can be shown that every voltage source
the impedance
is equivalent to a pair of equal but opposite
current sources, one entering node and the other leaving node
, whose value is
(5)
Fig. 3 represents, for phase , the circuit transformations suggested above.
Notice that, at a nonterminal node, the total current injection
due to mutual coupling will be the contribution of outcoming
branches minus that of the incoming branch. Fig. 4 illustrates
this idea for a 3-branch system (only one phase is shown).
An improvement to the conventional decoupled approach described in the previous section can be achieved by using the
ideas presented above. Depending on the load-flow technique
adopted, two different schemes can be considered.
1) Methods explicitly solving an equation system. In terms
of phases, rather than buses, the block structure of (2) is
(6)

778

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004

Fig. 5.

The 10-node system three-phase diagram.

Fig. 6.

The 25-node system three-phase diagram.

Fig. 4. Current injections accounting for mutual coupling for a 3-branch


system.

The procedure presented in this section allows writing (6)


as

(7)

where the right-hand side vector has been modified to account for the mutual network terms, that is, the equivalent bus current injections (note that, in [14], off-diagonal blocks are simply ignored). Two possibilities arise to
solve (7), namely: 1) The current mismatch vector is simultaneously computed for the three-phases; 2) The current mismatch vector is sequentially computed for each
phase and the respective equation is solved. As the solution of each phase benefits from previously updated
bus voltages, better convergence rates are expected for
the second approach (the number of iterations could be
even smaller than that of the three-phase conventional
solution).
This category of solution methods is currently being
investigated and will not be further pursued in this paper.
Note, however, that, if convergence does not significantly
deteriorate, large computational savings are expected as
a consequence of three 2N 2N equation systems, comprising 12N nonzero elements each, being solved, instead
of a 6N 6N system containing 108N nonzero elements.
2) Methods exploiting the radial structure. In this case,
there is no need to perform the voltage-source to current-source transformation. The backward sweep first
estimates branch currents as usual, while three arrangements are possible for the forward sweep:
Fully decoupled: (3) is solved so that no mutual
coupling is taken into account.
Exactly coupled: Voltages are updated by means of
, are saved for
(1). Branch voltage sources,
future use.
Approximately coupled: (4) is applied. Instead of re, they are retrieved from
computing the terms
a previous iteration.

Fig. 7. The 1007-node system single-phase diagram.

The key idea consists of starting with a certain


number of fully decoupled forward sweeps. Then,
a single exactly coupled forward sweep is carried
out, followed by the required number of approximately coupled forward sweeps to achieve convergence. If the exact forward sweep is performed
too late, then extra iterations may be required. If it
is applied from the very beginning, then the values
may not be accurate enough.
Since the forward sweeps resort to the respective
3 branch impedance matrix to update the
3
next bus voltages, without actually building the
whole network Jacobian, the computational saving

RAMOS et al.: QUASI-COUPLED THREE-PHASE RADIAL LOAD FLOW

779

TABLE I
NUMBER OF ITERATIONS, COMPUTATIONAL EFFICIENCY, AND RELATIVE ERRORS

TABLE II
LARGEST COMPONENT

OF THE POWER MISMATCH


FOR THE 1007-BUS SYSTEM

VECTOR

TABLE III
LARGEST COMPONENT

OF THE POWER MISMATCH


FOR THE 13-BUS SYSTEM

VECTOR

Fig. 8. Voltage magnitude profile along the lateral leading to the terminal bus
1028 (phase b) for the 1007-bus system.

Fig. 9. Voltage magnitude profile along the lateral leading to the terminal bus
675 (phase b) for the 13-bus system.

attained by the adopted solution may be not so


dramatic as in 1).
IV. TEST RESULTS
In this section, scheme 2) described above, based on a combination of different forward/backward sweeps, will be tested
and compared with both an exact three-phase load flow [15] and
its fully decoupled counterpart. Two versions of the proposed
methodology will be applied:

(a) only the first forward sweep is fully decoupled;


(b) the first two forward sweeps are fully decoupled.
Results corresponding to the following unbalanced systems
will be reported:
A 8.66-kV, 10-node system whose three-phase diagram
is shown in Fig. 5 (both the electrical data and the exact
solution are included in the Appendix). While the threephase branch sections of this system are symmetrically
coupled, all bus loads are quite unbalanced. In addition,
large voltage drops are expected, particularly at phase
which feeds a total load of 450 kW.
A 13-node system downloaded from [28]. This IEEE
test system comprises several load types and a regulating
transformer. As we have not developed a fully featured
commercial code, its tap settings have been assigned their
converged values. Also, branch shunt susceptances are
ignored which, according to [24], provides virtually exact
results.
A 4.16-kV, 25-node system reported in [4] (see the threephase diagram in Fig. 6). This network comprises unsym-

780

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004

TABLE IV
NUMBER OF ITERATIONS, COMPUTATIONAL EFFICIENCY AND RELATIVE ERRORS FOR THE 1007-BUS SYSTEM AND DIFFERENT LOAD FACTORS

metrical branch sections but the loading level is not as high


as in the previous cases.
A 13.8-kV, 1007-node real system whose one-line diagram is shown in Fig. 7.
Table I collects the most important figures obtained when
solving the above networks on a 933-MHz, 128-Mb PC. For the
exact method, both the number of iterations and absolute execution times are provided. For the approximate methods, from left
to right, the number of iterations, relative computational saving
and largest voltage magnitude error are given. Relative computational saving, defined as

refers to the whole iterative process. Percentages shown are


the average of several runs, as individual execution times are
slightly different under Windows. The convergence threshold is
for all experiments reported in Table I.
Figs. 8 and 9 represent the voltage magnitude profile along
the lateral and phase leading to largest errors, for the 1007-bus
and 13-bus systems, respectively. It can be observed in Fig. 8
that, because of the coupling, voltage magnitudes start to rise
from a given bus to the lateral end. This effect is neglected by
the fully decoupled method, which explains the resulting larger
errors.
Tables II and III present, for the same two systems, the largest
component of the power mismatch vector at the end of each
iteration. These two systems constitute extreme cases, both from
the point of view of convergence rate and imbalance. Note that
convergence rate for the 13-bus network is nearly lineal.
From the results presented above, the following conclusions
can be reached.
1) The proposed quasicoupled methodology is about two orders of magnitude more accurate than the fully decoupled
approach. Voltage magnitude errors are abnormally large
for the 13-bus system, also requiring more iterations to
achieve convergence.
2) About 40% computational saving can be obtained with
version (a) of the proposed methodology, compared to the
50% reduction achieved by the fully decoupled algorithm,
which cannot guarantee accurate enough results. These
figures do not apply to the 13-bus system, constituting
probably one of the hardest test cases for any load-flow
algorithm.

TABLE V
LOADS (kW AND kVAR) FOR THE 10-NODE SYSTEM

3) Version (b) of the proposed scheme takes more time than


version (a), because of the extra iteration required for convergence in all tested cases.
Table IV expands, for the 1007-bus system, the data presented
in Table I by increasing each bus load proportionally. Apart from
the larger number of iterations, the above comments still apply.
A noteworthy difference arises for the last two cases, where version (b) of the proposed methodology takes less iterations and
provides more accurate results than version (a). This suggests
that computation of voltage sources modeling branch coupling
should be delayed under stressed conditions until better branch
currents are available.
V. CONCLUSIONS
A well-known approach to approximately computing
three-phase load flows is based on ignoring the mutual coupling among phases so that three separate single-phase load
flows can be run. However, the results provided by this procedure may not be accurate enough in certain circumstances.
This paper further explores the possibilities for a phase-decoupled methodology to be applied to unbalanced distribution
networks without losing accuracy. With this goal in mind, a procedure is proposed by which the mutual terms are accounted for
by adding equivalent current injections to each bus or inserting
voltage sources in series with each branch. Bus current injections are more appropriate when NewtonRaphson-based load
flows are employed, while branch voltage sources are preferable
when forward/backward sweeps are adopted.
Finally, several results are provided so as to numerically assess the accuracy and efficiency of the approximately coupled
methodology discussed in the paper.

RAMOS et al.: QUASI-COUPLED THREE-PHASE RADIAL LOAD FLOW

TABLE VI
SELF AND MUTUAL IMPEDANCES (
) FOR THE 10-NODE SYSTEM

TABLE VII
EXACT SOLUTION FOR THE 10-NODE SYSTEM

APPENDIX
See Tables VVII.
REFERENCES
[1] W. F. Tinney and C. E. Hart, Power flow solution by Newtons method,
IEEE Trans. Power App. Syst., vol. PAS-86, pp. 14491460, 1967.
[2] B. Stott and O. Alsac, Fast decoupled load-flow, IEEE Trans. Power
App. Syst., vol. PAS-93, pp. 859869, 1974.
[3] W. H. Kersting, A method to teach the design and operation of distribution system, IEEE Trans. Power App. Syst., vol. 103, no. PAS-7, pp.
19451952, July 1984.
[4] S. K. Goswami and S. K. Basu, Direct solution of distribution systems,
Proc. Inst. Electr. Eng. C, vol. 138, Jan. 1991.
[5] T.-H. Chen et al., Distribution system power flow analysisa rigid
approach, IEEE Trans. Power Delivery, vol. 6, pp. 11461152, July
1991.
[6] D. Shirmohammadi, H. W. Hong, A. Semlyen, and G. X. Luo, A compensation-based power flow method for weakly meshed distribution and
transmission networks, IEEE Trans. Power Syst., vol. 3, pp. 753762,
May 1988.
[7] G. X. Luo and A. Semlyen, Efficient load flow for large weakly meshed
networks, IEEE Trans. Power Syst., vol. 5, pp. 13091316, Nov. 1990.
[8] D. Rajicic, R. Ackovski, and R. Taleski, Voltage correction power
flow, IEEE Trans. Power Delivery, vol. 9, pp. 10561062, Apr. 1994.
[9] P. A. N. Garca, J. L. R. Pereira Jr., S. Carneiro, V. M. da Costa, and N.
Martins, Three-phase power flow calculations using the current injection method, IEEE Trans. Power Syst., vol. 15, pp. 508514, May 2000.
[10] R. G. Cspedes, New method for the analysis of distribution networks,
IEEE Trans. Power Delivery, vol. 5, pp. 391396, Jan. 1990.
[11] A. G.A. Gmez Expsito and E. R.E. Romero Ramos, Reliable load
flow technique for radial distribution networks, IEEE Trans. Power
Syst., vol. 14, pp. 10631069, Aug. 1999.
[12] E. R.E. Romero Ramos and A. G.A. Gmez Expsito, A general-purpose load flow technique for balanced distribution networks, in
IASTED Conf., Marbella, Spain, Sept. 2000.
[13] R. D. Zimmerman and H. D. Chiang, Fast decoupled power flow unbalanced for radial distribution systems, IEEE Trans. Power Syst., vol.
10, pp. 20452052, Nov. 1995.

781

[14] W.-M. Lin, Y.-S. Su, H.-C. Chin, and J.-H. Teng, Three-phase unbalanced distribution power flow solutions with minimum data preparation, IEEE Trans. Power Syst., vol. 14, pp. 11781183, Aug. 1999.
[15] C. S. Cheng and D. Sjirmohammadi, A three-phase power flow method
for real-time distribution system analysis, IEEE Trans. Power Syst., vol.
10, pp. 671679, May 1995.
[16] K. N. Miu et al., Fast service restoration of large-scale distribution systems with priority customers and constraints, IEEE Trans. Power Syst.,
vol. 13, pp. 789795, Aug. 1998.
[17] A. B. Morton and I. M. Y. Mareels, An efficient-brute force solution
to the network reconfiguration problem, IEEE Trans. Power Delivery,
vol. 15, pp. 9961000, July 2000.
[18] V. Borozan, D. Rajicic, and R. Ackovski, Minimum loss reconfiguration of unabalanced distribution networks, IEEE Trans. Power Delivery, vol. 12, pp. 435442, Jan. 1997.
[19] J. Zhu, G. Bibro, and M.-Y. Chow, Phase balancing using simulated
annealing, IEEE Trans. Power Delivery, vol. 14, pp. 15081513, Nov.
1999.
[20] T. H. Chen and J. T. Cherng, Optimal phase arrangement of distribution
transformers connected to a primary feeder for system unbalance improvement and loss reduction using a genetic algorithm, IEEE Trans.
Power Syst., vol. 15, pp. 9941000, Aug. 2000.
[21] M. E. Baran and M.-Y. Hsu, Volt/Var control at distribution systems,
IEEE Trans. Power Syst. , vol. 14, pp. 312318, Feb. 1999.
[22] M. T. Bishop and R. E. Lee, Distribution system line loss reduction
trough enhanced capacitor location techniques, IEEE Trans. Power
Syst., vol. PS-1, pp. 190197, Apr. 1986.
[23] H. L. Willis, Power Distribution Planning: Reference Book. New
York: Marcel Dekker, 1997.
[24] W. H. Kersting, Distribution System Modeling and Analysis. Boca
Raton, FL: CRC Press, 2002.
[25] G. Kron, Tensorial analysis of integrated transmission systems, Part I,
the six basic reference frames, AIEE Trans., vol. 71, 1952.
[26] H. L. Nguyen, NewtonRaphson method in complex form, IEEE
Trans. Power Syst., vol. 12, pp. 13551359, Aug. 1997.
[27] A. G.A. Gmez Expsito and E. R.E. Romero Ramos, Augmented
rectangular load flow model, IEEE Trans. Power Syst., vol. 17, pp.
271276, May 2002.
[28]. [Online]. Available: http://ewh.ieee.org/soc/pes/dsacom/testfeeders.html

Esther Romero Ramos was born in Spain in 1967. She received the electrical
engineering and Dr.Eng. degrees in 1992 and 1999, respectively, from the University of Sevilla, Sevilla, Spain.
From 1992 to 1993, she was with Sainco, Sevilla, Spain. Since 1993, she
has been with the Department of Electrical Engineering, University of Sevilla,
Sevilla, Spain, where she is currently an Associate Professor. She is interested
in state estimation, load-flow problems, and analysis and control of distribution
systems.

Antonio Gmez Expsito was born in Spain in 1957. He received the electrical
engineering and Dr.Eng. degrees from the University of Sevilla, Sevilla, Spain.
Since 1982, he has been with the Department of Electrical Engineering, University of Sevilla, where he is currently a Professor and Head of the Department.
His primary areas of interest are sparse matrices, load flow, reactive power optimization, state estimation, and computer relaying.

Gabriel lvarez Cordero was born in Spain in 1969. He received the electrical
engineering degree in 1998 from the Univeristy of Sevilla, Sevilla, Spain.
From 1998 to 2000, he was with Isotrol, Sevilla, Spain, and from 2000 to
2003, he was with Norcontrol, Madrid, Spain. Since 2003, he has been with
REE, Madrid, Spain. He is interested in three-phase load-flow problems.

Вам также может понравиться