Sunteți pe pagina 1din 8

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO.

4, NOVEMBER 2005 1775

Analysis of Radial Distribution Systems With


Embedded Series FACTS Devices Using a
Fast Line Flow-Based Algorithm
Ping Yan, Member, IEEE, and Arun Sekar, Senior Member, IEEE

Abstract—Analysis of radial distribution systems with em- modifications to conventional bus-oriented algorithms result in
bedded series Flexible AC Transmission System (FACTS) devices additional iterative loops, since the control variables need to be
is facilitated by a formulation of power flow equations with bus indirectly calculated.
voltage magnitudes and line flows as independent variables. Since
control variables such as the line and bus reactive powers figure The major objectives of FACTS devices installed on a dis-
directly in the formulation, handling the control action of FACTS tribution feeder are to improve voltage profiles, correct power
devices in distribution systems is direct and easily implemented. factor, and reduce line losses. Distribution lines have high
Using the Breadth-First-Search (BFS), the bus incidence matrix R/X ratios, leading to convergence problems in traditional
of a radial distribution system is first rendered strictly upper di- approaches. Modifications and extensions to standard load
agonal, leading to reduced computational effort. All the common
FACTS device models under steady-state conditions are easily flow algorithms with FACTS devices are reported. Salem
incorporated in the new framework by a simple process of “vari- et al. [7] have exploited the analogous model of thyristor-con-
able swapping.” Using IEEE standard systems, the line flow-based trolled-series-capacitor (TCSC) to improve voltage control in a
(LFB) formulation is shown to provide easy implementation with radial distribution system. Garcia et al. [8] derived a modified
multiple series and shunt FACTS devices in the system and enable Newton method in rectangular coordinates by requiring an
direct evaluation of the FACTS device ratings.
augmented Jacobian matrix to incorporate the additional series
Index Terms—Distribution power flow analysis, flexible AC FACTS devices relationships between each control action and
transmission system (FACTS), power system planning and mod- controlled variable. Most of these contributions on FACTS
eling, voltage controls.
device applications to distribution systems focus on using the
Newton–Raphson formulation. The difficulties encountered in
I. INTRODUCTION convergence in such formulations due to the ill-conditioned
nature of the radial system and the high R/X ratios of the
F LEXIBLE AC Transmission System (FACTS) devices are
playing a leading role in efficiently controlling the line
power flow and improving voltage profiles of the power system
distribution line sections are well documented [9].
Distribution power flow methods reported in the literature and
actually implemented prefer to calculate line flows and voltage
network. These new devices can increase the reliability and
magnitudes using forward and reverse sweeps along a radial line
efficiency of transmission and distribution systems. They offer
[10], [11]. Line current and bus voltage phasors with simple
greater flexibility and control in operation. Conventional power
voltage drop calculations in the section impedances enable easy
flow analyses such as Newton–Raphson [1] and Fast Decou-
handling of the highly unbalanced nature of the distribution
pled Algorithm [2] have been adapted to include the FACTS
networks. Luo and Semlyen [12] used branch power variables
device models [3] in transmission systems. Peterson and Meyer
rather than complex current injections and applied tree-labeling
[4] first proposed an efficient method, which combines the
technique in a radial distribution network. Cheng and Shirmo-
state variables of the load tap change (LTC) transformers and
hammadi [13] extended the method to three-phase load flow so-
phase shifters with the bus voltages and their phase angles in
lution with real-time radial distribution system functions. Baran
Newton–Raphson load flow algorithm. The method requires the
and Wu [14] presented power flow equations for the radial distri-
Jacobian matrix to be modified for incorporation of the LTCs
bution system, called Dist-Flow, for determining optimal capac-
and phase shifters. Fuerte-Esquivel and Acha [5], [6] derived
itor sizing and location. Exposito and Ramos [15] use a linear
general admittance-based models for typical FACTS devices
framework embedded in a nonlinear outer loop and show that
and developed a reliable Newton–Raphson load flow algorithm
the convergence rate is enhanced due to the “intuitive idea that
with embedded FACTS device models. Since FACTS devices
the more linear an equation system, the better its convergence
are both series and shunt connected in the power system,
rate.” The results in their paper show how the Newton–Raphson
algorithm fails to converge in a number of cases. Analysis of
Manuscript received November 1, 2004; revised May 16, 2005. Paper no. embedded FACTS devices in line-flow-based (LFB) algorithms
TPWRS-00577-2004.
P. Yan is with American Electric Power, Columbus, OH 43215 USA (e-mail: should be advantageous since the control variables are directly
pyan@AEP.com). available for determination.
A. Sekar is with the Electrical and Computer Engineering Department, The main objective of this paper is to develop an LFB
Tennessee Technological University, Cookeville, TN 38505 USA (e-mail:
arunsekar@tntech.edu). formulation of power balance equations for analyzing a radial
Digital Object Identifier 10.1109/TPWRS.2005.856986 distribution system that will efficiently incorporate embedded
0885-8950/$20.00 © 2005 IEEE
1776 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

series and shunt FACTS devices. The LFB equations use


bus voltage magnitudes and line power flows as independent
variables and directly relate the FACTS device variables with
system operating conditions. The line loss terms are the only
nonlinear terms in the formulation. By adding them to bus
power injections, the coefficient matrix of LFB equations is
rendered linear. A preliminary Breadth-First-Search (BFS)
ordering of the branches transforms the coefficient matrix
structure to strictly upper/lower diagonal and leads to simple
backward/forward substitution for calculating real and reactive
line power in each branch and voltage at each bus. Fig. 1. Generic line segment with FACTS parameters.
The FACTS device models are described first, and the de-
velopment of LFB equations follows. Numerical examples, in- C. SVC
cluding multiple FACTS devices in the standard IEEE systems,
SVC based on thyristors without the gate turn-off capability
illustrate the power of the new approach. The procedure exhibits
is considered as a shunt-connected static VAR generator or ab-
good convergence characteristics, high reliability, and compu-
sorber. Their output is adjusted to exchange capacitive or induc-
tational efficiency. A balanced distribution feeder modeled by
tive current. As an important component for voltage control, it
the positive sequence impedance is used in the paper, since the
is usually installed at the receiving bus. In the formulation, the
aim of this paper is to demonstrate the advantages of the LFB
SVC has been considered a shunt branch with a compensated
formulation in handling the embedded FACTS devices. FACTS
reactive power setting by available inductive and capaci-
devices can be assumed to be cost-effective when deployed on
tive susceptances.
the main distribution feeder.
D. UPFC
II. STATIC MODELING OF FACTS DEVICES
The UPFC can fulfill the multiple functions of power flow
The concept of FACTS first proposed in 1978 is finding control, including reactive shunt compensation, series compen-
increasing appeal in power system planning and operation. sation, and phase shifting. It can independently control real
FACTS devices [16] initiative launched by EPRI resulted in power flow, reactive power flow, and voltage magnitude by
the development of series and shunt devices: series compen- being integrated into a generalized power controller combining
sators such as Thyristor Controlled Series Capacitor (TCSC), the functions of TCSC, TCPAR, and SVC, so that the UPFC
Thyristor Controlled Voltage Regulator (TCVR), Thyristor controls the magnitude and angular position and injection
Controlled Series Reactor (TCSR), and Thyristor Controlled voltage in real time. The aim is to maintain or vary the real
Phase Angle Regulator (TCPAR) and shunt compensators such and reactive power flow in the line to satisfy load demand and
as Static Var Compensator (SVC) and STATCON. The com- system operating conditions. The steady-state model of the
bined series and shunt FACTS’s sample is Unified Power Flow UPFC is made from the basic series and shunt FACTS models.
Controller (UPFC) [17]. To develop the static models of most
FACTS devices, TCSC, TCVR, SVC, and UPFC are specially E. Generic Line Segment With FACTS Parameters
discussed in this section and summarized into an integrated Fig. 1 shows the generic model of a line segment embedded
FACTS model. with series-class and shunt-class FACTS devices. Depending
upon the FACTS device type, its function in the generic model
A. TCSC can be easily implemented. Although designation of sending
TCSC is defined as a capacitive reactance compensator, and receiving ends is arbitrary, line flow directions are indi-
which consists of a series capacitor bank shunted by a cated close to the receiving end. A transformer with tap and
thyristor-controlled reactor in order to provide a smoothly phase angle parameters is located at the sending end. Addition-
variable series capacitive reactance. In the steady-state power ally, and are resistance and reactance of branch ; and
flow study, the TCSC can be considered as a static capacitor or are sending and receiving bus numbers of branch ; and
reactor offering a reactance with a series compensated and are active and reactive power flow at the receiving end of
branch represented by lumped equivalent series parameters branch . The generic line model is used to derive line voltage
connected. In most cases, the shunt susceptances of a branch drop equations in the LFB formulation. The phase angle param-
usually are neglected. Therefore, the TCSC’s static capacitor eter is ignored since it is has no significance in radial distribution
will be directly in series with the line impedance. systems.

B. TCVR III. LFB RADIAL DISTRIBUTION POWER FLOW MODEL


TCVR is considered as the common voltage regulator. It is Since line flows and bus voltage magnitudes are of practical
able to smoothly vary voltage magnitude with a tap changing in importance in the operation of a distribution system, and the
the control range of . A static model of FACTS devices control these quantities, a power flow model
TCVR with a tap ratio is connected in a series impedance based on real and reactive line power flows and bus voltage mag-
of the distribution line. nitudes will enable easy handling of series and shunt devices.
YAN AND SEKAR: ANALYSIS OF RADIAL DISTRIBUTION SYSTEMS 1777

Real and reactive power balance equations at all buses except Taking the magnitude of both sides of (6) and rearranging, we
the slack bus can be written using the incidence matrix of the get the following:
network graph [18]. Since all shunt connections are excluded
in the incidence matrix, their real and reactive power contribu-
tions are accounted for separately in the power balance equa-
tions. Real and reactive power loads, shunt capacitors, and line (7)
charging susceptances can be treated as shunt branches. Fol-
lowing the traditional classification of slack, voltage-controlled, Dividing (7) and setting and
and load buses, the LFB equations are formulated as three sets of
, we get
equations viz., bus real and reactive power balance and branch
voltage equations.
(8)
A. General Power Balance Equations
Using a bus incidence matrix with rows corresponding to The right side of (8) is written with .
all busses other than the slack, the bus real and reactive power There are a total of line voltage equations. The total number of
balance can be written as follows: unknown voltage magnitudes will drop to when
dispersed generators are specified as voltage-controlled buses.
(1) Because the voltage of a PV bus is known, the value is shifted
(2) to the right-hand side of (8). Equation (8) can be written in a
matrix form as
is defined as a modified bus incidence matrix with all
“ ” in set to zero, which makes it easy to include the line (9)
losses in the power balance equations by using the vectors of
branch real and reactive power losses and . is a diag- where is a bus incidence matrix corresponding to the PV
onal matrix, whose diagonal elements are sums of charging and buses. is a vector of square of voltage of PV buses and
compensating susceptances at each bus. and are the the slack bus. is a diagonal matrix of order equal to with
bus injection power vectors defined as and the values of a tapped transformer equal to the square of the tap
, where , , and are ac- value. and are obtained from by setting, respec-
tive and reactive generator and load powers at bus , and tively, the negative and positive values in to zero. and
are active and reactive power losses in line . and are real are diagonal line resistance and reactance matrices. The vector
and reactive line flow vectors at the receiving end. is the un- represents the term on the right side of (8) for all the lines.
known voltage vector except at the slack bus. is the square
of voltage magnitude at bus . C. LFB Power Flow Equations
If dispersed generators exist in the distribution system, the
Equations (1), (2), and (9) are put in matrix form as follows:
corresponding generator buses are classified into two kinds
viz., constant injected power PQ buses or voltage-controlled
PV buses. Let be the total number of buses and and
the number of voltage-controlled and load buses, respectively.
Allowing for one slack bus, the total number of various buses is

(3) (10)

The number of unknowns in (2) is reduced from to


, and the equation is rewritten as This is the LFB power flow equation. The total number of
unknowns is
(4)
(11)
where all symbols with subscript 1 are the relevant matrices or
vectors containing PQ buses only. Equation (10) is written in symbolic form as follows:

B. Branch Voltage Equations (12)


Referring to Fig. 1, the branch voltage drop equation can be
written as The coefficient matrix is a constant matrix that needs
to be factorized only once during the solution process. The co-
efficient matrix is nonsingular since the radial network is of tree
(5)
structure with all the branches connected, and the rows cor-
responding to tree branches in a network graph are indepen-
(6) dent [18]. is the variable vector of active, reactive line flows
1778 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

Fig. 2. IEEE 13-node test feeder.

, and square of bus voltage magnitudes ; is a con-


stant vector of the injection powers and PV bus voltages;
is a variable vector of branch losses and charging and compen-
sating powers. An iterative power flow analysis is started with Fig. 3. Graph of the IEEE 13-node feeder.
the vector set to zero. Using the solution vector in an itera-
tion, the right-hand side of (12) is updated by calculating vector
and adding to vector . Patel [19] has documented exten-
sive testing of the LFB formulation for interconnected systems
showing good convergence behavior of such an iterative proce-
dure, even in highly ill-conditioned systems.

IV. NATURE OF RADIAL DISTRIBUTION NETWORK GRAPH


The radial distribution network graph has a tree structure with
no loops. The total number of lines equals the number of buses
minus one. Since the pattern of incidence matrix depends on the
order of lines and nodes, the incidence matrices in (10) have a
structure that depends on the order in which the lines are read
from the data. Further, the incidence matrices are square and
nonsingular. By reordering the lines appropriately, the incidence
matrix can be made purely upper-diagonal. Cheng and Shirmo- Fig. 4. Arbitrary order incidence matrix.
hammadi [13] demonstrate this by “ordering layers away from
the root node (source bus).” Zhang and Cheng [20] show that this An example of using the BFS algorithm is illustrated using
results in an upper triangular incidence matrix. This will lead the radial distribution system with thirteen nodes and twelve
to gain in computation time in power flow solution, as shown lines of [22] and [23] shown in Fig. 2. Lines have only series
later. The BFS algorithm used to achieve this is explained in impedances. The nodes are numbered arbitrarily. This example
this section. distribution network graph is redrawn in Fig. 3 with bus num-
Although BFS and Depth-First Search [21] are two different bers in sequence starting from 1. Although no. 1 is given to the
patterns of tree search methods of graph in widespread use, here only source bus here, the others encircled are arbitrary num-
BFS is selected to search radial distribution network graph in bers as given in the original data list. This is useful when the
the LFB model. The basic idea of BFS is to point out to as network is reconfigured to meet the demand under different
many buses as possible before penetrating deep into a tree. This load and feeder scenarios. The flows in the branches are al-
means that we visit all the buses adjacent to the current level ways oriented away from the source node, and so the direction
before going on to another one. The brief description of BFS arrows are ignored.
to renumber buses and branches may be summarized in the fol- The incidence matrix in Fig. 3, including the source node (at
lowing three steps for building an optimal BFS tree. root, originally named as 650) called Bus 1, is shown in Fig. 4.
1) Start at source bus as the first level and fan out to the To conform to LFB load flow equations of the later section,
“downstream” buses as the next levels. the rows of the matrix are related to buses and its columns to
2) On the same level, all bus numbers are ordered consecu- branches. The BFS renumbering is applied to the 13-node test
tively. feeder. The optimal BFS tree of the IEEE 13-node feeder is
3) Branch renumbering is similar to that of the bus renum- shown in Fig. 5. Its reordered incidence matrix, including the
bering. At any level, a branch number is one less than the source node, is shown in Fig. 6. After the row corresponding to
“upstream” bus number. the source node is deleted from Fig. 6, bus incidence matrix
YAN AND SEKAR: ANALYSIS OF RADIAL DISTRIBUTION SYSTEMS 1779

V. DECOUPLED LFB MODEL OF RADIAL


DISTRIBUTION SYSTEMS

The matrix (10) may be split into three “decoupled” equation


sets and rewritten as follows:

(14)
(15)
(16)

With BFS ordered lines, the bus incidence matrix as (13)


and are upper triangle matrixes and is a lower triangle
matrix in nature. Using flat voltage profile as starting value, the
line flows and may be directly obtained by its backward
substitution of (14) and (15). Using the updated and into
(16), the square voltage magnitudes can be calculated by
Fig. 5. BFS numbering tree of the IEEE 13-node system. the forward substitution. There is no need for factorization of
the coefficient matrix. The decoupled set of equations is easier
to code in a program.
The convergence behavior and computation time required
of the LFB formulation are studied and compared with the
Newton–Raphson algorithm. IEEE 13-, 34-, and 123-node sys-
tems are calculated using the Newton–Raphson algorithm and
four variations of the LFB formulation. Neglecting the loss
term on the right side of (12) results in an approximate
linear model. The solution is obtained without iterations. This
model is called linear LFB. The other variations are the LFB
model, including the loss term and the decoupled formulation
(D-LFB). Calculations are carried out with a PC. For a con-
vergence criterion of , all three iterative methods took
four iterations in each case. Table I shows the computation
time in the various cases. For comparing the accuracy of the
results, the maximum absolute error and the mean square de-
Fig. 6. Reorder incidence matrix for Fig. 5. BFS optimal tree.
viation of the bus voltage values are calculated in the new
formulations using the Newton–Raphson results as accurate
becomes the upper triangle matrix in (13). This sparse upper tri- ones. Table II shows the results from the three variations for
angle incidence matrix results in reducing computational effort the various cases studied. The conclusion is that the proposed
during the iterative process, shown in (13) at the bottom of the LFB formulation is computationally superior and acceptably
page. accurate.

(13)
1780 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

TABLE I
COMPARISON OF CONVERGENCE COMPUTATION TIMES

TABLE II
ACCURACY COMPARISONS BASED ON THE FULL AC METHOD

Fig. 7. Voltage profiles after and before FACTS installed on the IEEE 13-node
test feeder.

TABLE III
SUMMARY OF FACTS TYPE AND IMPLEMENT IN IEEE 13-NODE FEEDER

VI. LFB-BASED FORMULATION WITH


EMBEDDED FACTS DEVICES
B. Bus Voltage Control Using Transformer Tap or TCVR
FACTS devices use shunt Q injections, tap changing, and se-
ries capacitors as control variables to control the bus voltages The objective of changing the transformer tap is to achieve bus
and line reactive power flows. In the proposed LFB-based de- voltage control. In the generic model in Fig. 1, the transformer is
coupled power flow, the unknown vectors are , , and . The usually inserted into a branch close to the sending end. Equation
known vectors include bus injections, tap values, and square of (8) corresponding to line will be rearranged as follows:
generator bus voltages. Depending upon the type of FACTS de-
vice, the formulation permits “swapping” or exchanging an un- (19)
known variable with a known value and enables the calculation
of FACTS device settings. Three representative case studies are and become the unknown variables in this equation.
made with different types of FACTS devices included in line Equation (19) replaces the original equation for line in (16).
between bus and in a distribution system. The method of The value can be treated as the new variable replacing
“variable swap or exchange” is used in each case. SVC and in the unknown vector . There is no change in the overall
UPFC functions can be accomplished by integrating the three structure of the equations. The right-hand side is to be computed
individual formulations. using (19) for the line with the transformer tap, and the solution
corresponding to in the original equation is interpreted as
A. Line Flow Control Using Series Capacitor in TCSC the product .
In this case, the voltage magnitude at bus will be fixed as C. Bus Voltage Control With Q Injection in SVC
the controlled voltage and the controlling capacitor added to the
reactance matrix in (16). Equation (8) corresponding to line In SVC, changing the reactive power Q injection at a bus con-
will be rearranged as follows: trols the bus voltage magnitude. Equation (16) will have one less
unknown, since bus voltage is controlled to a specified value.
(17) The bus will now be treated as a PV bus, and the Q injection can
be calculated using the line flows coming into the controlled bus.
and become the unknown variables in this equation,
VII. TEST RESULTS
which is rewritten as follows:
IEEE 13- and 34-bus test systems form the basis for testing the
(18) LFB formulation. The decoupled LFB procedure is applied to the
IEEE 13-node test feeder of Fig. 2 with the addition of a single
Equation (18) replaces the original equation for line in (16). series FACTS device and the IEEE 34-node test feeder modified
The product can be treated as the new variable replacing with multiple FACTS devices. All lines have only resistance and
in the unknown vector . As can be seen, there is no change reactance values. Base load flow studies of the example systems
in the overall structure of the equations. The right-hand side is without any FACTS devices are calculated. By including the
to be computed using (18) for the line with the series capacitor. FACTS devices, the effect on the voltage profiles is then studied.
The solution corresponding to in the original equation is to All parameters of the modified systems come from [22] and
be interpreted as the product . [23]. All convergence tolerances are selected as .
YAN AND SEKAR: ANALYSIS OF RADIAL DISTRIBUTION SYSTEMS 1781

Fig. 8. IEEE 34-node test feeder.

TABLE IV
IEEE 34 -NODE NUMBERINGS DEFINED BY BFS AND THE ORIGINAL

TABLE V
DEVICES IMPLEMENT AND LOCATIONS

To control the voltage at bus 2 of the IEEE 13-node feeder to


1.05 p.u., a single series FACTS device, either TCSC or TCVR,
is considered in line 1–2 on Fig. 5 to realize this control objec-
tive. The solution converged in four iterations. Table III summa- Fig. 9. Comparison of FACTS’ effect on voltage profiles.
rizes the resulting FACTS device parameters to realize the con-
trol objectives. The investigation for these two type FACTS to
with a TCSC controlled by series capacitor is not close to zero.
realize the same control objective shows almost the same values
The effects on the specified voltage profiles are shown in Fig. 9.
of their voltages of each bus. In addition, their max voltage dif-
ference at the same bus is below 0.005(p.u.). The voltage pro-
VIII. CONCLUSIONS
files in Fig. 7 demonstrate the effect of the FACTS device on
improving the voltage profile. FACTS devices offer a flexible and comprehensive solution
The LFB approach presented in this paper handles multiple to voltage profile control in distribution systems. Traditional
FACTS devices in a system with equal ease. The IEEE 34-bus methods such as voltage regulators, series, and parallel capacitor
radial distribution system configuration of Fig. 8 is used to study switching can be adapted to work more effectively with FACTS
the effect of multiple FACTS devices embedded in the system. control.
BFS numbering is listed in Table IV along with the original Methods of power flow analysis of distribution networks are
numbering in the diagram. The IEEE 34-bus system is modi- generally different from the Newton–Raphson algorithm used
fied with two FACTS controllers with both of TCSC, TCVR, in interconnected system analysis. This paper proposes a line
or one TCSC and another TCVR. The results of FACTS device flow-based formulation, which uses line power flows as vari-
parameters are summarized in Table V, which lists FACTS in- ables instead of line currents. Since line reactive flows and bus
stallation locations and control objectives. The LFB approach voltage magnitudes are of interest in controlling voltage pro-
easily converged as long as the equivalent reactance of the line files, the line flow-based algorithm provides a direct and simple
1782 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4, NOVEMBER 2005

approach to handle single or multiple devices. This will enable [7] M. R. Salem, L. A. Talat, and H. M. Soliman, “Voltage control by tap-
easy determination of size and placement of such devices. changing transformers for a radial distribution network,” in Proc. Inst.
Elect. Eng., Gener., Transm., Distrib., vol. 144, Nov. 1997, pp. 517–520.
The line flow-based equations have constant coefficient [8] P. A. N. Garcia, J. L. R. Pereira, and S. Carneiro, “Voltage control de-
matrices and need no updating during the iterative procedure. vices models for distribution power flow analysis,” IEEE Trans. Power
Breadth first search of the topology of the radial network Syst., vol. 16, no. 4, pp. 586–593, Nov. 2001.
[9] C. Trevino, “Cases of difficult convergence in load flow studies,” in Proc.
leads to a coefficient matrix structure of LFB equations that is IEEE Paper 71-62-PWR, Summer Meeting, Los Angeles, 1970.
already in a triangular form, which results in just one forward [10] W. H. Kersting, Distribution System Modeling and Analysis. Boca
or backward substitution in each iteration. The BFS algorithm Raton, FL: CRC, 2002.
[11] DEWorkstation, Electric Power Research Institute, Palo Alto, CA.
can handle very large systems at high speed and is the preferred [12] G. X. Luo and A. Semlyen, “Efficient load flow for large weakly meshed
method used in web search engines [24]. This paper proposes a networks,” IEEE Trans. Power Syst., vol. 5, no. 4, pp. 1309–1316, Nov.
decoupling of the LFB equations, which reduces the computa- 1990.
[13] S. Cheng and D. Shirmohammadi, “A three phase power flow method
tion time. The new formulation is shown to be equally accurate for real time distribution system analysis,” IEEE Trans. Power Syst., vol.
as the full AC analysis from test runs on IEEE standard systems. 10, no. 2, pp. 671–679, May 1995.
The convergence behavior is same as the full AC analysis. [14] M. E. Baran and F. F. Fu, “Optimal sizing of capacitors placed on a radial
distribution system,” IEEE Trans. Power Del., vol. 4, no. 1, pp. 735–742,
This paper demonstrates the ease with which single or mul- Jan. 1989.
tiple FACTS devices can be handled by the LFB formulation. [15] A. G. Exposito and E. R. Ramos, “Reliable load flow technique for radial
Simple “variable exchange or swap” requiring little change to distribution networks,” in Proc. IEEE PE-344-PWRS-0-12, 1996.
[16] J. Gotham and G. T. Geydt, “FACTS device models for power studies,”
the overall solution methodology enables the evaluation of the in Proc. 27th NAPS, Bozeman, MT, Oct. 2–3, 1995, pp. 514–519.
control parameters. [17] L. Gyugyi, T. R. Rietman, S. L. Willians, T. R. Rietman, D. R. Torgerson,
IEEE 13- and 34-node standard distribution systems are used and A. Edris, “The unified power flow controller: A new approach to
power transmission control,” IEEE Trans. Power Del., vol. 10, no. 2, pp.
to illustrate the method of evaluating the FACTS device param- 1085–1097, Apr. 1995.
eters. Single and multiple FACTS devices are considered. Both [18] G. W. Stagg and A. H. El_Abiad, Computer Methods for Power System
series and shunt devices are employed to improve the voltage Analysis. New York: McGraw-Hill, 1971.
[19] J. R. Patel, “Power flow analysis using modified tableau formulation
profile. The required device parameters, such as the series ca- of circuit equations,” Master’s thesis, Tennessee Technological Univ.,
pacitance and the shunt VARs, are directly determined. Cookeville, TN, 1992.
Commercially available distribution analysis software using [20] F. Zhang and C. S. Cheng, “A modified newton method for radial distri-
bution system power flow analysis,” IEEE Trans. Power Syst., vol. 12,
line currents as variables can be easily modified for LFB formu- no. 1, pp. 389–397, Feb. 1997.
lation since the data structure is the same in both cases. [21] R. J. Wilson and J. J. Watkins, Graphs: An Introductory Ap-
Work is in progress in formulating the optimization problem proach. New York: Wiley, 1990.
[22] W. M. Kersting and L. Willis, “Radial distribution test systems,” IEEE
for determining the “best” manner of deploying the FACTS de- Trans. Power Syst., vol. 6, no. 3, pp. 975–985, Aug. 1991.
vices to achieve the desired results in practical situations. This [23] IEEE Distribution Feeder Test Result. IEEE Distribution System Anal-
paper assumes balanced distribution feeder for developing the ysis Subcommittee. [Online]. Available: http//ewh.ieee.org/soc/pes/
dsacom/testfeeders.html
procedure. The authors are working on extending the procedure [24] M. Najork and J. L. Wiener, “Breadth-first search crawling yields high-
to unbalanced feeders. quality pages,” in Proc. ACM 1-58 113-348-0/01/0005, WWW10, Hong
Kong, May 2–5, 2001.
REFERENCES
[1] W. F. Tinney and C. Z. Hart, “Power flow solution by Newton’s method,”
IEEE Trans. Power App. Syst., vol. 86, no. 11, pp. 1449–1456, Nov.
1967.
[2] B. Stott and O. Alsac, “Fast decoupled load flow,” IEEE Trans. Power Ping Yan (S’98–M’01) received the Ph.D. degree from the Electrical and
App. Syst., vol. 93, no. 3, pp. 859–869, May/Jun. 1974. Computer Engineering Department, Tennessee Technological University,
[3] N. G. Hingorani and L. Gyugyi, Understanding FACTS: Concept and Cookeville, and holds a PE license from the state of Ohio.
Technology of Flexible AC Transmission Systems. New York: IEEE She is now with American Electric Power, Columbus, OH. She was with Elec-
Press, 1999. tric Power System Management, Asian Institute of Technology, Bangkok, Thai-
[4] N. M. Peterson and W. Meyer, “Automatic adjustment of transformer land, as a Research Associate and Zhejiang Power System Bureau, Hangzhou,
and phase-shifter taps in the Newton power flow,” IEEE Trans. Power China, as a Power System Engineer and Lecturer.
App. Syst., vol. PAS-90, no. 3, pp. 103–108, Jan./Feb. 1971.
[5] C. R. Fuerte-Esquivel and E. Acha, “Newton–Raphson algorithm for
the reliable solution of large power networks with embedded FACTS
devices,” in Proc. Inst. Elect. Eng., Gener., Transm., Distrib., vol. 143,
Sep. 1996, pp. 447–454.
[6] , “A Newton-type algorithm for the control of power flow in elec-
trical power networks,” IEEE Trans. Power Syst., vol. 12, no. 4, pp. Arun Sekar (M’84–SM’88) is a Professor of electrical and computer engi-
1474–1480, Nov. 1997. neering at the Tennessee Technological University, Cookeville.

S-ar putea să vă placă și