A Modified Backward/forward Method For Fast Solving Radial Distribution Networks
A Modified Backward/forward Method For Fast Solving Radial Distribution Networks
- U
n-1
-U
n
=
= U
o
-U
1
-U
2
- - U
n-1
(10)
If the loads power depends on the voltage, and if the lines
shunt admittances are considered, the generic branch of the
feeder in fig. 1 can be represented as series impedance with
two shunt admittances at the ending nodes, each having value
of half the total shunt admittance of the branch, fig. 2.
Fig. 2 Circuit representation of a generic network branch.
The solution of the system can be carried out by means of
an iterative process in which the network is examined from
the terminal nodes; the iterative process is organised in the
following steps:
1 the voltages at all the nodes are initialised at the value U
o
;
2 on the basis of the known values of the nodal voltages the
loads currents and those derived on the shunt admittances of
the branches derived from the ending nodes are calculated;
3 the currents circulating in the network branches are
calculated;
4 starting from the ending branch, to whose ending node the
voltage U
o
is given, and going up to the root node, all the
voltage nodes are calculated, U'
j
, also considering that at the
supply node, U'
o
;
5 the difference U
o
=U'
o
-U
o
is evaluated;
6 each nodal voltage, U'
j
, is decreased of the quantity U
o
;
7 the error, given by the magnitude of U
o
, is compared
with a prefixed value ; if err =|Uo| < the process stops,
otherwise the process starts again from step 2.
If the network has some branching nodes, namely nodes
from which more than two branches spread out, fig. 3,
following the procedure above detailed, from the terminal
nodes up to the root node, some problem would arise since
each branching node should get as many voltage values as
many branches there are below it.
For this kind of networks, the procedure here proposed
includes a pre-processing step in which a main feeder and
some laterals are identified, fig. 4; the node origin of each of
these laterals is a branching node of the original system (nodes
2, 4 and 8 for the network in fig. 3). In this way, both the main
feeder and the laterals have the structure of a single line with
many lumped loads without laterals. Of course, else than the
root node for the laterals, some of their internal nodes, both of
the main feeder and of the laterals can be branching nodes; in
these nodes, in addition to the current derived from the load at
the node, the currents circulating in the branches derived from
this node (not belonging to the feeder or the lateral) must be
considered. The division of the original system in a main
feeder and many laterals can be done following different
criterions; besides, the choice for one or another criterion does
not influence, in the solution methodology here developed,
neither the speed of convergence nor the total CPU time. Each
branch, each node and each branching node indeed, is
considered only once and these sets do not vary as compared
to the sets of possible main feeders and of possible laterals
obtainable from a given system. As a result, for networks
having laterals the steps of the procedure become:
1 for all the network nodes the voltages are put to U
o
;
2 the load currents and those derived on the shunt
admittances are calculated;
3 the laterals not having internal branching nodes are
Fig. 3 Radial network; the nodes numbering is done so that the id number j of any node is greater than the id number of any node above node j; the numbering
is carried out giving to any branch the id node of the relevant EB.
Fig. 4 - Identification of a main feeder (FP) and of the laterals (FL
1
FL
3
) for
the radial system in fig. 3.
considered and starting from the terminal nodes, the current in
all the branches are calculated; in this way, the current on the
branch whose SB is the branching node also belonging to the
main feeder or to another lateral, and supplying the lateral is
calculated;
3' the laterals having internal branching nodes are
considered; those having as internal branching nodes of the
SB of the laterals treated at step 3 are first considered. For
these laterals, starting from the terminal nodes, the currents in
all the branches are calculated; for the branches whose EB is a
branching node, the current circulating in the branch is
evaluated adding to the current in the branch below that of the
load and that of the first branch of the relevant lateral
considered at step 3; in this way, the current in the branch
connected to the root node of the lateral can be obtained;
3" step 3' is repeated for all the laterals having internal
branching nodes; therefore all the branch currents flowing
from the branching nodes on the main feeder are known;
4 the currents on all the branches of the main feeder are
evaluated;
5 for the nodes on the main feeder the voltages are
calculated, starting from the terminal node (having voltage U
o
in the first iteration or the value calculated at the following
step 7 in the other iterations) and going up to the supply node;
6 once the root node voltage is calculated, U'
o
, the difference
with the imposed voltage value U
o,imp
, U
o
=U'
o
-U
o,imp
is
calculated;
7 the voltages of all the nodes belonging to the main feeder
are reduced of the quantity U
o
;
8 starting from the branching nodes belonging to the main
feeder and going on the laterals branches, starting from the
branch connected to the main feeder and going towards the
terminal branches, (following a reverse order in the
examination of the lateral feeders done at steps 3, 3' and 3")
the voltages of all the nodes of the lateral are calculated;
9 once all the laterals have been considered the convergence
criterion is tested comparing the magnitude of U
o
with a
prefixed ; if |U
o
|< , the process stops, otherwise the
procedure is repeated from step 2.
From what was said above, the process does not
substantially differ from the classical backward/forward;
besides, the methodology has the following advantages:
- during the backward phase the node voltages of the laterals
are not evaluated, only those of the root node and of the
internal nodes of the main feeder are considered;
- the convergence criterion is based on the evaluation of the
error in the calculation of the voltage of the root node; this
error is defined as the difference between the value calculated
starting from the terminal node of the main feeder and the
presumed value for the same node:
err=|U
o
|=|U'
o
- U
o
| (11)
- the speed of convergence can be increased by means of the
updating (during the backward phase) of the load currents. In
this aim, consider the line in fig. 5 showing the main feeder or
one of the laterals. Since the bus voltages evaluations is done
starting from the terminal branch up to the root node, giving
the value U
o
to the terminal node n, for the voltage at node n-1
we have:
U
n-1
=U
o
+ U
n
(12)
Where the voltage drop U
n
is given by:
U
n
= Z
n
J
n
= Z
n
(I
Ln
+ I
cn
) (13)
The current I
cn
is given by:
I
cn
= U
n
Y
cn
(14)
Where Y
cn
is half of the total shunt admittance of the branch n
(for the internal nodes of the feeder this admittance is given
by the summation of the shunt admittances relevant to the
branches also having the considered node as terminal node).
In the hypotheses that for the loads it is valid the constant
power load model, the current required by the load at node n,
I
Ln
, is given by:
I
Ln
= S
n
/U
n
= S
n
/U
o
(15)
The evaluation of the load current on the basis of the rated
voltage U
o
gives a value that is certainly lesser than that
related to the final solution, U
n
(<U
o
); the evaluation of this
current is more precise if one gives to the supply voltage at
node n a value that is closest to the actual one, namely taking
into account the voltage drop on branch n; on the basis of this
assumption, the current on the last branch of the main feeder
or of the lateral can be evaluated with the following
expression:
J
n
= I
L
n
+ I
c
n
= S
n
/(U
n-1
- Z
n
J
n
) + U
n
Y
c
n
=
= S
n
/(U
n-1
- Z
n
(S
n
/U
n
)) + U
n
Y
c
n
(16)
For branch n-1, for voltage calculation at node n-2, the
current circulating in the branches is given by:
J
n-1
= J
n
+ S
n-1
/(U
n-2
- Z
n-1
(J
n
+ S
n-1
/U
n-1
)) + U
n-1
Y
c
n-1
(17)
Fig. 5 Main or lateral feeder representation, for load current calculation.
If node n-1 is a branching node, the load current of the load
derived at node n-1 must be summed up with the currents
circulating on the first branches of the laterals having as root
node the node n-1. As a result, the current circulating in a
generic branch inside the feeder (or the lateral), not
considering the terminal branch, is given by:
J
n-1
= J
n
+ A
n-1
/(U
n-2
- Z
n-1
(J
n
+ S
n-1
/U
n-1
)) +
+ U
n-1
Y
c
n-1
+ I
lat
n-1
(18)
Where the last term expresses the summation of the
currents derived towards the laterals that spread up from node
n-1. The expression (18) links the current in the generic
branch with that in the branch below and with the voltages at
the nodes of the same branch; for the generic branch j, the
voltage of the bus that is farthest from the source node, EBj, is
already known, whereas that of the node that is closest to the
source node, SBj, can be evaluated through the expression:
U
SBj
=U
EBj
+ Z
j
J
j
(19)
Keeping into account the latter updating, the steps of the
procedure are modified in the following way:
1 the voltages of all the terminal nodes of the main feeder
and of the laterals are set to U
o
;
2 the laterals not having other branching nodes except the
relevant supply node are considered; starting from the
terminal branches and using eqn. 16 (for the terminal branch)
and 17 (for the other branches), the currents in all the
branches are calculated; at the same time also the voltages of
all the internal nodes of the feeder are calculated using eqn.
19; in this way the current on the first branch of the lateral is
obtained, i.e. that connected to the supply node (which is a
branching node of the original system) as well as the voltage
in the supply node of the lateral;
3' starting from the terminal branches, the currents in all the
branches of the laterals having as internal nodes the branching
nodes that are the supply nodes for the laterals examined at
the preceding step (for these laterals, in correspondence of
these nodes the load current must be summed up with the
branch current evaluated at the preceding step) are evaluated
using eqns. 16 and 18; at the same time the voltage at all the
nodes are calculated; in this way, the current in the branch
connected to the root node is obtained as well as the voltage at
the same node;
3" - step 3' is repeated until all the laterals are considered; the
currents in the branches derived from the branching nodes on
the main feeder are all known and all the voltages of the
supply nodes of the laterals are also known. The values of
these voltages are saved into a vector having a size which
equals the number of laterals obtained from the original
system; they represent the voltages of the supply nodes of the
laterals evaluated starting from the currents circulating in the
lateral as a result of a given voltage value in the terminal node
of the same lateral;
4 the main feeder is then considered; following the same
procedure of the preceding step, the currents in all the
branches and the voltages at all the nodes are calculated; in
particular, at the end of this step, the voltages of the source
node and of the branching nodes belonging to the main feeder
of the original systems are evaluated. Of course the voltage
value of the root node, U'
o
, does not match with the value
imposed at the same node and, in a similar way, the values of
the voltages in the branching nodes on the main feeder do not
match with those calculated at the end of step 3" for the same
nodes considered as supply nodes for the laterals;
5 for the supply node of the system the difference between
the calculated value, U'
o
, and the imposed value, U
o
, is
calculated, U
o
=U'
o
-U
o
;
6 the voltages of all the nodes belonging to the main feeder
are reduced of the quantity U
o
; the values of the voltages so
calculated represent the actual values at the end of the current
iteration; in this way, the values of the voltages in the
branching nodes belonging to the main feeder are fixed;
7 for the laterals whose supply nodes belong to the main
feeder, the differences between the voltage values calculated
at step 3" and those evaluated at the previous step for the
relevant supply nodes are calculated; the voltages of all the
nodes of these laterals are modified subtracting this difference
to the values calculated at step 3; in this way the voltages of
the branching nodes in the laterals are fixed;
8 for the laterals whose supply nodes belong to the laterals
examined at the preceding step, the same evaluations already
executed are performed for the calculation of the node
voltages. In this forward phase, the main feeder and laterals
are examined in the reverse order as compared to that
followed in the backward phase;
9 once all the laterals have been examined the convergence
criterion is tested comparing the magnitude of U
o
with a
prefixed ; if |U
o
|< , the process stops, otherwise goes back
to step 2 of the procedure.
III. APPLICATIONS
The developed method has been applied to the solution of a
radial system with 33 buses, whose electrical data are reported
in [13]; the implementation has been carried out in
FORTRAN 90 language and the program run on a Mainframe
IBM S/390-2003/225. The division of the original system into
a main feeder and some laterals has been done following the
criterion of selecting always the feeder with the maximum
number of branches between the supply node and the terminal
nodes. For the identification of the main feeder the supply
node is the root node of the entire system, whereas, for the
identification of the laterals, the relevant supply nodes are the
branching nodes belonging to the main feeder or to the just
considered laterals. This process orders all the feeders in the
correct sequence in the aim of the application of the proposed
algorithm; in the backward phase the feeders are considered
starting from the last selected, whereas in the forward phase,
the feeders are examined in the same sequence adopted during
their selection. The results of the applications, in terms of
number of iterations and CPU time (the CPU times reported in
what follows are averaged out of 10 runs for each case
examined),
TABLE. I NUMBER OF ITERATIONS REQUIRED TO SOLVE THE 33 BUSES RADIAL SYSTEM WITH THREE DIFFERENT METHODOLOGIES FOR ALL THE COMBINATIONS
OF THE TOTAL LOAD INCREASE, OF THE REACTIVE LOAD AND OF THE R/X RATIO OF THE LINES.
A 0.5 0.5 0.5 1 1 1 1.5 1.5 1.5 2 2 2
Q NR B/F
M-B/F
NR B/F
M-B/F
NR B/F
M-B/F
NR B/F
M-B/F
0.5 0.5 4 4 4 4 4 4 4 5 5 4 5 5
0.5 1 4 4 4 4 4 4 4 5 5 5 6 6
0.5 1.5 4 4 4 4 5 5 5 6 5 5 6 6
0.5 2 4 4 4 4 5 5 5 6 6 5 7 7
1 0.5 4 4 4 4 5 5 4 6 5 5 7 6
1 1 4 4 4 4 5 5 5 6 6 5 7 7
1 1.5 4 4 4 4 5 5 5 7 7 5 8 8
1 2 4 4 4 5 6 6 5 8 7 6 11 10
1.5 0.5 4 4 4 4 5 5 5 7 6 5 8 8
1.5 1 4 4 4 5 6 6 5 7 7 5 10 10
1.5 1.5 4 5 5 5 6 6 5 8 8 6 14 13
1.5 2 4 5 5 5 7 7 6 10 10
2 0.5 4 5 4 5 6 6 5 8 8 6 12 12
2 1 4 5 5 5 7 6 5 9 9 6 19 18
2 1.5 4 5 5 5 7 7 6 12 11
2 2 4 5 5 5 8 8 6 19 18
have been compared with those obtained solving the network
with the Newton-Raphson method and with the classical
backward/forward. Assuming the network data as rated
values, the following data have been modified: the load factor,
A
(=S/S
n
), the reactive load factor,
Q
(=Q/Q
n
), and the
incremental factor of the lines resistance,
R
(=R/R
n
), giving
to these indices respectively the values 0.5; 1; 1.5 and 2. In
Tab. 1, for each set of values assumed by the factors, the
number of iterations required to reach the solution with a
convergence factor of 0.00001, applying the Newton-Raphson
(N-R) method, the backward/forward (B/F) method and the
proposed method (M-B/F) is reported; Tab. 2 reports, for each
of the three methods and for each number of iterations, the
CPU time required. The results indicate that the proposed
methodology in many cases produces the reduction of the
number of iterations and, even when this is not true, there is in
any case a reduction in the calculation times. The Authors
have tested the proposed methodology on networks with 69
and 85 nodes, respectively considered in Ref. [1] and [4]. The
solution has been reached, in the first case, in 5 iterations and,
in the second case, in 6 iterations; the CPU times respectively
are 0.0014 s and 0.0021 s.
IV. CONCLUSIONS
A solution method for radial distribution systems has been
here set up. The proposed method is based on the classical
backward/forward method with some important modifications
for the calculation time reduction. The original system is
divided into a main feeder and some laterals all having the
same structure of feeders supplied at one end and without
laterals. The electrical features linking the main feeder and the
laterals are the voltages at the branching nodes (these buses
are both internal nodes, for the main feeder or the laterals, or
supply nodes for the laterals) and the currents derived from
the first branches of the laterals. Each feeder is solved starting
from the terminal branch for which it is known the voltage at
the EB (either because imposed at the first iteration or as a
result of the evaluations starting from the second iteration).
For the laterals, in the backward phase, the interesting
quantities are the node voltages and the current in the starting
branch; for the main feeder, these quantities are only the bus
voltages. In the forward phase the node voltages are modified
on the basis of the difference between the calculated values
and those imposed in the supply node of the system and in the
branching nodes. Since the calculation of the voltages is done
on the backward phase, going from the terminal nodes up to
the root node, current values that are closer to the final values
can be used, producing thus an acceleration in the whole
method allowing the reduction of the total number of
iterations required to reach the final solution with a given
error margin.
TAB. 2 AVERAGE CPU TIMES, IN SECONDS, RELATED TO THE NUMBER OF
ITERATIONS AND TO THE SOLUTION METHODOLOGY.
Iterations NR B/F M-B/F
4 0.053 0.00068 0.00067
5 0.068 0.00086 0.00082
6 0.084 0.00102 0.00098
7 0.00121 0.00114
8 0.00144 0.00136
9 0.00158 0.00144
10 0.00173 0.00159
11 0.00192 0.00174
12 0.00210 0.00191
13 0.00205
14 0.00243
18 0.00286
19 0.00331
The reduction of the number of iterations compensates the
increase in the number of required operations for the
calculation of the improved values of the currents. The results
have shown that the final solution is reached in a few
iterations, with quite low values of the error margin and with
limited calculation times.
As compared to the application of the classical
backward/forward method it can be noted that in all the
considered cases, a reduction of the total number of iterations
or, with the same number of iterations, the reduction of the
calculation times caused by a lesser number of operations due
to the particular way to implement the proposed methodology
has been obtained. Due to the favorable features of the
proposed procedure, the future work will be addressed
towards the introduction of PV buses simulating the presence
of dispersed generation units.
V. REFERENCES
[1] M. E. BARAN, F. F. WU - "Optimal sizing of capacitors placed on a
radial distribution system", IEEE Trans. PWRD, vol. 4, no. 1, pp. 735-
743, January 1989.
[2] G. B. JASMON, L. H. C. C. LEE - "Distribution network reduction for
voltage stability analysis and load flow calculations", Electric Power &
Energy Systems, vol. 13, no. 1, pp. 9-13, February 1991.
[3] H. D. CHIANG - "A decoupled load flow method for distribution power
networks: algorithms, analysis and convergence study", Electric Power
& Energy Systems, vol. 13, no. 3, pp. 130-138, June 1991.
[4] D. DAS, D. P. KOTHARI, A. KALAM - "Simple and efficient method
for load flow solution of radial distribution networks", Electric Power &
Energy Systems, vol. 17, no. 5, pp. 335-346, October 1995.
[5] D. DAS, H. S. NAGI, A. KALAM - "Novel method for solving radial
distribution networks", IEE Proc. - Gener. Transm. Distrib., vol. 141, no.
4, pp. 291-298 July 1994.
[6] M. H. HAQUE - "Efficient load flow method for distribution systems
with radial or mesh configuration", IEE Proc. - Gener. Transm. Distrib.,
vol. 143, no. 1, pp. 33-38, January 1996.
[7] W. M. LIN, J. H. TENG - "Phase-decoupled load flow method for radial
and weakly-meshed distribution networks", IEE Proc. - Gener. Transm.
Distrib., vol. 143, no. 1, pp. 39-42, January 1996.
[8] F. ZHANG, C. S. CHENG - "A modified Newton method for radial
distribution system power flow analysis", IEEE Trans. PWRS, vol. 12,
no. 1, pp. 389-397, February 1997.
[9] S. GHOS, D. DAS - "Method for load-flow solution of radial
distribution networks", IEE Proc. - Gener. Transm. Distrib., vol. 146, no.
6, pp. 641-648, November 1999.
[10] M. H. HAQUE - "A general load flow method for distribution systems",
Electric Power Systems Research, vol. 54, no. 1, pp. 47-54, April 2000.
[11] J. NANDA, M.S. SRINIVAS, M. SHARMA, S.S. DEY, L.L. LAI -
"New findings on radial distribution system load flow algorithms",
Power Engineering Society Winter Meeting, 2000, IEEE, vol. 2, pp.
1157-1161, 2000.
[12] A. AUGUGLIARO, L. DUSONCHET, M. IPPOLITO, E. RIVA
SANSEVERINO - "An efficient iterative method for load-flow solution
in radial distribution networks", IEEE Porto Power Tech 2001, 10-13
September 2001.
[13] M. E. BARAN, F. F. WU - "Network reconfiguration in distribution
systems for loss reduction and load balancing", IEEE Trans. PWRD,
vol. 4, no. 2, pp. 1401-1407, April 1989.
VI. BIOGRAPHIES
Antonino Augugliaro (1949) received the Doctors degree in Electrical
Engineering from the University of Palermo, Italy, in 1975. Since 1978 to
1994 he has been Associate Professor and now he is Full Professor of
Electrical Power Generation Plants at the Faculty of Engineering of the
University of Palermo. His main research interests are in the following fields:
simulation of electrical power system; transmission over long distances; mixed
three-phase/six-phase power system analysis; optimisation methods in
electrical distribution systems design and operation; distribution automation.
(Dipartimento di Ingegneria Elettrica, Universit di Palermo, Viale delle
Scienze, I-90128 Palermo, Italy, T +39 91/6566111, Fax +39 91/488452)
Luigi Dusonchet (1948) received the Doctors degree in Electrical
Engineering from the University of Palermo, Italy, in 1975. Since 1978 to
1990 he has been Associate Professor and now he is Full Professor of
Industrial Electrical Systems at the Faculty of Engineering of the University of
Palermo. His main research interests are in the following fields: simulation of
electrical power system; transmission over long distances; mixed three-
phase/six-phase power system analysis; optimisation methods in electrical
distribution systems design and operation; distribution automation.
(Dipartimento di Ingegneria Elettrica, Universit di Palermo, Viale delle
Scienze, I-90128 Palermo, Italy, T +39 91/6566111, Fax +39 91/488452)
Mariano Giuseppe Ippolito (1965).received the Doctors degree in Electrical
Engineering in 1990 and the Ph.D. in 1994 from the University of Palermo
(Italy). From 1995 to 2001, he is been working as a researcher at the
Department of Electrical Engineering in Palermo. Now he is Associate
Professor at the same Department. His main research interests are in the areas
of power systems analysis , optimal planning, design and control of electrical
distribution system. He also made some contributions to the advancement of
power quality study. (Dipartimento di Ingegneria Elettrica, Universit di
Palermo, Viale delle Scienze, I-90128 Palermo, Italy, T +39 91/6566204, Fax
+39 91/488452, e-mail: [email protected])
Stefano Mangione (1955) received the doctors degree in electrical
engineering from the University of Palermo, Italy, in 1981. In 1982 he joined
the Dept. of Electrical Engineering of the same University. Since 1983, he has
been Associate Professor of Power Systems at the University of Palermo. His
research interests are in the field of optimization methods on electrical
distribution systems; grounding systems design and electromagnetic
compatibility. (Dipartimento di Ingegneria Elettrica, Universit di Palermo,
Viale delle Scienze, I-90128 Palermo, Italy, T +39 91/6566111, Fax +39
91/488452)
Eleonora Riva Sanseverino (1971) received the Doctors degree in Electrical
Engineering from the University of Palermo, Italy, in 1995. Since 1995 shes
been working in the Research Group of Electrical Power Systems. In February
2000 shes successfully finished her Phd in Electrical Engineering at the same
University. Afterwards shes had a researcher position at the Dept. of
Electrical Engineering in Palermo. From December 2001 till October 2002
shes been working as a researcher at the National Council of Research in the
field of Computer Systems and Computer Networks. Now she is Associate
Professor in Power Systems at the University of Palermo. Her main research
interest is in the field of optimisation methods for electrical distribution
systems design, operation and planning. (Dipartimento di Ingegneria Elettrica,
Universit di Palermo, Viale delle Scienze, I-90128 Palermo, Italy, T +39
91/6566111, Fax +39 91/488452)