Three-Phase Backward/Forward Power Flow

Solution Considering Three-Phase Distribution

Abstract- This paper proposes a three-phase backward/forward mathematical formulas to model the distribution transformers
power flow solution which can handle distribution transformers in a branch current-based analysis, but their method needs to
of various common winding connections in unbalanced system of derive the individual formulae for different winding
multi voltage levels. The proposed algorithm is based on the connections from scratch. The authors of [13] propose a
power summation method and takes the power losses into
account considering the mutual coupling between three phases.
method to model Y-D transformer connection in three-phase
The transformers are modeled by transformer nodal admittance backward/forward power flow and in short circuit programs
matrix and mathematical formulas concerning to and also generalized it for different transformer connections.
backward/forward sweep calculations of transformers are They present the delta-side currents of the transformer by
presented in detail for power summation-based process. The product of a singular matrix and the transformer wye-side
validity and effectiveness of the proposed method is currents. Then they propose a technique to overcome the
demonstrated by the test results. singularity problem in order to calculate the primary currents
of the transformer in backward sweep. The authors of [14]
I. INTRODUCTION use the nodal admittance matrix of transformers in a three-
phase backward/forward power flow based on the current
One of the key calculations for any system is the summation method. They develop an approach to solve the
determination of the steady-state behavior, which is termed as singularity problem of some of mutual admittance
distribution power flow for a distribution system. The single- submatrices of nodal admittance matrix which is occurred in
phase power flow methods are normally used in the systems most of transformer connection types. In [15], the distribution
which their unbalances can be neglected. In distribution transformers and their loads are integrated and modeled by
systems, the three-phase balanced hypothesis can not their equivalent loads in power flow algorithm and also the
generally be applied and a three-phase power flow algorithm transformers are assumed to be ideal for simplification. In this
with complete three-phase models is required. The power case, the distribution network can only analyze in one voltage
flow solution techniques for distribution networks can be level, i.e. medium voltage level.
basically divided into three classes: the Newton-Raphson- Although, the current summation method is used
based method (including fast-decoupled versions) [1,2], Zbus more than the other backward/forward sweep methods in
Gauss-based method [3,4,5] as well as backward/forward three-phase power flow solution, the power summation
sweep methods [6-14]. Due to the radial structure of method has an advantage, it uses active and reactive powers
distribution network along with the high resistance/reactance as flow variables which makes it more convenient and
(R/X) ratio of its lines, it is an ill conditioned system and applicable than current summation method especially when
hence the Fast-Decoupled and Newton methods are distribution system contains some voltage-specified buses.
unsuitable for most of the distribution power flow programs. This paper proposes a three-phase backward/forward
The convergence producer of Zbus Gauss method is severely power flow solution which can handle distribution
dependant on the number of voltage-specified buses and it is transformers of various common winding connections in
suitable to solve power flow program in distribution system unbalanced system of multi voltage levels. The proposed
while the number of voltage-specified buses are small. algorithm is based on the power summation method and takes
Backward/forward sweep based methods are widely used to the power losses into account considering the mutual
solve the distribution power flow programs due to their low coupling between three phases. Each transformer is modeled
memory and computation requirements and their robust by its nodal admittance matrix as it is introduced in [5] and
convergence characteristics. They include three different mathematical formulas concerning to backward/forward
methods, the impedance summation method, the power sweep calculations of transformers are presented in detail for
summation method and the current summation method. power summation-based process. In order to solve the
According to the literature, the last method is more common singularity problem of some of mutual admittance
than the two first methods especially in three-phase submatrices of nodal admittance matrices, the technique of
applications [8-14]. One of the important issues in three- [14] is used and modified in accordance with the
phase backward/forward power flow solution is the modeling requirements of the power summation analysis.
of distribution transformers. The authors of [12] derive the
 II. BACKWARD/FORWARD POWER FLOW ALGORITHM BASED (k ) (k ) ( k −1)
ON THE POWER SUMMATION METHOD ΔVφi = Vφi − Vφi φ = [a, b, c] (5)

Let the root node be the slack node with known where, k denotes the iteration number. If any of these voltage
voltage magnitude and angle, and let initial voltage for all mismatches is greater than a convergence criterion, the three
other nodes be equal to the root node voltage. Also let initial first steps are repeated until convergence is achieved.
power loss for all branches is equal to zero. The iterative
algorithm for solving the radial system consists of four steps. III. INCORPORATING DISTRIBUTION TRANSFORMERS IN POWER
These steps at iteration k are: SUMMATION-BASED ALGORITHM
1. Nodal power calculation: Since the absorbed power by
some loads (i.e. constant current and constant impedance The three-phase transformers are modeled by their
loads) and the injected reactive power by the capacitors vary nodal admittance matrix. This model can be represented as
due to the variations of nodal voltages in iterative process, follows:
they should be recalculated at all iterations. For a constant ⎡ I p ⎤ ⎡Y pp Y ps ⎤ ⎡V p ⎤
current load, substitution of its absorbed current in well ⎢ I ⎥ = ⎢Y ⎥.⎢ ⎥ (6)
⎣ s ⎦ ⎣ sp Yss ⎦ ⎣Vs ⎦
known equation S ( k ) = V ( k ) .I * leads to calculation of its where vectors V p , I p and Vs , I s are three-phase line-to-
absorbed power. For a constant impedance load / a capacitor,
firstly its absorbed / injected current and thereby its absorbed neutral bus voltages and injection currents at the primary and
/ injected power should be recalculated at all iterations. The secondary sides of the transformer respectively. Y pp , Y ps , Ysp
absorbed powers of constant power loads are unchanged at all and Yss are sub matrices of transformer nodal admittance
iterations. matrix ( Yt ). The authors of [5] present the nodal admittance
2. Backward sweep to sum up branch power: Starting from
the branches connected to the end busses and moving toward matrices for nine common connections of three-phase
the root node, the power in branch n is: transformers considering the copper and core losses, the
winding connection, the phase-shifting between primary and
S nφ = S iφ + Lossnφ + ∑ S mφ φ = [a, b, c] (1) secondary windings, and the off-nominal tapping, see table I
m∈M
and equation (7).
where:
S nϕ is the power of phase ϕ of branch n , i denotes the ⎡1 0 0⎤ ⎡ 2 − 1 − 1⎤
⎢ ⎥ yt ⎢
YI = yt .⎢0 1 0⎥ YII = .⎢− 1 2 − 1⎥⎥
ending node of branch n, S iϕ is the absorbed power by the 3
⎢⎣0 0 1⎥⎦ ⎢⎣− 1 − 1 2 ⎥⎦
connected load to phase ϕ at the ending node of branch n,
Loss nφ is the power loss of phase φ of branch n , M is the ⎡− 1 1 0⎤
yt ⎢
YIII = . 0 − 1 1 ⎥⎥ (7)
set of branches which are connected to the ending node of 3 ⎢
branch n , S mϕ is the power of phase ϕ of branch m. ⎢⎣ 1 0 − 1⎥⎦
3. Forward sweep to update nodal voltages: Starting from the where yt is the per-unit transformer leakage admittance. It is
first branch, which is connected to the slack bus, and moving mentionable to say that YII and YIII are singular.
toward the ending branches:
* TABLE I
⎛ S nφ ⎞ Submatrices of nodal admittance matrix for common transformer connections
J nφ = ⎜ j ⎟ φ = [ a , b, c ] (2)
⎜ Vφ ⎟
⎝ ⎠ Transformer connection Self admittance Mutual admittance
where:
J nϕ is the current of phase ϕ of branch n and j denotes the Primary Secondary Y pp Yss Y ps Ysp
sending node of branch n .
GrY GrY YI YI −YI −YI
Vai,b,c = Vaj,b ,c − Z na ,b,c × J an,b,c (3)
GrY Y YII YII −YII −YII
Where Z na ,b ,c is the 3 × 3 series impedance matrix of branch
GrY D YI YII YIII YIIIt
n as it is discussed in [10].
Loss nφ = (Vφj − Vφi ).J nφ
*
φ = [ a , b, c ] (4) Y GrY YII YII −YII −YII
Y Y YII YII −YII −YII
According to equation (3), Vφj − Vφi for each phase is a
function of the currents vector of three phases, therefore the Y D YII YII YIII YIIIt
power losses for each phase, equation (4), is calculated D GrY YII YI YIIIt YIII
considering the mutual coupling with the other phases.
4. The voltage mismatches calculations: After the three steps D Y YII YII Y t
III
YIII
above are executed during each iteration, the voltage D D YII YII −YII −YII
mismatch for all phases at each bus (say bus i) are calculated
as below:
 In order to handle distribution transformers in ⎡Va(1+ 2) ⎤
[1 1 1].⎢⎢Vb(1+2) ⎥⎥ = 0
backward/forward algorithm based on the power summation
method, the following steps should be done. (15)
A. Backward sweep ⎢Vc(1+ 2) ⎥
⎣ ⎦
In backward sweep, the secondary powers and Therefore, if two of three laniary dependant equations in (14)
secondary voltages of the transformer are known and its are solved simultaneously with equation (15), the singularity
secondary currents can be computed using the well known problem will be solved and nonzero-sequence component of
equation I s = ( S s Vs ) * . Afterwards, using equation (6), the V can be calculated by using A' .V (1+ 2) = B ' , in which A' is
primary voltages and currents as well as primary powers can obtained by replacing one of the rows of A , e.g. its last row,
be computed as follows:
with [1 1 1] and B ' is obtained by replacing the same row
Ysp .V p = ( I s − Yss .Vs ) (8)
of B with 0.
I p = Y pp .V p + Y ps .Vs (9) As a result, in backward sweep, equation (8) for all
S p = V p .I *p (10) transformer connection in table 1 except GrY-GrY can be
replaced with equations (16), (17), (18) as below:
Note 1: Equation (4) does not use to calculate the power loss
V p = V p(1+2) + V p( 0) (16)
of transformers branches, since they are incorporated in the
basic model of the transformers as is discussed in [5]. (V paf + V pbf + V pcf )
Note 2: According to table 1, Ysp is singular only for GrY- V p( 0 ) = (17)
3
GrY connection, therefore calculating V p needs some where V paf , V pbf and V pcf are the primary voltages which are
additional process when singularity is occurred. calculated in forward sweep.
B. Forward sweep V p(1+ 2 ) = Ysp' ( I s' − Yss' .Vs ) (18)
In forward sweep, the primary currents and voltages
of transformer are known and its secondary voltages can be where Ysp' is obtained by replacing the last row of Ysp with
computed as follows:
[1 1 1] , I s' and Yss' are obtained by replacing their last row
Y ps .Vs = ( I p − Y pp .V p ) (11)
with 0.
Note: According to table 1, Y ps is singular only for GrY-GrY Similar way can be used to solve the equation (11) in forward
connection, therefore calculating Vs needs some additional sweep.
process when singularity is occurred.
C. Solving the singularity problem IV. NUMERICAL RESULTS
Let A.V = B , where V = [Va Vb Vc ]t , A is a Some test systems were used to validate the
nonsingular matrix ( A = YII or YIII ) and B is right hand side proposed method and evaluate its convergence characteristic.
matrix. V can be represented by summation of nonzero- A. Model validation
sequence and zero-sequence components as below: IEEE four-node test feeder which its data and
⎡Va ⎤ ⎡Va ⎤ ⎡Va ⎤
(1+ 2 ) ( 0) solutions are available in [16] were used to validate the
⎢V ⎥ = ⎢V (1+ 2) ⎥ + ⎢V ( 0) ⎥ (12)
proposed algorithm. The single line diagram of the test feeder
⎢ b⎥ ⎢ b ⎥ ⎢ b ⎥ is shown in Fig. 1. The primary purpose of this test feeder is
⎢⎣Vc ⎥⎦ ⎢Vc ⎥ ⎢Vc ⎥
(1+ 2 ) ( 0 )
to provide a simple system for the testing of all possible
⎣ ⎦ ⎣ ⎦
then equation A.V = B can be rewritten as below: three-phase transformer connections. The proposed algorithm
was applied on the test feeder for balanced and unbalanced
⎡Va ⎤ ⎛ ⎡Va(1+ 2 ) ⎤ ⎡Va( 0 ) ⎤ ⎞ ⎡Va(1+ 2 ) ⎤ loading and different transformer connections. In all cases,
⎜⎢ ⎢ ( 0) ⎥ ⎟ ⎢ ⎥
⎢ ⎥ (1+ 2 ) ⎥ the results match very well with those listed in [16]. The
A.⎢Vb ⎥ = A.⎜ ⎢Vb ⎥ + ⎢Vb ⎥ ⎟ = A.⎢Vb(1+ 2 ) ⎥ = B ,
⎜ ⎢ (1+ 2 ) ⎥ ⎢ ( 0 ) ⎥ ⎟ ⎢Vc(1+ 2 ) ⎥ results correspond with unbalanced loading and different
⎢⎣Vc ⎥⎦ ⎜ Vc ⎟
⎝⎣ ⎦ ⎣Vc ⎦ ⎠ ⎣ ⎦ step-down transformer connections are presented in tables II
A = YII or YIII (13) and III.
Phase a:
⎡V a
⎤
(1+ 2 )
1 2 3 4 1500 KVA, 0.85 lag
⎢ ⎥
(1+ 2 ) 2000ft 2500ft Phase b:
⎥=B
Load
A.⎢V b (14) 2000 KVA, 0.9 lag
Infinite Phase c:
⎢V ⎥
(1+ 2 ) bus 2500 KVA, 0.95 lag
⎣ c ⎦ Three-phase transformer
In other words, both V and its nonzero-sequence component 6000 KVA, 12.47KV – 4.16 KV, Z =1.0+j6.0%

( V (1+ 2) ) are applicable in equation A.V = B . Moreover, since Fig. 1. IEEE four-node test feeder
V (1+ 2) dose not contain zero-sequence component, the
following equation is established:
 Table II
IEEE four-node test feeder with unbalanced load and different step-down transformer connections - voltages (phase to neutral for wye connections and
line to line for delta connections)
Transformer Node V1 (volt) V2 (volt) V3 (volt)
connections ID Calculated IEEE results calculated IEEE results Calculated IEEE results
2 7164<-0.1 7164<-0.1 7112<-120.2 7110<-120.2 7086<119.3 7082<119.3
Gr Y – GrY 3 2305<-2.3 2305<-2.3 2254<-123.6 2255<-123.6 2201<114.8 2203<114.8
4 2177<-4.1 2175<-4.1 1935<-126.8 1930<-126.8 1841<103.1 1833<102.8
2 7115<-0.2 7113<-0.2 7145<-120.4 7144<-120.4 7112<119.5 7111<119.5
Gr Y – D 3 3895<-2.8 3896<-2.8 3971<-123.8 3972<-123.8 3875<115.7 3875<115.7
4 3435<-5.7 3425<-5.8 3652<-130.1 3646<-130.3 3311<108.7 3298<108.6
2 7113<-0.2 7112<-0.2 7145<-120.4 7144<-120.4 7113<119.5 7112<119.5
Y–D 3 3895<-2.8 3896<-2.8 3971<-123.8 3972<-123.8 3875<115.7 3874<115.7
4 3435<-5.7 3425<-5.8 3652<-130.1 3646<-130.3 3311<108.7 3298<108.6
2 12344<29.8 12341<29.8 12373<-90.5 12370<-90.5 12305<149.6 12302<149.5
D–D 3 3900<27.2 3902<27.2 3973<-93.9 3972<-93.9 3871<145.7 3871<145.7
4 3439<24.3 3431<24.3 3655<-100.2 3647<-100.4 3306<138.7 3294<138.6

Table III
IEEE four-node test feeder with unbalanced load and different step-down transformer connections – currents
From Ia (amper) Ib (amper) Ic (amper)
Transformer node
connections to Calculated IEEE results Calculated IEEE results Calculated IEEE results
node
1-2 229.8<-35.9 230.1<-35.9 344.5<-152.6 345.7<-152.6 452.1<84.9 455.1<84.7
Gr Y – Gr Y
3-4 689<-35.9 689.7<-35.9 1033.3<-152.6 1036<-152.6 1358<84.9 1364<84.7
1-2 307.4<-41.3 308.5<-41.5 313.5<-145.4 314.6<-145.5 387.5<86 389<85.9
Gr Y – D
3-4 1080<-71 1083.8<-71 847.6<177 849.9<177 1095.2<63.2 1098.7<63.1
1-2 309<-41.5 309.8<-41.7 314.4<-145.1 315.5<-145.2 385.7<86 387.2<85.9
Y–D
3-4 1080<-71 1083.8<-71 848<177.1 849.9<177 1095<63.2 1098.7<63.1
1-2 360.2<-41 361.7<-41 282.5<-153 283.5<-153 365.2<93.3 366.5<93.2
D–D
3-4 1081<-41 1084<-41 847.4<-153 849.7<-153 1095<93.3 1099<93.2

Table IV Table V
Iteration number and execution time for different test systems Iteration number for different loading conditions of IEEE 13-node test feeder
Current summation Number of iteration
Proposed method
method Case Current
Test system Loading Proposed
No. summation
Iteration Time (s) Iteration Time (s) method
method
2 bus test 1 0.25(P+jQ) 4 4
4 0.0005 6 0.00006
system 2 0.5(P+jQ) 5 5
IEEE 13-node 3 0.75(P+jQ) 6 6
6 0.015 7 0.016
test feeder 4 1.0(P+jQ) 6 7
IEEE 37-node 5 1.25(P+jQ) 7 8
4 0.046 4 0.047
test feeder 6 1.5(P+jQ) 8 9
7 1.75(P+jQ) 9 11
8 2(P+jQ) 13 15
B. Convergence characteristic
Three test systems with different sizes, i.e. a two From the results, it is observed that as the load increases, the
buses test system, IEEE 13-node test feeder and IEEE 37- number of iterations for convergence increases in both
node test feeder, were solved using proposed algorithm. The method. However, the proposed method is converged by less
data of the first system are available in [5] and the data of the iteration.
second and third systems are available in [16]. Each test
system contains a three-phase transformer, i.e. D-GrY V. ACKNOWLEDGEMENT
connection for first system, GrY-GrY connection for second
system and D-D connection for third system. In all cases the The authors gratefully acknowledge the supports of
results match very well with those listed in corresponding Khouzestan Electric Power Distribution Company in this
references. The test systems were also solved using the research work.
current summation method similar to that in [14].
The number of iterations and the execution time for REFERENCES
converging the proposed method are often less than the
[1] W. M. Lin and J. H. Teng, “Three-Phase Distribution Network Fast-
current summation method, see table IV. Decoupled Power Flow Solution”, International Journal of Electric Power
Table V compares the iteration number of the Energy System, Vol. 22, No. 5, pp. 375-380, June 2000.
proposed method and current summation method under [2] R. D. Zimmerman and H. D. Chiang, “Fast Decoupled Power Flow for
different loading condition of IEEE 13-node test feeder. Unbalanced Radial Distribution Systems”, IEEE Transactions on Power
System, Vol. 10, No. 4, pp. 2045-2052, Nov.1995.
[3] T.-H. Chen, M.-Sh. Chen, K.-J. Hwang, P. Kotas and E. A,Chebli, [10] C. S. Cheng and D. Shirmohammadi, “A Three-Phase Power Flow
“Distribution System Power Flow Analysis – a Rigid Approach”, IEEE Method for Real-Time Distribution System Analysis”, IEEE Transactions on
Transactions on Power Delivery, Vol. 6, No. 3, pp. 1146-1152, July 1991. Power Systems, Vol. 10, No. 2, pp. 671-679, May 1995.
[4] J.C.M. Vieira Jr, W. Freitas and A. Morelato, “Phase – Decoupled [11] S. Khushalani, J. M. Solanki and N. N. Schulz, “Development of Three-
Method for Three-Phase Power Flow Analysis of Unbalanced Distribution Phase Unbalanced Power Flow Using PV and PQ Models for Distributed
Systems”, IEE Proceeding on Generation, Transmission and Distribution, Generation and Study of the Impact of DG Models”, IEEE Transactions on
Vol. 151, No. 5, pp. 568 – 574, Sept. 2004. Power Systems, Vol. 22, No. 3, pp. 1019-1025, August 2007.
[5] T.-H. Chen, M.-Sh. Chen, T. Inoue, P. Kotas and E. A. Chebli, “Three – [12]. M. E. Baran and E. A. Staton, “Distribution Transformer Models for
Phase Cogenerator and Transformer Models for Distribution System Branch Current Based Feeder Analysis”, IEEE Transactions on Power
Analysis”, IEEE Transactions on Power Delivery, Vol. 6, No. 4, pp. 1671- Systems, Vol. 12, No. 2, pp. 698-703, May 1997.
1681, October 1991. [13] W. H. Kersting, W. H. Phillips and W. Carr, “A New Approach to
[6] A. Augugliaro, L. Dusonchet, S. Favuzza, M. G. Ippolito and E. Riva Modeling Three-Phase Transformer Connections”, IEEE Transactions on
Sanseverino, “A New Backward/Forward Method for Solving Radial Industry Applications, Vol. 35, No. 1, pp. 169-175, Jan./Feb.1999.
Distribution Networks with PV Nodes”, Electric Power Systems Research, [14] P. Xiao, D. C. Yu and W. Yan, “A Unified Three-Phase Transformer
Vol. 78, pp. 330-336, 2008. Model for Distribution Load Flow Calculations”, IEEE Transactions on
[7] D. Shirmohammadi, H. W. Hong, A. Semlyen and G. X. Luo, “A Power Systems, Vol. 21, No. 1, pp. 153-159, Feb. 2006.
Compensation-Based Power Flow Method for Weakly Meshed Distribution [15] T-H Chen and Y-L Chang, “Integrated Model of Distribution
and Transmission Network”, IEEE Transactions on Power Systems, Vol. 3, Transformer and Their Loads for Three-Thase Power Flow Analyses”, IEEE
No. 2, pp. 753-762, May 1988. Transactions on Power Delivery, Vol. 11, No. 1, pp. 507-513, Jan. 1996.
[8] D. Thukaram, H. M. Wijekoon Banda and J. Jerome, “A Robust Three [16] Radial Distribution Test Feeders. [Online]. Available:
Phase Power Flow Algorithm for Radial Distribution Systrms”, Electric http://ewh.ieee.org/soc/pes/dsacom/testfeeders.html
Power Systems Research, Vol. 50, pp. 227-236, 1999.
[9] G. X. Luo and A. Semlyen, “Efficient Load Flow for Large Weakly
Meshed Networks”, IEEE Transactions on Power Systems, Vol. 5, No. 4,
pp.1309-1316, Nov. 1990.