Electric Power Systems Research 81 (2011) 1499–1506

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

A generalized power flow analysis for distribution systems with high penetration
of distributed generation
Hany E. Farag ∗ , E.F. El-Saadany, Ramadan El Shatshat, Aboelsood Zidan
Department of Electrical and Computer Engineering, University of Waterloo, 200 University Ave. W, Waterloo, ON, Canada N2L3G1

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, the element incidence matrix has been extended to develop a comprehensive three-phase
Received 30 November 2010 distribution system power flow program for radial topology. Three-phase overhead or underground pri-
Received in revised form 1 March 2011 mary feeders and double-phase or single-phase line sections near the end of the feeder laterals have been
Accepted 2 March 2011
considered. Unbalanced loads with different types including constant power, constant current and con-
Available online 1 April 2011
stant impedance are modeled at the system buses. Substation voltage regulator (SVR) consisting of three
single phase units connected in wye or two single-phase units connected in open delta are modeled to
Keywords:
satisfy the desired voltage level along the feeder. The mathematical model of distributed generation (DG)
Distribution systems
Unbalanced power flow
connected as PQ and PV buses are integrated into the power flow program to simulate the penetration
Distributed generation (DG) of DGs in the distribution systems. The proposed method has been tested and compared with different
Step voltage regulator IEEE test feeders result. The developed algorithm has been used to study the impact of both SVR and high
Coordination penetration of DG on voltage profile and system power losses.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction as state variables to solve the power flow problem [7–10]. Teng
[11] proposed a network topology based three-phase distribu-
Distribution system represents the final link between the bulk tion power flow algorithm. The algorithm developed two matrices,
power system and the consumers, therefore it is crucial to have namely bus-injection to branch-current matrix (BIBC) and branch-
an accurate analysis for such systems. Power flow programs are current to bus-voltage matrix (BCBV); in order to obtain the power
typically used in both operational and planning stages. There flow solution. The advantage of this method is that it does not need
are different applications (i.e. distribution automation, network matrix decomposition, forward/backward substitution of the Jaco-
optimization, Var. planning, and switching state estimation) that bian matrix or the Y admittance matrix as in other methods. In [12]
require power flow analysis [1]. The distribution system has some an approach for three-phase power flow for large scale distribution
distinct characteristics that are different from the transmission sys- systems is presented. The method is based on the implicit Z-bus and
tem due to the following [2]: the modified Gauss–Seidel method. The factorization only needs to
be done once for the sub-admittance matrix (YAA , YBB and YCC ) in
• It works in radial or weakly meshed topology. the solution procedure which reduces the CPU execution time.
• Distribution lines usually have high R/X ratio. Recently, due to the continuous growth of distributed gener-
• Significant unbalance may be found. ation DG penetration in the distribution networks, many studies
were performed on distribution systems with the presence of DGs.
The goal of these studies is to show the impact of DG on power qual-
Many power flow algorithms have been proposed for distribu-
ity and its control capabilities to provide fast voltage regulation.
tion systems. In general, these methods are categorized as node
The authors in [13–15] showed that with a proper control scheme,
based and branch based methods [3]. In node based methods (e.g.
inverter-based DG has superior capability for fast load voltage reg-
network equivalence method, Z-bus method, Newton–Raphson
ulation and mitigation of unbalanced voltage disturbances.
algorithm and Fast decouple algorithm), node voltage or current
In [16] a helpful list of the various types of DGs and their con-
injections are used as state variables to solve the power flow prob-
nection to the grid as well as their suitable models for power flow
lem [4–6]. However, in branch based methods (sweep based and
studies (PV and PQ or negative load) were offered. The model of
loop impedance methods) branch currents or powers are used
DGs as PV or PQ depends on its operational mode and control char-
acteristics [17]. In [18,19] various distribution system equipment
models for distribution power flow (i.e. DGs and voltage regulators)
∗ Corresponding author. Tel.: +1 519 888 4567; fax: +1 519 746 3077.
were presented. DGs were modeled as PQ (i.e. negative load) or as
E-mail address: [email protected] (H.E. Farag).

0378-7796/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2011.03.001
1500 H.E. Farag et al. / Electric Power Systems Research 81 (2011) 1499–1506

PV bus using fictitious node with impedance. Teng [20] proposed

three mathematical models with equations of DGs for power flow
analysis. The models are: (1) constant power factor model, (2) vari-
able reactive power model and (3) constant voltage model. Without
SVR modeling, the authors in [21] studied the impact of DG mod-
els on the voltage profile and power losses for the IEEE 37 bus test
feeder.
Fig. 2. Exact lumped load model.
However, none of these references tackle the model of all dis-
tribution system components or compare their results with power
flow results of a generalized test feeder like the IEEE radial distri- in Eq. (1).
bution test feeders [22]. In this paper, a simplified, efficient and ⎡ ⎤
generalized algorithm for solving unbalanced three-phase radial Zaa Zab Zac Zan
distribution system power flow problem is presented. The algo- ⎢ Zba Zbb Zbc Zbn ⎥
[Zabcn ] = ⎣
Zcn ⎦
(1)
rithm is simple because it depends mainly on the bus incidence Zca Zcb Zcc
matrix which represents the relation between bus-injection cur- Zna Znb Znc Znn
rents to branch currents. It is also efficient as it shows good
Eq. (2) is designed to include the effects of the neutral or ground
convergence characteristics in all operating conditions. Further-
wire and to be used in the unbalanced power flow calculation [1,23].
more, it is generalized because it incorporates the three phase
model of feeders, unbalance due to both loads and different type

Zaa−n Zab−n Zac−n
of phases, exact load modeling, SVR model, and distributed gen- [Zabc ] = Zba−n Zbb−n Zbc−n (2)
erator model in its different operation modes. The algorithm has Zca−n Zcb−n Zcc−n
been tested and compared with the results of the IEEE 13 and 37
bus unbalanced radial distribution test feeders, which are available The relation between the bus voltages and branch currents in
in [22]. Different case studies with and without both SVR and DGs Fig. 1 can be expressed as shown in Eq. (3). The feeder shunt capac-
are presented to show the impact of both of them on the voltage itive current in each bus can be calculated using Eq. (4).
profile and system losses and how they can be coordinated to avoid  a
   a
Vm Vna Zaa−n Zab−n Zac−n Iline
the interference between them. b
Vm = Vnb − Zba−n Zbb−n Zbc−n b
Iline (3)
c
Vm Vnc Zca−n Zcb−n Zcc−n c
Iline
2. Distribution system modeling
 a   a
Ish Yaa Yab Yac Vbus
2.1. Feeders modeling 1
b
Ish = Yba Ybb Ybc b
Vbus (4)
c 2 c
Ish Yca Ycb Ycc Vbus
For transmission system, the three phase currents are balanced
and transposition of the conductors pretty much eliminates the
variation in the line parameters. However, distribution systems 2.2. Uniformly distributed load lumped model
do not have these two characteristics. Balancing the loads on the
three phases of a distribution system is almost next to impossible Fig. 2 shows the general configuration of the exact model of uni-
leading to three phase currents that are typically away from the formly distributed loads [23]. In this work, the node at one-fourth
balanced condition. Distribution lines are seldom transposed nor of the way from the source end has been presented as a dummy
can it be assumed that the conductor configuration is an equilat- bus.
eral triangle. When these two characteristics are dominant, it is
necessary to introduce a more accurate method for calculating the 2.3. Load modeling
line impedance.
In this work, Carson’s equations of a three-phase grounded four- Behavior of loads can be modeled by studying the changes in
wire system are used. Carson’s equations allow the computation of their active and reactive power requirements due to changes in
conductor self impedance and the mutual impedance between any system voltages, either analytically or experimentally. There are
number of conductors above ground. Fig. 1 shows a three-phase three well established models for loads representation in power
line section between bus n and m [23]. A 4 × 4 matrix, which takes system studies [24]:
into account the self and mutual coupling terms, can be expressed
as shown in Eq. (1). For a well-grounded distribution system, VN 1. Constant power: for which load power and reactive power are
and Vn are assumed to be zero, and Kron’s reduction can be applied assumed to be constant irrespective of bus load voltage; which

Fig. 1. Three phase feeder model.

 H.E. Farag et al. / Electric Power Systems Research 81 (2011) 1499–1506 1501

  a 
Qa Vdum Va
Qspec = Qb , Vdum,i = b
Vdum , Vi = Vb (9b)
Qc c
Vdum Vc

where Qspec is the dummy reactive power specified value. Vit is the
t
calculated voltage of PV specified node ‘i’ at iteration t. Vdum,i is the
voltage of the dummy node of PV specified node i at iteration ‘t’. ˇi
is the sensitivity reactance; it depends on feeder parameters, DG
size and loads. The dummy node voltage is updated as:
t spec
Vdum,i = Vi − |Vit | (10a)

t+1 t t
Fig. 3. PV node model using a dummy branch. Vdum,i = Vdum,i + Vdum,i (10b)

As the system is unbalance, if the DG is a synchronous machine

means they are usually voltage independent. then the positive sequence voltage has been considered as a ref-
spec
erence (Vi ). If the DG is inverter-based, both positive sequence
P = P0 , Q = Q0 (5)
voltage and phase voltages can be taken as a reference.
2. Constant current: in which the active and reactive powers are To take the DG reactive power limits in consideration, the calcu-
assumed to be proportional to the load bus voltage. Usually it lated DG reactive power is compared with minimum and maximum
represents the combination of constant impedance and constant limits. If the calculated reactive power violates the upper or lower
power loads. limits, the DG switch from PV to PQ mode and the reactive power is
kept at its limits. When DG operates in PQ mode, it is represented
V V in power flow algorithm as a negative load.
P = P0 , Q = Q0 (6)
V0 V0

3. Constant impedance: this is the traditional representation, in 3. Problem formulation

which, at system nominal frequency, the active and reactive con-
sumed powers are assumed to be proportional to the square of The proposed method is based on the relation between bus-
supply voltage. The load is represented by a constant admittance injected currents and branch currents. For an N bus radial
added to the system admittance matrix at the concerned load distribution network; there are only N − 1 lines (elements) and
node. branch currents can be expressed in terms of bus currents [25]. For
2 2
an element ij connecting between nodes ‘i’ and ‘j’ the bus current
V V of node j can be expressed as a linear equation.
P = P0 , Q = Q0 (7)
V0 V0 
where V0 represents the initial operating voltage and P0 , Q0 are Ij = Iij − Ijk(j) (11)
active and reactive power corresponding to the initial operating
voltage. where k(j) is the set of nodes connected to node j. For the slack
bus the power is not specified so it is excluded and the relationship
between bus currents and branch currents are derived as a non-
2.4. Step voltage regulator modeling SVR
singular square matrix.

Step voltage regulator consists of an autotransformer and a load Ibus = KIbranch (12)
tap changing mechanism. The voltage change can be obtained by
changing the taps of the autotransformer series winding by line The matrix K is element incidence matrix. It is a non singular
drop compensator (LDC) control circuit. The effective regulator square matrix of order (N − 1). The elemental incidence matrix is
ratio is given by [23]: constructed in a simple way same like bus incidence matrix. In
   this matrix K each row is describing the element incidences. The
VAn aRa 0 0 Van elements are numbered in a conventional way, i.e. the number of
VBn = 0 aRb 0 Vbn (8) element ‘ij’ is j − 1. where:
VCn 0 0 aRc Vcn
• The diagonal elements of matrix K are ones. The variable j is
where aRa , aRb and aRc are the effective turns ratios for the three
single phase regulators. In this paper, the number of required taps denoting the element number K(j, j) = 1.
• For each jth element let m (j) is the set of element numbers con-
has been calculated using LDC control circuit after power flow con-
vergence. Then power flow has been repeated to consider the SVR nected at its receiving end K(j, m(j)) = −1.
• All the remaining elements are zeros. It can be observed that all
impact by applying Eq. (8) [19].
the elements of matrix K below the diagonal are zeros.

2.5. Distributed generator DG modeling

By using Eq. (12) and after the construction of the element inci-
dence matrix K, we can get the branch currents with respect to the
Distributed generators DG are represented by PV node using
injection bus currents as shown in Eq. (13).
a dummy node and dummy branch which inject reactive power
to the specified node to maintain the specified voltage value [19] Ibranch = k−1 Ibus (13)
as shown in Fig. 3. The dummy reactive power injection value is
calculated as follows: The idea of the bus incidence matrix has been extended in this
t paper to work in case of unbalanced radial distribution systems.
Vdum,i − |Vit |
Qspec = |Vit | (9a) Fig. 4 shows the flowchart for the proposed algorithm. Detailed
ˇi explanations of this algorithm are discussed in the following steps:
1502 H.E. Farag et al. / Electric Power Systems Research 81 (2011) 1499–1506

Step 1: take initial values for the magnitude of all bus voltages
equal to 1.0 p.u (flat start). For any phase which fails to present,
set the value of its voltage equal to zero.
Step 2: construct the element incidence matrices Ka , Kb and Kc for
phases a, b and c, respectively. Note that the size of matrix Ka is
(Na − 1)× (Na − 1), the size of matrix Kb is (Nb − 1) × (Nb − 1) and
the size of matrix Kc is (Nc − 1) × (Nc − 1), where Na , Nb and Nc
are the total number of buses at which phases a, b and c exist,
respectively.
a
Ibus = ILa + Ish
a − Ia
G
b
Ibus = ILb + Ish
b − Ib (14)
G
c
Ibus = ILc + Ish
c − Ic
G

where ILa,b,c is the load current in each phase and it can be calcu-
lated using Eq. (15). Note that n is the load exponent. It equals zero
for constant power, one for constant current and two for constant
impedance load.

a,b,c n a,b,c
abs(Vbus ) × conj(SL,spec )
ILa,b,c = (15)
a,b,c
conj(Vbus )

a,b,c
Ish is the feeder shunt capacitive current and it can be calculated
using Eq. (4). IGa,b,c is the distributed generator current and it can be
a,b,c
calculated using Eq. (16). Qg,spec is fixed when a DG is represented
as PQ and it is updated in each iteration when the DG operates in
PV mode.
Step 3: using the initial bus voltages, the specified load powers,
the feeder shunt admittance matrix, and the specified DG power,
calculate the bus currents at each phase.

a,b,c a,b,c
(Pg,spec − jQg,spec )
IGa,b,c = (16)
a,b,c
conj(Vbus )

Step 4: the branch currents in each phase can be expressed as:

a
Ibranch = Ka−1 Ibus
a
b −1 b
Ibranch = Kb Ibus (17)
c
Ibranch = Kc−1 Ibus
c

Step 5: the voltage drop at each branch (i) in each phase can be
expressed by:
 a   a
Vbranch i Zaa i Zab i Zac i Ibranch i
b
Vbranch i
= Zba i Zbb i Zbc i
b
Ibranch i
(18)
c
Vbranch Zca Zcb Zcc c
Ibranch
i i i i i

Step 6: calculate the voltages of the receiving ends with respect to

the voltages of the sending end using the backward sweep, subject
to the phase at bus j is found.
⎡ a
⎤  
Vbus j
a
Vbus a
Vbranch
i i
⎣ b
Vbus j
⎦= b
Vbus − b
Vbranch (19)
i i
c c
Vbus c
Vbranch
Vbus j i i

Step 7: the absolute difference between the new bus voltages

and the initial bus voltages gives the mismatch which represents
the convergence factor. In addition, if a DG operates in PV mode,
the absolute difference between the specified positive sequence
voltage and the calculated one must be taken in the mismatch.
Furthermore, it is possible to calculate the mismatch by compar-
ing the specified and the calculated injected active and reactive
power at each bus.
Step 8: repeat the process from step 3 to step 7 until convergence
Fig. 4. A flowchart for the proposed algorithm.
is achieved based on the required mismatch.
 H.E. Farag et al. / Electric Power Systems Research 81 (2011) 1499–1506 1503

4. Studied systems

The IEEE 13 and 37 bus unbalanced radial distribution test feed-

ers have been chosen to test the effectiveness of the proposed
algorithm. The two test feeders contain three phase feeder model,
different load modeling, different load types (1-phase, 2-phase and
3-phase), and SVR model with LDC control technique. Although the
IEEE 13 bus test feeder is very small, it is a comprehensive feeder.
It contains overhead and underground lines with variety of phas-
ing, shunt capacitor banks and unbalanced spots and distributed
loads. In another side, the IEEE 37 bus test feeder is a three-wire
delta operating with very unbalanced loading. In this feeder, all
loads are spot loads, all line segments are underground and the
SVR consisting of two single-phase units connected in open delta.
The test feeders data and power flow results are available in [22].
The algorithm has been coded in MATLAB. Several case studies have
been carried out to illustrate the robustness of the algorithm and
Fig. 5. Line voltages in per unit for the base case of the IEEE 37 bus test feeder.
the effect of both SVR and DGs on the operation of the distribution
systems. The algorithm shows good convergence in all operating
conditions. The number of required iterations for convergence was
around 5 iterations. bus-voltage matrix (BCBV). Fig. 5 shows the voltage profile of the
line voltages in the test feeder for the proposed method, IEEE results
and the developed algorithm in [11]. As shown in the figure, the
4.1. Case 1: base case for IEEE 13 bus test feeder results are almost the same. Table 2 shows the radial flow summary
of the test feeder. As shown in the table, the number of iterations
After convergence of the program, the numbers of taps that are in our proposed method is 6 and the number of iterations for the
required to maintain the voltages within the specified standard lim- developed algorithm in [11] is 9 iterations. These results show that
its have been calculated using LDC technique. It has been found that the proposed method has good convergence characteristics in large
those taps are 10, 8, and 11 for phases a, b, and c, respectively. The scale distribution feeders. In addition, the total reactive power loss
program was converged after 5 iterations. Table 1 shows the results based on the developed algorithm in [11] is not accurate due to
of the proposed algorithm compared with the results of IEEE [22]. neglecting the shunt capacitance of the distribution lines. These
The table shows that the proposed algorithm results are approxi- results indicate the importance of our comprehensive power flow
mately the same, which indicate the effectiveness of the proposed algorithm.
algorithm.

4.2. Case 2: base case for IEEE 37 bus test feeder 4.3. Case 3: IEEE 37 bus feeder with DGs

In this section, the proposed method has been compared with Table 3 shows the data for eight DGs connected to the IEEE 37
the IEEE results and the developed algorithm in reference [11]. The bus feeder. We assume that all DGs have the capability to operate
algorithm developed in [11] consists of two matrices; they are bus- in PQ and PV modes except DG # 2 and 8, which operate at 0.9 lag
injection to branch current matrix (BIBC) and branch-current to power factors (absorb reactive power) in all operating conditions.

Table 1
IEEE 13 bus load flow results of the proposed algorithm compared with IEEE results.

Node VA–N VB–N VC–N

Test results IEEE results Test results IEEE results Test results IEEE results

Mag. Ang. Mag. Ang. Mag. Ang. Mag. Ang. Mag. Ang. Mag. Ang.

650 1.000 0.0 1.000 0.0 1.000 −120 1.000 −120 1.00 120 1.00 120
RG60 1.0625 0.0 1.0625 0.0 1.050 −120 1.050 −120 1.0687 120 1.0687 120
632 1.0220 −2.47 1.0210 −2.49 1.0402 −121.70 1.042 −121.72 1.0182 117.78 1.0174 117.83
633 1.0190 −2.54 1.018 −2.56 1.0390 −121.74 1.0401 −121.77 1.0152 117.77 1.0148 117.82
634 0.9948 −3.20 0.9941 −3.23 1.021 −122.20 1.0218 −122.22 0.9963 117.30 0.9960 117.34
645 – – – – 1.0319 −121.87 1.0329 −121.9 1.0160 117.81 1.0155 117.86
646 – – – – 1.030 −121.95 1.0311 −121.98 1.0141 117.85 1.0134 117.9
671 0.9909 −5.28 0.9900 −5.30 1.0515 −122.31 1.0529 −122.34 0.9780 115.97 0.9778 116.02
680 0.9909 −5.28 0.9900 −5.30 1.0515 −122.31 1.0529 −122.34 0.9780 115.97 0.9778 116.02
684 0.9900 −5.30 0.9881 −5.32 – – – – 0.9763 115.85 0.9758 115.92
611 – – – – – – – – 0.9744 115.73 0.9738 115.78
652 0.9834 −5.22 0.9825 −5.25 – – – – – – – –
692 0.9909 −5.28 0.9900 −5.31 1.0515 −122.31 1.0529 −122.34 0.9780 115.97 0.9777 116.02
675 0.9846 −5.53 0.9835 −5.56 1.0545 −122.47 1.0553 −122.52 0.9763 115.95 0.9758 116.03

Total KW KVAR

Test results IEEE results Test results IEEE results

Load 3466.352 3466.128 1400.6 1400.119

Substation power 3577.2 3577.191 1725.10 1724.772
Losses 110.848 111.063 324.5 324.653
1504 H.E. Farag et al. / Electric Power Systems Research 81 (2011) 1499–1506

Table 2
Radial power flow summary for the IEEE 37 bus test feeder.

A–B B–C C–A

IEEE results Proposed Ref. [11] IEEE results Proposed Ref. [11] IEEE results Proposed Ref. [11]

Load KW 732.331 732.306 732.349 638.824 638.77 638.83 1090.1 1090.1 1090.099
KVAR 359.538 359.525 359.54 313.874 313.848 313.875 529.558 529.553 529.558
Losses KW 26.671 26.671 26.63 13.804 13.804 13.735 20.088 20.088 19.893
KVAR 18.769 18.995 21.283 9.953 9.952 12.239 17.733 17.961 19.854

Total KW KVAR

IEEE results Proposed Ref. [11] IEEE results Proposed Ref. [11]

System input 2521.827 2521.725 2521.64 1249.435 1250.073 1256.39

Load 2461.255 2461.63 2461.28 1202.970 1202.93 1202.973
Losses 60.563 60.563 60.2582 46.455 46.908 53.3765

Proposed Ref. [11]

Iterations 6 9
Time elapsed 0.0193s 0.02099s

Table 3
DGs locations, types and ratings in the IEEE 37 bus test feeder.

DG # Location Pgmax (KW) Type PF

1 701 600 3-phase Controllable

2 703 450 3-phase 0.9 lag
3 775 450 3-phase Controllable
4 742 200 3-phase Controllable
5 722 200 3-phase Controllable
6 725 200 3-phase Controllable
7 710 250 3-phase Controllable
8 741 150 1-phase 0.9 lag

4.3.1. Case 3-a: when DGs operate in PQ mode

First we considered that the controllable DGs operate at max-
imum generation power and unity power factor. In this case the
number of required taps for SVR based on LDC is 4 and 5 for phases
AB and CB, respectively. Fig. 6 shows the voltage profile when all
DGs operate in PQ mode and SVR operates at 4 and 5 taps compared
Fig. 7. Voltage profile in per unit of the feeder after SVR operation and DGs operate
with the voltage profile in the base case.
at different modes and set points.

4.3.2. Case 3-b: when DGs operate in PV mode

Total real power loss in KW
When a DG switches from power factor control (PFC) mode to
voltage control (PV) mode, its voltage set point is required to be 40
adjusted such that it does not conflict with other DGs and with 35
existing utility voltage control devices. To demonstrate this issue,
30
consider that the six controllable DGs shown in Table 3 are adjusted
to regulate their local positive sequence voltage. Figs. 7 and 8 show 25
20
15
10
5
0
DGs in PV Vset =1 DGs in PV Vset =1.025 DGs in PQ PF =1.0

Fig. 8. Total real power loss after SVR operation and DGs operate at different modes
and set points.

the voltage profile for phase CA and the total real power loss after
SVR actions when DGs operate in PV mode with set points equal 1.0
and 1.025 per unit and when DGs operate with unity power factor.
The figures show that the least improvement in voltage profile and
total system loss is achieved with improper setting for DGs in PV
mode. In contrast, the best improvement occurs when DGs operate
in PV mode with proper set points. This interaction beside other
Fig. 6. Line voltages in per unit with and without DGs when DGs operate in PQ
interactions between DGs and utility controllers make the trend of
mode. the research aims to tackle these challenges by converting the dis-
 H.E. Farag et al. / Electric Power Systems Research 81 (2011) 1499–1506 1505

tribution system from its passive appendage into active network. Total Power Loss in KW
An important aspect of active network is to find a suitable control
70
structure that will take advantage of the inherent scalability and
robustness benefits of DGs. The control structure monitors the sys- 60 heavy load light load
tem by making measurements or state estimation and based on 50
these chooses control actions that are implemented on the system. 40
When active network is applied, a communication and coordina-
30 No SVR
tion protocol could be constructed in order to select the proper
actions and avoid the interference between SVR and DGs. 20 with SVR

10

5. Results discussion 0
No DGs DGs in DGs in No DGs DGs in DGs in
PQ PV PQ PV
Fig. 9 shows the absolute of the line voltage between phases A
and B when both SVR and DGs are connected individually. As shown Fig. 11. Total real power loss at different scenarios.
in the figure, SVR has a significant impact on the voltage profile. DGs
cause raise in the voltage in case of unity power factor mode due
to the impact of the injected active power. The raise of voltages 6. Conclusion
increases when controllable DGs operate in PV mode due to the
impact of the injected reactive power. Fig. 10 shows the number The idea of using the element incidence matrix has been
of required taps when the SVR is activated at different DGs scenar- extended to be used in solving a generalized unbalanced three
ios. The figure shows that the number of required taps is reduced phase power flow for radial distribution systems. This method
when DGs are connected and operate at unity power factor and could be easily implemented because it depends mainly on the con-
they reduced further when DGs operate in PV mode with proper struction of the element incidence matrices. The results have been
set point. Fig. 11 shows the total real power loss at different sce- compared with IEEE results to test the effectiveness of the proposed
narios. The figure shows that DGs have significant impact on the algorithm. Both SVR and distributed generators are incorporated in
system power losses compared with the SVR. Moreover, DGs help the algorithm to show their effect on the voltage profile and system
in achieving greater reduction in the system real power losses dur- losses. The results show that the SVR has a powerful effect on the
ing heavy loading conditions. In contrast, DGs may have negative voltage profile if it is compared with DGs and it cannot be deacti-
impacts on the real power loss in case of light loading conditions. vated even with high penetration of DG even if those DGs operate
in PV mode. In contrast, DGs have a significant impact on the total
system power losses, if it is compared with the impact of the SVR.
The results also, showed that coordination between the SVR and
distributed generators, especially when DGs operate in PV mode is
required and when done properly led to significant improvement
in the system performance.

References

[1] J. Teng, A direct approach for distribution system load flow solutions, IEEE Trans.
Power Deliv. 18 (3) (2003) 882–887.
[2] Y. Zhu, K. Tomsovic, Adaptive power flow method for distribution systems with
dispersed generation, IEEE Trans. Power Deliv. 17 (3) (2002) 822–827.
[3] W.C. Wu, B.M. Zhang, A three-phase power flow algorithm for distribution sys-
tem power flow based on loop-analysis method, Elect. Power Energy Syst. 30
(2008) 8–15.
[4] A.V. Garcia, M.G. Zago, Three phase fast decoupled load flow for distribution
networks, IEEE Proc – Gener. Transm. Distrib. 143 (2) (1996) 188–192.
[5] P.A.N. Garcia, J.L.R. Pereira, Three-phase power flow calculations using the cur-
rent injection method, IEEE Trans. Power Syst. 15 (2) (2000) 508–514.
Fig. 9. The impact of DGs and SVR on the voltage profile when they are connected [6] J.-H. Teng, C.-Y. Chang, A novel and fast three-phase load flow for unbalanced
individually. radial distribution systems, IEEE Trans. Power Syst. 17 (4) (2002) 1238–1244.
[7] A.G. Bhutad, S.V. Kulkarni, S.A. Khaparde, Three-phase load flow methods for
radial distribution networks, in: Conf. on Convergent Technologies for Asia-
Pacific Region, TENCON 2, 2003, pp. 781–785.
Number of Taps [8] T.H Chen, M.S. Chen, K.f. Hwang, P. Kotas, E.A. Chehli, Distribution system
8 power flow analysis – a rigid approach, IEEE Trans. Power Deliv. 6 (3) (1991)
1146–1153.
7 [9] D. Thukaram, H.M.W. Banda, I. Jerome, A robust three-phase power flow algo-
rithm for radial distribution systems, Elect. Power Syst. Res. 55 (3) (2000)
6
191–200.
5 [10] R. Ranjan, B. Venkatesh, A. Chaturvedi, D. Das, Power flow solution of three-
phase unbalanced radial distribution networks, Elect. Power Comp. Syst. 32 (4)
4 Tap_AB (2004) 421–433.
[11] J. Teng, A direct approach for distribution system load flow solution, IEEE Trans.
3 Tap_AC
Power Deliv. 18 (3) (2003) 882–887.
2 [12] J. Teng, A modified Gauss–Seidel algorithm of three-phase power flow analysis
in distribution networks, Elect. Power Energy Syst. 24 (2) (2002) 97–102.
1 [13] M. Marei, E.F. El-Saadany, M. Salama, A novel control algorithm for the DG
0 interface to mitigate power quality problems, IEEE Trans. Power Del. 19 (3)
(2004) 1384–1392.
without DG with DG in PQ with DG in PV [14] Y.A.-R.I. Mohamed, E.F. El-Saadany, A control method of grid-connected PWM
voltage source inverters to mitigate fast voltage disturbances, IEEE Trans. Power
Fig. 10. Number of required taps at different cases. Syst. 24 (1) (2009) 489–491.
1506 H.E. Farag et al. / Electric Power Systems Research 81 (2011) 1499–1506

[15] Y.A.-R.I. Mohamed, E.F. El-Saadany, A control scheme for PWM voltage-source [20] J. Teng, Modeling distributed generations in three-phase distribution load flow,
distributed generation inverters for fast load-voltage regulation and effective IET Gen. Transm. Distrib. 2 (3) (2008) 330–340.
mitigation of unbalanced voltage disturbances, IEEE Trans. Ind. Electron. 55 (5) [21] Y. Zhu, K. Tomsovic, Development of three-phase unbalanced power flow using
(2008) 2072–2084. PV and PQ models for distributed generation and study of the impact of DG
[16] S. Tafreshi, E. Mashhour, Distributed generation modeling for power flow models, IEEE Trans. Power Syst. 22 (3) (2007) 1019–1025.
studies and a three-phase unbalanced power flow solution for radial distri- [22] Radial Distribution Test Feeders. [Online].
bution systems considering distributed generation, Elect. Power Syst. Res. 79 Available:.http://ewh.ieee.org/soc/pes/dsacom/testfeeders/.
(4) (2009) 680–686. [23] H. William, Kersting, Distribution System Modeling and Analysis, CRC Press,
[17] M. Ding, X. Guo, Z. Zhang, Three phase power flow for weakly meshed distribu- LLC, 2002.
tion network with distributed generation, in: Power and Energy Engineering [24] H. Zeineldin, J. Kirtley, Islanding operation of inverter based distributed
Conference APPEEC, 2009, pp. 1–7. generation with static load models, in: IEEE Power Energy Society General
[18] S. Naka, T. Genji, Y. Fukuyama, Practical equipment models for fast distribu- Meeting-Conversion and Delivery of Electrical Energy in the 21st Century, 2008,
tion power flow considering interconnection of distributed generators, in: IEEE pp. 1–6.
Power Engineering Society Summer Meeting (2), 2001, pp. 1007–1012. [25] R. Danaraj, S. Kodad, T. Das, An algorithm for radial distribution power flow in
[19] T. Oomori, T. Genji, T. Yura, T. Watanabe, S. Takayama, Y. Fukuyama, Develop- complex mode including voltage controlled buses, Indian J. Sci. Technol. 1 (2)
ment of equipment models for fast distribution three-phase unbalanced load (2007).
flow calculation, Elect. Eng. Jpn. 142 (3) (2003) 8–19.