You are on page 1of 7

IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO.

3, AUGUST 2007 1019

Development of Three-Phase Unbalanced Power


Flow Using PV and PQ Models for Distributed
Generation and Study of the Impact of DG Models
Sarika Khushalani, Student Member, IEEE, Jignesh M. Solanki, Student Member, IEEE, and
Noel N. Schulz, Senior Member, IEEE

Abstract—With the increased installations of distributed genera- power industry, help defer the costs of expansion, and have pos-
tors (DGs) within power systems, load flow analysis of distribution itive environmental impacts. A DG alters the power flow of the
systems needs special models and algorithms to handle multiple system, thus impacting the overall system losses and voltage
sources. In this paper, the development of an unbalanced three-
profile of the system. A lot of work has been done for radial
phase load flow algorithm that can handle multiple sources is de-
scribed. This software is capable of switching the DG mode of op- power flow solution [1]–[10]. Cheng and Shirmohammadi [11]
eration from constant voltage to constant power factor. The algo- address the load flow solution, incorporating DG for terrestrial
rithm to achieve this in the presence of multiple DGs is proposed. distribution systems. Initial development of such a power flow
Shipboard power systems (SPS) have other special characteristics for radial terrestrial distribution systems with distributed gen-
apart from multiple sources, which make the load flow difficult erator was addressed in [12]. Butler et al. [13] developed a
to converge. The developed software is verified for a distribution three-phase load flow algorithm for shipboard power systems
system without DG using the Radial Distribution Analysis Package
(RDAP). The developed software analyzes an IEEE test case and (SPS) and handle multiple sources by collapsing them into a
an icebreaker ship system. System studies for the IEEE 37-node single source. A comparison of distribution power flows for
feeder without the regulator show the effect of different models and a balanced SPS is addressed by Lewis and Baldwin [14]. An
varying DG penetration related to the increase in loading. System SPS is an ungrounded delta-connected system where genera-
losses and voltage deviations are compared. tion, transmission, and distribution are tightly coupled. Anal-
Index Terms—Distributed generation, IEEE 37-node feeder, ra- ysis of SPS, due to its distinctive characteristics, leads to a fur-
dial distribution analysis package, shipboard power systems. ther complication in distribution power flow because of almost
the same nominal voltages of all generator nodes. This paper
details the power flow development for three-phase unbalanced
I. INTRODUCTION terrestrial distribution systems and SPS with distributed gener-
ator nodes modeled as PQ and PV nodes. Unlike the develop-
O NE of the key calculations for any system is the deter-
mination of the steady-state behavior, which is termed
as distribution power flow for a distribution system. Distribu-
ment in [12], this development can handle multiple DGs and
allows for switching the DG mode from constant voltage to con-
tion automation needs fast and efficient power flow solutions. stant power factor. Comparisons of the power flow results with
The loading of a distribution feeder is inherently unbalanced standard distribution power flow software are made, and lim-
due to a large number of unequal single-phase loads and the itations of the standard software are addressed. An SPS load
nonsymmetrical conductor spacing of three-phase underground flow solution is obtained. System studies showing the impact
and overhead line segments. Due to these factors, conventional of considerable DG penetration on steady-state behavior of the
power flow programs used for transmission system studies do California distribution feeder are shown. PQ and PV represen-
not show good convergence properties for distribution systems. tations along with different penetration levels are incorporated
These programs also assume a perfectly balanced system so that in the case studies.
a single-phase equivalent can be used. The rise in power de-
II. COMPONENT MODELING
mand has led to installation of small power units called dis-
tributed generators (DGs), which give high fuel flexibility. A Because of the limited use of matrix operations, the ladder
DG, if properly planned and controlled, can be beneficial to the iterative method is selected for the load flow. Reference [12]
reviews previous work related to the ladder iterative method.
This method involves two sweeps of calculations. In the forward
Manuscript received March 24, 2006; revised February 20, 2007. This
work was supported in part by NSF Career under Grant ECS 0196559 and
sweep, the end voltages are initialized for the first iteration, and
in part by the Office of Naval Research under Grants N00014-02-1-0623 and currents are calculated starting at the buses at the load end of the
N00014-03-1-0744. Paper no. TPWRS-00158-2006. radial branch and solved up to the source bus by applying the
S. Khushalani and J. M. Solanki are with Mississippi State University, current summation method. The backward sweep starts at the
Mississippi State, MS 39762 USA (e-mail: sakru1@yahoo.com; jigneshm-
solanki@yahoo.com). source bus and calculates voltages using the current calculated
N. N. Schulz is with the Department of Electrical and Computer Engi- from the forward sweep until the load end of the radial branches.
neering, Mississippi State University, Mississippi State, MS 39762 USA The voltages from the backward sweep are used for the next it-
(e-mail: schulz@ece.msstate.edu).
Color versions of one or more of the figures in this paper are available online
eration in the forward sweep calculations. Convergence occurs
at http://ieeexplore.ieee.org. when the calculated source voltage in the backward sweep cor-
Digital Object Identifier 10.1109/TPWRS.2007.901476 responds to specified source voltage.
0885-8950/$25.00 © 2007 IEEE
1020 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 3, AUGUST 2007

TABLE I 2) If the voltage mismatch is within the specified tolerance,


COMPONENT MODELS the PV node voltage has converged to the specified value.
If a voltage mismatch at the PV node is not less than the
specified tolerance, then reactive power compensation Q
generated by that PV node in order to maintain the voltage
at specified value needs to be calculated as follows:

(2)

where is the positive sequence sensitivity impedance


matrix whose size is . The diagonal elements of this
matrix are the absolute value of the positive sequence of the
sum of series line impedances between each PV node and
the source node. The off-diagonal elements are the sums of
the common series line impedances between two PV nodes
and source node. is . Thus, is and
is the magnitude of reactive current injection. The DG can
operate in lagging as well as leading power factor mode.
Thus, the injection of current will depend on the error dif-
ference . If is positive, then reactive power is
supplied by DG, and is negative then reactive power
is absorbed by DG.
The reactive current injection is thus

(3)

where , and are the angles of the converged


voltage at the th node.
3) There is a limit to which the DG can produce reactive
power. This limit is decided in this program by setting the
power factor limits between 0.8 and 1, lagging/leading
Modeling equations for various components of the distribu-
tion system are tabulated in Table I. Details of the models can
be found in [12] and [15]. Reference [16, Tables II and III] helps (4)
to calculate the impedance of overhead lines from the given
phasing, space ID, material, and stranding. Reference [16, Ta- If during computation the reactive power of any of the DGs
bles IV–VI] used the tables to calculate the impedance of under- goes outside its limits, it is fixed at the limiting value, and
ground concentric or tape shield cables. Transformers are mod- this node is now treated as a PQ node. The limiting value
eled as in [15], and only the updated equations are given here. is calculated as the three-phase reactive power limit; thus,
Four different types of transformer connections have been mod- the total per-phase reactive current that the DG can inject
eled: delta-wye, wye-delta, wye-wye, and delta-delta. before its limit is hit is given by
The DG connections can be wye or delta. Depending on the
control, the DG may be set to output power at either constant
power factor for small DG or constant voltage for large DG. (5)
Thus, two types of DG models need to be developed: constant
PQ, modeled as negative load with currents injecting into the
node, and PV nodes for which the calculations are as below. 4) These currents are then added, to the load currents and
1) Initially, the generator real power and positive sequence currents calculated due to DG real power injection ,
voltage are specified. The reactive power is initialized to at the th node
zero. After a load flow has converged, the positive sequence
voltage magnitude mismatch at the PV node is checked

(6)
(1)
Load flow runs again to check the voltage magnitudes and
where of PV nodes. new . If after load flow the of the PV node con-
KHUSHALANI et al.: DEVELOPMENT OF THREE-PHASE UNBALANCED POWER FLOW 1021

Fig. 2. IEEE 37-node feeder.

Fig. 1. Flow chart for load flow with multiple DGs.

verted to PQ node is within the limits, the node is switched


back to PV node. Fig. 1 summarizes the solution algorithm.

Fig. 3. Shipboard Power System.


III. SYSTEM DESCRIPTIONS
The IEEE 37-node feeder is an actual feeder in California.
The data for the feeder were obtained from IEEE test case
archive for distribution feeders [17]. The diagram of the feeder
is shown in Fig. 2. The regulator was removed in order to
clearly see the effect of the DG on the system. The data are
characterized by
Loads—Spot loads, single-phase and three-phase bal-
anced and unbalanced loads, delta connected, constant
kW, kVAR, constant Z and constant I type;
Overhead and Underground Lines—Three-phase lines
with different spacing of phases;
Transformer—Substation and inline transformers are
delta-delta.
The 18-node SPS of an icebreaker ship was also analyzed for
which the data were obtained from [14]. The diagram of the Fig. 4. Renumbered Shipboard Power System.
system is shown in Fig. 3.
The system was represented and renumbered as in Fig. 4 so
that the analysis could be done using the developed unbalanced Overhead and Underground Lines—Three-phase
power flow software. The data are characterized by cables;
Loads—Spot loads, three-phase balanced loads, delta con- Transformer—Inline transformer is delta-delta with tap
nected, constant kW, kVAR; ratio of one.
1022 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 3, AUGUST 2007

TABLE II TABLE III


MODIFIED IEEE 37–NODE VOLTAGES SPS VOLTAGES AND CURRENTS

TABLE IV
SPS GENERATION

a large reactive current injection. The load flow took four


iterations to converge.

B. System Studies
The IEEE 37-node feeder [17] without the regulator was
studied under different DG modeling and varying penetration.
This study helps to demonstrate how the percentage of the
distributed generation as well as the loading on the system
coupled with the DG model affect the final voltage results.
IV. RESULTS This highlights the variation in results depending on the DG
model.
A. Load Flow 1) DG is connected to node 734.
2) DG penetration is defined as
The unbalanced load flow software was developed in
MATLAB. Using Radial Distribution Analysis Package % where
(RDAP) [18]), a commercially available distribution power
flow package, the authors compared results on a common 3) Anticipating the future load growth, DG penetration is in-
IEEE distribution test system [17] to validate the results of the creased by increasing the real and reactive power of loads
developed software. in all the phases of nodes 727, 728, 729, 730, 731, 732, 733,
1) Results of the original feeder (without DG) obtained by 735, 736, 737, 738, 740, 741, 742, and 744. The penetra-
the developed software and those obtained by RDAP are tion of DG is increased in steps of 3.5% up to 35%, which
compared. Voltages are as shown in Table II, and the cur- corresponds to an increase of loading in steps of 5% from
rents are shown in Table VI in the Appendix. Results ob- 5% to 50%, respectively
tained from the developed program closely match the re- 4) Fig. 5(a)–(c) shows a comparison of the voltage deviation
sults obtained from RDAP for the original feeder. The load from 1 p.u. for different DG models and varying DG pen-
flow took one iteration to converge. RDAP can only handle etration. The x-axis of Fig. 5(a)–(c) represents the node
modeling a DG as a negative load, so results are not avail- number, the y-axis of Fig. 5(a) and (b) represents the per-
able for the PV model. centage DG penetration, the y-axis of Fig. 5(c) represents
2) The voltages in per-unit and currents in amperes of 18 node the percentage load increase, and the z-axis of Fig. 5(a)–(c)
SPS are shown in Table III. Node 1 is slack node, and nodes represents the voltage deviation from 1 p.u.
2–4 are modeled as PV nodes. The generation values are Fig. 5(a) is obtained with a DG modeled as a constant
shown in Table IV. The results were compared to results in PQ node. The surface plot of Fig. 5(a) clearly indicates
[14]. These results are obtained with 0.001 p.u. tolerance the voltage deviation is low for the downstream nodes of
for load flow as well as PV node convergence. The cables the feeder. Voltage deviation is high for nodes 744 and
are short, and even a very small voltage difference leads to 728 because load at 728 is a three-phase load, and hence,
KHUSHALANI et al.: DEVELOPMENT OF THREE-PHASE UNBALANCED POWER FLOW 1023

Fig. 5. IEEE 37-node feeder study with PQ and PV models and varying DG penetration level (a) PQ model. (b) PV model. (c) PV model with 50% penetration.
(d) Loss comparison.

the loading is increased in all three phases and 744 is di- is the least when DG is modeled as PV node with 50% DG
rectly connected to this node. The minimum voltage de- penetration.
viation is 0.01933 p.u. Fig. 5(b) is obtained with a DG 5) Twelve cases as shown in Table V are defined as follows.
modeled as a PV node that switches to PQ node in case a) Cases 0–3 have 0% DG penetration.
of a reactive power limit hit. The same observations can be b) Cases 4–6 have varying DG penetration but are mod-
made for surface plot of Fig. 5(b). The minimum voltage eled as PQ node.
deviation is 0.0169 p.u. Comparing surfaces of Fig. 5(a) c) Cases 7–9 have varying DG penetration but are mod-
and (b) demonstrates that the difference in voltage devia- eled as PV node, which switches to PQ node in case
tion for downstream nodes during low DG penetration is of a limit hit.
not much. However, the difference in voltage deviation for d) Cases 10–12 have 50% DG penetration but are mod-
downstream nodes during high DG penetration is consid- eled as PV node, which switches to PQ node in case
erable. Fig. 5(c) is obtained with the DG modeled as PV of a limit hit. However, for all three cases, the limit
node and constant penetration, i.e., 50%, which switches does not hit with this penetration level.
to PQ node in case of a reactive power limit hit. However, 6) Fig. 5(d) shows the system loss comparison for all these
there was no reactive power limit hit for the load increases cases. Losses for cases 1–3, which correspond to no DG
shown. The surface plot of Fig. 5(c) clearly indicates that penetration, are quite high. Losses for cases 4–6, which
the voltage deviation is low for the downstream nodes of correspond to the PQ model, are slightly higher than cases
the feeder, but it increases with an increase in loading. The 7–9, which correspond to the PV model. For 50% DG pen-
minimum voltage deviation is 0.0016 p.u. Based on these etration, losses are much less. Losses for case 12 are more
observations, for an increase in load, the voltage deviation than that of cases 10 and 11 due to an increase in loading.
1024 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 22, NO. 3, AUGUST 2007

TABLE V TABLE VI
CASE SCENARIOS MODIFIED IEEE 37-NODE CURRENTS

Fig. 6. Real power contributions from DG and substation along with total real
power loading.

[19], will provide additional opportunities to study the impact


For a 50% increase in load, losses decrease by 3.4% due to of various DG models on unbalanced power flow analysis.
50% DG penetration.
7) Fig. 6 shows the real power generated by the DG relative APPENDIX
to that from the substation in all the 12 cases. This graph Table VI shows the modified IEEE 37-node currents.
is also a representation of the DG injection. For cases 0–3,
there is no DG penetration, and hence, the entire real power ACKNOWLEDGMENT
contribution is from the substation. For cases 4–6 and 7–9, The authors would like to thank Dr. T. Baldwin from Florida
the DG penetration increases, and hence, the substation State University for the icebreaker shipboard power system data.
contribution decreases. For cases 10–12, DG penetration
is 50%, and thus, the substation contributes 50%. REFERENCES
8) The power flow took one iteration to converge for all cases [1] T.-H. Chen et al., “Distribution system power flow analysis—A rigid
approach,” IEEE Trans. Power Del., vol. 6, no. 3, pp. 1146–1152, Jul.
where the DG was modeled as a PQ node, whereas a max- 1991.
imum of three iterations were required for all cases where [2] R. D. Zimmerman and H. D. Chiang, “Fast decoupled power flow for
DG was modeled as a PV node. unbalanced radial distribution systems,” IEEE Trans. Power Syst., vol.
10, no. 4, pp. 2045–2052, Nov. 1995.
[3] J.-H. Teng, “A direct approach for distribution system load flow solu-
V. CONCLUSION tions,” IEEE Trans. Power Del., vol. 18, no. 3, pp. 882–887, July 2003.
This paper describes a revised general three-phase unbal- [4] W. M. Lin et al., “Three-phase unbalanced distribution power flow so-
lutions with minimum data preparation,” IEEE Trans. Power Syst., vol.
anced power flow algorithm that allows for the incorporation 14, no. 3, pp. 1178–1183, Aug. 1999.
of DGs modeled as either PV or PQ nodes. The algorithm was [5] Y. J. Jang and J. K. Park, “Three-phase power flow method based on
tested on both an IEEE 37-node test feeder and an 18-node fast-decoupled method for unbalanced radial distribution system,” [On-
line]. Available: http://eeserver.korea.ac.kr/~bk21/arch/bk21conf/54.
icebreaker shipboard power system. Comparing the results pdf.
of the unbalanced power flow without DG with RDAP, an [6] J.-H. Teng, “A network-topology-based three-phase load flow for dis-
established software package, demonstrates the accuracy of the tribution systems,” Proc. Nat. Sci. Council, vol. 24, no. 4, pp. 259–264,
2000.
algorithm for an established test case. Advanced studies on the [7] W. Xu et al., “A generalized three-phase power flow method for the
IEEE 37-node test case with DG demonstrate the impact of DG initialization of EMTP simulations,” in Proc. Int. Conf. Power System
model type, size, and load variations on the results. Technology, 1998, vol. 2, pp. 875–879.
[8] H. M. Mok et al., “Power flow analysis for balanced and unbalanced
Further studies, including using the recently published paper radial distribution systems,” [Online]. Available: http://www.itee.uq.
on induction machine models and the IEEE 34-node test case edu.au/~aupec/aupec99/mok99.pdf.
KHUSHALANI et al.: DEVELOPMENT OF THREE-PHASE UNBALANCED POWER FLOW 1025

[9] M. A. Laughton, “Analysis of unbalanced polyphase networks by Jignesh M. Solanki (S’06) received the B.E. degree
method of phase co-ordinates,” Proc. Inst. Elect. Eng., vol. 15, no. 8, from V.N.I.T., Nagpur, India, in 1998, the M.E.
pp. 1163–1172, Aug. 1968. degree from Mumbai University, Mumbai, India, in
[10] W. H. Kersting, Distribution System Modeling and Analysis. Boca 2000, and the Ph.D. degree from the Electrical and
Raton, FL: CRC, 2002. Computer Engineering Department of Mississippi
[11] C. S. Cheng and D. Shirmohammadi, “A three-phase power flow State University, Mississippi State, MS, in 2006.
method for real-time distribution system analysis,” IEEE Trans. Power He is now an engineer working for Open Systems
Syst., vol. 10, no. 2, pp. 671–679, May 1995. International, Minneapolis, MN. He was involved in
[12] S. Khushalani and N. N. Schulz, “Unbalanced distribution power flow research activities at IIT Bombay, Bombay, India. His
with distributed generation,” in Proc. IEEE Transmission and Distri- research interests are power system analysis and its
bution Conf., Dallas, TX, May 2006. control.
[13] M. M. Medina, L. Qi, and K. L. Butler-Purry, “A three-phase load flow
algorithm for shipboard power systems (SPS),” in Proc. IEEE Power
Eng. Soc. Transmission and Distribution Conf. Expo., Sep. 2003, vol.
1, pp. 227–233. Noel N. Schulz (SM’00) received the B.S.E.E. and
[14] T. L. Baldwin and S. A. Lewis, “Distribution load flow methods for M.S.E.E. degrees from Virginia Polytechnic Institute
shipboard power systems,” IEEE Trans. Ind. Appl., vol. 40, no. 5, pp. and State University, Blacksburg, in 1988 and 1990,
1183–1190, Sep.–Oct. 2004. respectively, and the Ph.D. degree in electrical
[15] W. H. Kersting, Distribution System Modeling and Analysis. Boca engineering from the University of Minnesota,
Raton, FL: CRC, 2002. Minneapolis, in 1995.
[16] W. H. Kersting, “Radial distribution test feeders,” in Proc. IEEE Power She has been an Associate Professor in the
Eng. Soc. Winter Meeting, 2001, vol. 2, pp. 908–912. Electrical and Computer Engineering Department
[17] Radial Distribution Test Feeders. [Online]. Available: http://www.ewh. at Mississippi State University, Mississippi State,
ieee.org/soc/pes/dsacom/testfeeders.html. MS, since July 2001. Her research interests are in
[18] WH Power Consultants, RDAP User Manual, ver. 3.0. Las Cruces, computer applications in power system operations,
NM, Sep. 1999. [Online]. Available: http://www.zianet.com/whpower. including artificial intelligence techniques.
[19] R. C. Dugan and W. H. Kersting, “Induction machine test case for the Dr. Schulz holds the TVA Endowed Professorship in Power Systems Engi-
34-bus test feeder-description,” in Proc. IEEE Power Eng. Soc. General neering. She is an NSF CAREER award recipient. She has been active in the
Meeting, Montreal, QC, Canada, Jun. 2006. IEEE Power Engineering Society and is serving as Secretary for 2004–2007.
She was the 2002 recipient of the IEEE/PES Walter Fee Outstanding Young
Sarika Khushalani (S’06) received the B.E. degree Power Engineer Award.
from Nagpur University, Nagpur, India, in 1998,
the M.E. degree from Mumbai University, Mumbai,
India, in 2000, and the Ph.D. degree from the
Electrical and Computer Engineering Department
of Mississippi State University (MSU), Mississippi
State, MS, in 2006.
She is now an engineer working for Open Systems
International, Minneapolis, MN. She was involved
in research activities at IIT Bombay, Bombay, India.
Her research interests are computer applications in
power system analysis and power system control.
Ms. Khushalani received a Honda Fellowship Award at MSU.

You might also like