IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

THE IEEE 34 NODE RADIAL TEST FEEDER AS

A SIMULATION TESTBENCH FOR
DISTRIBUTED GENERATION

Authors James O. Owuor, Josiah L. Munda, Adisa A. Jimoh

Tshwane University of Technology,
Pretoria, South Africa
[email protected], [email protected], [email protected]

Abstract— The advent of distributed generation introduces This paper aims to present results obtained in the
new challenges and possibilities of controlling reactive power implementation of the IEEE 34 node network and compare
and regulating voltage in distribution systems. In addition, them with those from the IEEE Distribution Subcommittee
dynamic behavior of the distribution system is influenced by results. The established network will then be used in the
the presence of Distributed Generation (DG). As a result, the simulation and analysis of various effects of integrating
impact of DG on distribution system radial feeders and laterals Distributed Generation (DG) into typical distribution
needs to be evaluated in terms of steady state normal and networks. The effects of different load models also heavily
transient conditions before implementation. Several computer impact on the accuracy of simulations. Results of such
programs exist to carry out such simulations and this paper
analyses can provide useful information for distribution
focuses on DigSILENT PowerFactory. Power Factory provides
system researchers and engineers.
steady state solutions for balanced and unbalanced load flow
for both transmission and distribution networks, short circuit The rest of the paper is presented in the following
calculations and, it also provides dynamic solutions for manner. Section 2 – covers the modeling of the IEEE 34
transient studies. The main objective of this work is to Node Test Feeder in DigSilent. In section 3 the steady state
establish a test bench in PowerFactory and report steady state results are compared with the results for the IEEE
results. The simulation test bench is to be used in ongoing benchmark system. Section 4 gives the conclusions and
research work for analysis of the impact of various forms of section 5, the references.
DG on distribution networks. The IEEE 34 Node test feeder is
modeled and results are compared with those of the IEEE II. IEEE 34 NODE TEST FEEDER MODEL
distribution system analysis subcommittee.
The IEEE Distribution System Analysis Subcommittee
Keywords- Distributed Generation (DG), Radial Distribution developed and published data for a variety of unbalanced
Networks radial distribution test feeders the primary purpose of which
is the evaluation of power system analysis software [4].
I. INTRODUCTION These test systems are for use by software developers and
Distributed generation (DG) is steadily acquiring a small, field engineers for validating their studies. The IEEE 34
yet significant share of electrical power generation. Utility Node Test Feeder represents an actual feeder in Arizona and
restructuring provides an economic driver, but DG also has the following features reflecting the physical world [8].
promises ancillary technical benefits to the distribution 1) Each section of the distribution system is modeled by
system infrastructure [9].
actual phase impedance values.
Distribution networks exhibit a wide range of 2) The system includes single and two phase laterals.
complexities due to the nature of loads and circuit 3) Loads on each phase of each section are specified in
connections [8]. Various commercial software platforms are real and reactive power
available for distribution system analysis. The choice of 4) Distributed load models represent load on feeders
software primarily depends on ease of use, suitability for the
with closely spaced load taps.
study amongst other factors. It is, however, a challenging and
time consuming task to get acquainted with all the details 5) The system includes voltage regulators and
and specialties of such programs, and many users are not capacitive reactive power compensation.
inclined to keep tabs on the evolution of similar products or 6) The system has very long distributions lines and is
do not even know them. Choice of a suitable network for lightly loaded.
studies and implementation in software for accurate 7) There is an in-line transformer for reducing voltage
simulations can also be a challenging and time consuming to 4.16 kV for a short section of the feeder.
task. Cost and actual availability of software is often an
additional challenge.

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

The configurations and detailed parameters of the 848

published feeder can be downloaded from the website at 822 846

http://ewh.ieee.org/soc/pes/dsacom/ testfeeders.html. A one 820 844

line diagram of this feeder is shown in Figure 1. 818

864
842

802 806 808 812 814 850 824 826 860 836
The IEEE 34 node network is chosen for this work 816
858
834
840

because of its topological features and wide variety of 832 862

800
components, which are representative of a typical rural 810
888 890
838
distribution network with single and three phase laterals and 852

long feeders which could incorporate distributed generation.

This distribution network is operated at 24.9 kV and has two 828 830 854 856

Figure 1. The IEEE 34 node test feeder.

voltage regulators, two capacitor banks and one low voltage
lateral at 4.16 kV. The substation is rated at 2500 kVA, with III. COMPONENT MODELS
a 69kV/24.9 kV transformer. Loads are modelled as three-
phase (balanced or unbalanced), single-phase spot or A. Shunt capacitors
distributed loads. Three-phase loads can be connected in wye The three-phase Y-N model was used to model the shunt
or delta while single-phase loads can be connected line-to- capacitors that are connected at nodes 844 and 848. The
ground or line-to-line. Loads are modeled as constant kW values of these capacitors are calculated using the relations
and kVAr (PQ), constant impedance (Z) or constant current in (1), where Q is each phase’s rated kVar, ω = 120π , and
(I) [2] and the program used for the original IEEE reults is
WH Power Consultants RDAP [3]. U is the rated phase to ground voltage. The Q values used
are given in (IEEE 34) and the loss factor tan δ is assumed
While the internal models for some components vary for
to be zero.
some components between the original software and
DigSilent PowerFactory in terms of parameters required, the Q
model is intended to be equivalent to the IEEE circuit and C= (1)
results in the implemented model were substantially similar ωU 2
to those in the IEEE document, which provides a reasonable
measure of validation for the DigSilent implemented model. B. Transformer and Voltage Regulator Models
The substation and inline transformers were modeled as a
The substation voltage in the IEEE results are an exactly two winding transformer with 32 tap positions. The winding
balanced set at 1.05 pu with angles of 00/-1200/1200 positive sequence leakage impedances were distributed
whereas in the DIgSILENT model the voltage was set to equally between the LV and HV sides 0.95+j2.04 pu on each
1.048 pu. It is not possible to set individual phase angles in side, the same distribution of reactances was adopted for
DIgSILENT from the external grid model, there is only both the transformers and voltage regulators. As
provision for setting the angle of phase A and all the other recommended in [6], when tap changes need to take care of
angles are calculated relative to it, the closest achievable daily variations in system conditions, they allow the ratio to
angle for phase A that resulted in minimum differences in vary in the range of ±10% to ±15%. The latter is adopted in
angles for all phases was found to be 1.990 , while this this work, with each tap change on the HV side resulting in
brings the angles for phases B and C to within tolerable an additional ±0.6469 kV (0.9375%), the tap position was
limits, the angle of phase A voltages have notably different adjusted to give measurably proximal results to the IEEE 34
values from the IEEE benchmark results, the use of benchmark results. The detailed models of the transformers
autotransformers in the place of voltage regulators specified and regulators and the relations between the mathematical
in the original IEEE 34 model means individual phase angles parameters in the models and the parameters in the type and
cannot be adjusted accordingly. Nevertheless, the voltages of element dialogs can be found in [1].
all nodes fall within 0.95 pu and 1.05 pu except for that of
node 890. The voltages for this node are 0.8946 pu, 0.9397 The two voltage regulators were implemented using
pu and 0.9202 pu for phases A, B and C respectively, which YNyn0 autotransformers. The tap positions were set such
fall out of the margin. This is due to the fact that the line that there was parity with the voltages given in [4] load flow
between nodes 888 and 890 is relatively long, at about 3.22 results. The OLTC basic arrangement for the transformers
km and given the voltage level of 4.6kV, this result is and regulators is shown in figure 3, and the equivalent
expected. This intuitively indicates a node at which voltage diagram [1] with an ideal transformer connected to the HV
correction measures are required. This undervoltage at the side is shown in figure 4. The OLTC controller keeps the
end of lateral 832 – 890 is also reflected in the base case substation secondary bus voltage U lv constant within the
IEEE load flow.
range
Three possible locations for three phase DG in future
studies include the main feeder and the three - phase laterals U lower limit ≤ U lv ≤ U upper limit (2)
832-890, 834-848 and 836-840.
and
U lower limit = U set − 0.5U DB (3)

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

Figure 4. Nominal π model for the distribution lines.

1
1 + z 'Y ' l 2
A −1 2 1
Yπ = = = ( G ' l + jωC ' l ) (10)
Figure 2. OLTC Basic arrangem
ment. B Z 'l 2
Thus the lines are moodeled in PowerFactory using
discrete R,L,C and G parameeters and are hence suited to both
steady state and transient stuudies. The equivalent pi circuit is
shown in figure 3. In this woork, the leakage conductances to
ground were assumed to be zero. The various tower
geometric parameters were also a taken into account in order
Figure 3. OLTC Equivalent cirrcuit. to form the reactance and resistance matrices.
U Upper limit = U set + 0.5U DB (4)
D. Load Models
where U set is the set point voltage annd U DB is the There were two types of load, spot loads, and distributed
deadband. In practice, many OLTCs are opperated with the loads. Spot loads were placeed at the appropriate nodes. The
Line Drop Compensator disabled for thhe purposes of distributed loads were modelled using the exact lumped load
simplicity in control and avoidance of error
e [10]. This model described in [3]. Inn this load model, 2/3 of a
assumption is therefore adopted in this workk henceforth and distributed load was put at a ¼ length of the line and the
the LDC component is neglected. other 1/3 of the distributed looad was put at the end of the line
(Figure 5). A comparison with w the [2] model gave correct
C. Line models computation results for the voltage
v drops and the line power
The lumped parameter pi model was ussed to model all losses for the constant poweer, constant current and constant
the lines by virtue of their relatively short lenngths, this model impedance load models specified in the same document. The
can be used with acceptable accuracy forr short lines or results are attached in sectionn 3.
relative long lines at low frequencies. The modeling of loads in stability studies is a complex
Defining as the series impedance per-unit
p length problem due to the unclear nature of aggregated loads [5].
corresponding to the line voltage drop and
a ’ as shunt Load models are typicallly classified into two broad
admittance representing current drawn to earth per unit categories: static and dynamiic. The load models specified in
length we have. [2] can be modeled using steeady state and dynamic models.
Steady state analysis is usually conducted using conventional
Z ' = R '+ jω L ' or extended power flow programs. Dynamic analysis
(5) complements steady state analysis,
a is more detailed, and
gives a better understandingg of voltage and other stability
Y ' = G '+ jωC '
(6) phenomena. Accurate anallysis for any study requires
appropriate representation off load models. Representation of
After series expansion of the hyperbolic funcctions that define load characteristics for staticc and dynamic analysis is well
the ABCD parameters, and considering onlly the A and B provided for in PowerFaactory and include constant
parameters we have impedance, constant current and constant power load
1 models.
A = cosh γ l = 1 + Z ' Y ' l 2 + ( Z 'Y ') l 4 + ........ (7)
24
⎛ sinh γ l ⎞ ⎡ 1 1 ⎤ (8)
⎟ = Z ' l ⎢1 + 6 Z ' Y ' l + 120 ( Z ' Y ' ) l + ......⎥
2 2 4
B = Z 'l ⎜
⎝ γl ⎠ ⎣ ⎦
If we consider up to the second order term inn expansions
(vii) and (viii) above, we can approximate thhem as follows
Zπ = B = Z ' l = R ' l + jω L ' l (9)
Figure 5. Exact lum
mped distributed load model.

The three basic load models

m are characterized in the
following manner [7].

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

1) Constant Impedance Load Model (constant Z): A Dynamic load models are also provided for with active
static load model where the real and reactive power is and reactive power Pext , Qext as input variables and delayed
proportional to the square of the voltage magnitude. It is voltage and frequency time contants, cosine and sine of
also referred to constant admittance load model. voltage angles as state variables for r.m.s simulations.
2) Constant Current Load Model (constant I): A static
load model where the real and reactive power is directly All the above load characteristics catered for in
PowerFactory make it an ideal tool for research work in
proportional to the voltage magnitude.
which voltage/frequency dependence of loads is an important
3) Constant Power Load Model (constant PQ): A static factor. Ongoing research using the feeder configuration
load model where the real and reactive powers have no includes effects of DG on different load models.
relation to the voltage magnitude. It is also referred to as a
constant MVA load model. IV. RESULTS
The above three load models can be described by the There was in general substantially good agreement
following polynomial expressions [6]. between the [2] and PowerFactory results. Load flow results
were obtained, the node voltages and angles were compared
a with the IEEE results. The percentage difference in voltage
⎛V ⎞
P = P0 ⎜ ⎟ (7) magnitude for the nodes with the corresponding IEEE result
⎝ V0 ⎠ was obtained as shown in fig. 7. Table 1 shows error values
for the node voltages and angles.

⎛V ⎞
b
The highest maximum error in phase A occurred at node
Q = Q0 ⎜ ⎟ (8) 814. The highest difference in phase angle of phase A
V
⎝ 0⎠ occurred at node 802. These results can be attributed to the
P and Q in (7) and (8) are the real and reactive powers of the fact that the grid source model can only be represented as a
load respectively and the subscript 0 identifies the values of balanced three phase source and the individual phase
the respective variables at the initial operating conditions. voltages and angles cannot be adjusted, the same is the case
The exponents a and b depend on the type of load that is for the substation transformer. These results can be expected
being represented, e.g. for constant power load models a = b given that the system has a high degree of unbalance
= 0, for constant current load models a = b = 1 and for between phases. Figure 7 shows a voltage profile for all the
constant impedance load models a = b = 2. nodes and it is evident that the average voltages for all nodes
fall within 0.95 pu and 1.05 pu, the undervoltage for phase A
An alternative model that is widely used to represent at node 890 notwithstanding.
voltage dependency of loads known as the ZIP model is
composed of the three models in items 1,2 and 3 above Table 1 compares the base case IEEE phase voltage
(Kundur 1994) and is modeled in power factory as follows magnitudes with those obtained in the PowerFactory results
and the absolute percentage magnitude difference.Table 1
also compares the base case IEEE phase voltage angles with
those obtained in the PowerFactory results and the absolute
⎛ ⎛ V ⎞ e _ aP ⎛V ⎞
e _ bP
⎛V ⎞
e _ cP
⎞ (9) percentage magnitude difference.
P = P0 ⎜ aP ⎜ ⎟ + bP ⎜ ⎟ + (1 − aP − bP ) ⎜ ⎟ ⎟
⎜ V ⎝ V0 ⎠ ⎝ V0 ⎠ ⎟
⎝ ⎝ 0⎠ ⎠ Three possible locations for three phase DG in future
studies include the main feeder and the three – phase laterals
⎛ e _ aQ e _ bQ e _ cQ
⎞ (10) 832-890, 834-848 and 836-838, base case voltage profiles
⎛V ⎞ ⎛V ⎞ ⎛V ⎞
Q = Q0 ⎜ aQ ⎜ ⎟ + bQ ⎜ ⎟ + (1 − aQ − bQ ) ⎜ ⎟ ⎟ were computed and results are shown in figures 8, 9 and 10.
⎜ ⎝ V0 ⎠ ⎝ V0 ⎠ ⎝ V0 ⎠ ⎟
⎝ ⎠
The end of lateral 832-890 is the only one in the network
where supplied by a 24.9/4.16 kV transformer, the base case profile
1 − aP − bP = cP (11) in fig.8 is indicative of a heavily loaded load centre and the
and sloping profile is typical. It is therefore a natural case for
1 − aQ − bQ = cQ (12) several studies involving the impact of connecting DG.
Lateral 834 – 838 has a rising profile as shown in fig.9, due
The inherent load characteristics can be modeled by to the shunt compensators installed at node 844, while the
specifying the exponents voltage profile is obviously improved by the compensation,
e _ aP, e _ bP, e _ cP and e _ aQ, e _ bQ, e _ cQ to achieve the lateral and it’s nodes could form a case for introducing
constant power, current or impedance behavior. The relative DG and performing voltage and other stability studies.
proportion of each coefficient can be freely defined using Lateral 836 -840 shown in fig. 10 has a relatively flat profile
the coefficients . This flexibility of specifying loads that is because the nodes are relatively lightly loaded. It forms a
provided for in PowerFactory will play an important role in case for studies involving the effect of increased loading and
simulations that will be carried out in further research work introduction of DG.
on the IEEE 34 test bench model.

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

TABLE 1: Base case % maximum and average node voltage and angle
errors vs IEEE results.

%Error Va δa Vb δb Vc δc
Maximum 4.4 41.14 3.69 1.22 5.13 2.46
Average 2.13 14.71 1.02 0.63 1.52 0.66

Figure 9: Voltage prrofile for lateral 834-848.

Figure 6. Voltage profiles comparison between IEEE annd DigSilent Models

for phase A.

Figure 10: Voltage prrofile for lateral 836-840.

V. CON
NCLUSIONS
This paper presents the use of DigSILENT
PowerFactory’s suitability forfo studies being carried out by
the authors involving integgration of DG into distribution
Figure 7. Voltage angles comparison between IEEE annd DigSilent Models networks, it also presents thee suitability of the IEEE 34 node
for phase A. network as a platform for thee same. It is not intended to be a
comparison between differeent software packages. Indeed
many suitable software packkages are available that can be
used in different degrees forr similar studies. The choice of
software and network weree primarily dictated by among
others a literature study of available
a software and software
that is actually available to thhe authors. Also considered was
the need for the model off a real network that exhibits
unbalanced phase loading. Inn addition, the need for a variety
of components and voltage levels that provides possibility
for various studies involvving integration of DG was
considered.
The modeling results substantially agree with those of the
IEEE Distribution Analysiis Subcommittee. The phase
Figure 8: Voltage profile for lateral 832-890. imbalances are clearly exhibbited from the voltage profiles.
Nodes with undervoltages that would provide interesting
case studies for stability are also
a clearly identified.
Results indicate that bothh the chosen software platform
and network are very suitaable for use by the authors in
ongoing research work invvolving integration of DG into
distribution networks.

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

REFERENCES
[1] DIgSILENT GmbH. "TechRef ElmTr2 V3." Two-
Winding Transformer (3-phase), Gomaringen:
DIgSILENT GmbH, 2007.
[2] IEEE 34 Node TestFeeder. "Distribution System
Analysis Subcommittee Report." IEEE transactions,
2003.
[3] W.H. Kerstin, "Distribution system Modeling and
Analysis", FL: CRC Press LLC, 2002.
[4] W H.Kersting, "Modeling and Analysis of Rural
Electric Distribution Feeders",. IEEE, 1992.
[5] Kodsi, Sameh and A. Canizares. "Modeling and
Simulation of IEEE 14 Bus System With FACTS
Controllers." IEEE Transactions (IEEE
TRANSACTIONS) 3 (2003).
[6] Prahba Kundur,. Power System Stability and Control.
McGraw Hill, 1994.
[7] M.M.A. Mithulananthan, C.A. Canizares and J. Reeve,
"Distribution System Voltage Regulation and Var
Compensation for Different Static Load Models."
Edited by IJEEE. (IJEEE) 37, no. 4, October 2000, pp
384-395.
[8] N. Mwakabuta and A. Sekar, "Comparative Study of
the IEEE 34 Node Test Feeder under Practical
Simplifications.", 39th North American Power
Symposium (NAPS 2007). IEEE, 2007.
[9] G. Narain and L. Hingorani, “FACTS Concepts and
Technology of Flexible AC Transmission Systems".
IEEE Press Text Book, 2000.
[10] T.A. Short, "Electric Power Distribution Handbook".
CRC Press LLC, 2004.
[11] T.Van Cutsem and C.D. Vournas, "Voltage Stability of
Electric Power Systems", Vol. 4. Kluwer Academic
Publishers, 1998.
[12] T.Van Cutsem, "Voltage Instability:
Phenomena,Countermeasures, and Analysis Methods."
(IEEE TRANSACTIONS), AUGUST 1999.

978-1-61284-993-5/11/$26.00 ©2011 IEEE