IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO.

2, APRIL 2003

619

Determination of Allowable Penetration Levels of

Distributed Generation Resources Based on Harmonic
Limit Considerations
Arijit Bhowmik, Member, IEEE, Arindam Maitra, Member, IEEE, S. Mark Halpin, Senior Member, IEEE, and
Joe E. Schatz

AbstractAnalytical methods are formulated in this paper to

predict allowable distributed generation resources on a radial
distribution feeder before voltage harmonic limits (as specified in
IEEE 519-1992) are exceeded. The final result is a determination
of allowable penetration levels of distributed generation resources
for a range of distribution feeders. Closed-form equations are
derived for typical feeder designs and different load allocation
patterns. These equations are validated with conventional harmonic analysis simulations which show that the closed-form
equation predictions are slightly conservative. These methods
will serve as efficient tools for power system planning and design
without requiring exhaustive studies to determine the impact of
distributed generation resources.
Index TermsDistributed generation, harmonic limits.

I. INTRODUCTION

EFORE the advent of distributed resources, the vast majority of distribution circuits (radial, open loop, or network
type systems) were usually designed to operate without any generators along the feeder. This implied that real power flowed
from a centralized generating unit downward through the distribution feeder to the end-users. The combination of technology
innovation in power electronics, electricity deregulation, economics, customer value, and energy demand are beginning to
converge causing the electric power industry to shift from a few
large concentrated generation centers to a more distributed and
dispersed generation infrastructure. From a harmonic distortion
standpoint, the two key questions that arise from the potential
use of distributed generation are as follows.
1) What (harmonic) impact will these distributed resources
have on a distribution circuit?
2) How much of the existing feeder load can be supplied
by these distributed resources before IEEE 519-1992 [1]
voltage limit compliance is violated?
Before distributed generators can be evaluated as an alternative to release feeder capacity, the utility, from a planning
prospective, must be able to predict the maximum distributed
Manuscript received January 10, 2001.
A. Bhowmik is wih Open Systems International, Inc., Minneapolis, MN
55447-1286 USA.
A Maitra is with EPRI PEAC Corp., Knoxville, TN 37932 USA.
S. M. Halpin is with the Department of Electrical and Computer Engineering
at Auburn University, Auburn University, AL 36849 USA.
J. E. Schatz is with Southern Company Services, Birmingham, AL 35291
USA.
Digital Object Identifier 10.1109/TPWRD.2003.810494

resources that can be connected to an existing feeder before harmonic related problems are encountered. The type and severity
of harmonic problems will depend on switching device technology, the nature of the characteristic harmonics, equipment
ratings, and loading conditions of the host distribution feeder.
The motivation behind this work is to recognize that from a
utility standpoint, distribution system configurations and end
use equipment operations vary widely and are difficult to predict. The number of studies and level of detail required, coupled
with this high degree of variability, make direct simulation evaluations an unfavorable approach for drawing broad conclusions.
From a harmonic modeling and simulation standpoint, a distributed generator is usually a converter-inverter type unit and
can therefore be treated as a nonlinear load injecting harmonics
into the distribution feeder. Under the present framework of 5191992, the supplier of electricity is responsible for the quality of
the voltage supplied. The end-user is responsible for limiting
harmonic current injections based on the size of the end-use load
relative to the capacity of the system. Distributed resources such
as microturbines, fuel cells, and photovoltaic arrays are small
relative to system capacity, but the smaller sizes are much more
likely to achieve significant penetration levels for economic reasons.
Present-day documents (standards under development) concerning interconnection requirements for distributed generation
specify compliance with the first row in the IEEE 519 current
limit tables. It is therefore reasonable to address the question of
allowable penetration by assuming a harmonic injection at the
limit levels specified. To obtain meaningful results, however, it
is necessary to consider a large variety of distribution system
designs and loading patterns and how the distributed resources
can be allowed to serve the load. To capture these wide ranges
of variations, a large number of individual (but specific) studies
would have to be conducted and general conclusions drawn.
In most planning applications, the details are not well known
enough to permit such investigations even if time was not an
issue.
Most of the previous publications on distributed harmonic
sources have dealt with relatively simple systems or specific
distribution feeders with specific harmonic injection patterns
[2][6]. Florescent lights, personal computers, battery chargers,
and other loads have received significant attention due to the
possible proliferation of these harmonic-producing loads. While
these results have been useful for developing a better understanding of the issues surrounding allowable penetration levels,

0885-8977/03$17.00 2003 IEEE

620

Fig. 1.

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 2, APRIL 2003

Block diagram of a distribution feeder.

they have usually been too system- or load-specific to allow the

broad generalizations needed for planning purposes.
In the following sections, analytical methods and techniques
for predicting allowable penetration levels are developed based
on generic feeder designs and typical load patterns. These results will allow utilities to predict, in the general planning stages
where little specific data is known, the maximum allowable distributed resources that can be installed. Once this maximum
penetration level is determined, it is straightforward to determine the amount of existing feeder load that can be covered
using distributed resources without exceeding voltage harmonic
limits. The direct applications of planning levels of penetration are in the determination of capacity (release) credits, rate
structures, incentives, and other financial concerns.
II. ASSUMPTIONS
Assumptions made to develop closed-form solution techniques are discussed in this section. The method formulation is
based on a balanced 3-phase, 4-wire radial distribution feeder.
Fig. 1 shows a basic (per-phase) diagram including the major
elements considered.The assumptions made are as follows.
1) Positive, negative, and zero sequence harmonic effects
can be considered separately; distribution lines are modeled using appropriate sequence network circuits.
2) Linear loads (when included) can be modeled using
equivalents.
simple
3) Nonlinear loads can be modeled using current injections
at appropriate harmonics.
4) The series impedance of the feeder and substation at harmonic frequencies can be determined from the fundamental frequency values of resistance and reactance.
5) Skin effect and line charging capacitance can be neglected.
6) Power factor correction capacitors are neglected.
7) The effects of phase angle diversity between distributed
harmonic sources can be ignored.
Assumptions 1, 3, 4, and 5 are acceptable approximations that
are often used in harmonic studies where either great accuracy
is not required or insufficient data exists. Assumptions 2, 6, and
7, however, require discussion and justification.
Individual linear load models are not used during initial
method formulation. Rather, the basis for the entire approach
for predicting the allowable penetration level of distributed
resources is based on the total feeder load as measured at the
substation. To be practical, total feeder loadings are derived
based on ampacities of three types of feeders [7] using standard

distribution primary system conductors (4/0 ACSR, 477 AAC,

and 795 ACSR). These total loads are then allocated along
the various feeders in order to demonstrate the conservatism
obtained by neglecting them in the initial development of the
method.
Because feeder optimization capacitor size and location are
difficult to generalize (for planning purposes), capacitor effects
on harmonic voltages along the feeder are not included during
method formulation. This important assumption could suggest
that results obtained could be seriously under-conservative.
From an economics perspective, however, the costs of using
tuned capacitor banks (to avoid problematic resonance) can be
easily shown to be very minor (approximately twice the cost of
fixed capacitor-only banks) relative to the savings offered by
the potential system capacity released by distributed generation
resources. For this reason, capacitor banks can be assumed to
either be not present or sufficiently tuned to avoid harmonic
resonance.
As shown in [8], the effects of phase-angle diversity are negligible on the harmonic voltage profiles for radial distributed
feeders at lower frequencies. As will be shown in the next section, the lower frequencies tend to be the most limiting, so ignoring higher frequency effects is not a significant issue.
III. DETERMINATION OF THE MOST LIMITING HARMONICS
Due to the linear nature of a power delivery network (for harmonic analysis), the severity of individual harmonics can be
evaluated independent of each other. The basic equation is presented in (1) and gives the required short-circuit ratio (scr) such
that a current harmonic limit value (as specified in the existing
519-1992 current harmonic limit tables) will produce the corresponding voltage harmonic limit value (taking 3% voltage distortion for individual harmonic) at any harmonic. Calculated scr
values that are less than 519-1992 tabulated scrs indicate built-in
conservatism or diversity allowance in the current limit table.
Calculated values much less than the 519-1992 range indicate
significant conservatism or diversity allowance.
(1)
The plot of calculated scrs for individual harmonics (up to
thirty-ninth) is shown in Fig. 2 based on using the first-row
current harmonic limit and a 3% individual voltage harmonic
limit in (1). The (relative) degree of conservatism or diversity
allowance can be estimated by considering the value at any harmonic with respect to the scr ratio of 20 that identifies the first
row in the 519-1992 current limit tables. It can be clearly seen
that the least conservative (most restrictive) current harmonic
limits are the ninth (zero sequence) and seventh (positive sequence).
Only odd harmonics limits are shown in Fig. 2. The generation of even harmonics is more restricted. Although IEEE
519 limits even-order harmonic values to 25% of the (corresponding) odd-order harmonics, their effect will be negligible
for balanced three phase converter-inverter circuits. In addition,
it is quite evident from Fig. 2 that the ninth harmonic current
may be the most restrictive, but it may be that some (or all)

BHOWMIK et al.: DETERMINATION OF ALLOWABLE PENETRATION LEVELS

621

Fig. 3. Uniformly distributed generation (single phase).

Fig. 2. Variations of

scr < 20 (supply voltages  161 kV).

distributed generators may not produce appreciable ninth harmonic. Nondiscretionary use of the ninth (or any other nontypical) harmonic could lead to over-conservative conclusions.

IV. METHOD FORMULATION

Based on Fig. 2, it is clear that limiting harmonic current
injections of distributed resource to the values specified
in 519-1992 will not necessarily produce a 3% individual
harmonic voltage in practical circuits. The key question is,
How much of a given feeder load demand can be supplied
by harmonic-injecting distributed resources before individual
voltage harmonic limits (3%) are exceeded? To approach the
problem in a general way, closed-form equations are needed.
The voltage harmonics at individual buses are evaluated
based on the total voltage drop from the substation to that bus.
The method formulation is derived for generic feeder designs
and typical load allocation patterns. A similar approach [9] has
been used for many 60-Hz studies to evaluate voltage drops on
a primary feeder. Because of the principle of superposition, no
fundamental-frequency source will be present in the network
when a harmonic is evaluated. Determining voltage drop,
therefore, is exactly the same as determining harmonic voltage
produced.
It is assumed that distributed generators will carry a part of
the existing load demand of a distribution circuit. Equations to
calculate the voltage distortion levels at the most restrictive harmonic frequency (as identified in the previous section) are developed in this section for three different types of load distributions. With capacitor models neglected, an individual voltage
harmonic will be maximum at the far end of any radial distribution feeder. The allowable substation current at this most restrictive harmonic frequency will determine the total demand that
can be supplied by distributed resources along the feeder before
the voltage harmonic limit (3%) at the far end is exceeded.
The variables used in the derivations in this section are defined as follows:
( ).
( mile).
worst case voltage distortion (V).
branch current at a distance
from the substation.
total current at the substation (A).
distance from the substation (miles).
length of the feeder (miles).
harmonic number.

Fig. 4. Linearly increasing generation (single phase).

A. Uniform Feeder Loading

The most simplistic loading (and associated distributed
generation) pattern is the uniform loading shown in Fig. 3.
Note that only the harmonic-producing distributed generation
is shown; linear loads are not yet considered except to establish
the pattern of the harmonic injections.
at any point x on the distribution
The line current
feeder is given in (2). The voltage drop at the extreme end of
the feeder due to this line current is given in (3). Note that (3)
is perfectly generic and can be used for any frequency provided
that appropriate impedances are used.
(2)
(3)
By completing the integral in (3), the total substation current
can be determined as shown in (4). This substation current will
be used to predict the allowable distributed generation for the
uniform load distribution case.
(4)

B. Linearly Increasing Loading

A more realistic load (and distributed generation) distribution
is the linearly increasing load pattern shown in Fig. 4.
at any point x on the distribution
The line current
feeder is evaluated, as shown in (5). The voltage drop at the
extreme end of the feeder due to this line current is given in

622

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 2, APRIL 2003

TABLE I
FEEDER CONFIGURATIONS

TABLE II
LINE IMPEDANCES (
=mile)

Fig. 5.

Linearly decreasing generation (single phase).

(6). Note that (6) is perfectly general and can be used for any
frequency provided that correct impedances are used.

TABLE III
SUBSTATION IMPEDANCES (
)

(5)
(6)
By completing the integral in (6), the total substation current
can be determined, as shown in (7). This allowable substation
current will be used to predict the allowable penetration of distributed generation for the linearly increasing load case.

TABLE IV
EVALUATION OF I
FOR SEVENTH HARMONIC

(7)

C. Linearly Decreasing Loading

As a final load (and distributed generation) distribution, consider the linearly decreasing load pattern shown in Fig. 5.
at any point x" on the distribution
The line current
feeder is given in (8).
(8)
in (8) can be represented as a function of the substation
current and the length of the feeder, as shown in (9). Using (8)
and (9), the total voltage drop at the extreme end of the feeder
due to this line current can be found, as shown in (10).
(9)
(10)
from (9),
By completing the integral in (10) and substituting
the total substation current can be determined as shown in (11).
Note that (11) is completely general and can be used for any
frequency provided that correct impedances are used.
(11)

V. PREDICTING THE ALLOWABLE NUMBER

OF DISTRIBUTED RESOURCES
The use of the closed-form expressions developed in the
previous section to determine distributed resource penetration
levels for typical feeder configurations is demonstrated in this
section. A 12.47-kV radial distribution feeder is selected for
the evaluations. Based on 519-1992 voltage limit criteria, the
) is set to 3% of the
maximum allowable voltage distortion (
nominal voltage. As shown in Table I, the feeders considered
include the larger and the smaller conductor sizes that are
typical of many practical distribution circuits. The sequence
impedance line parameters for the various feeder conductors
are shown in Table II. The substation sequence impedances are
shown in Table III and were derived from typical fault MVA
levels at 12-kV substations.
using (4), (7), and
The calculated values of allowable
(11) (to produce 3% voltage harmonic magnitude) for the different feeder configurations are shown in Tables IV and V for
positive (seventh) and zero (ninth) sequence harmonics, respectively. The ratio of these allowable substation currents to the
maximum current levels will predict the allowable penetration
percentage of distributed resources.

BHOWMIK et al.: DETERMINATION OF ALLOWABLE PENETRATION LEVELS

TABLE V
EVALUATION OF I
FOR NINTH HARMONIC

TABLE VI
ALLOWABLE PERCENTAGE OF DISTRIBUTED GENERATION FOR POSITIVE
SEQUENCE HARMONICS

The maximum levels of substation current at each harmonic

are determined based on the given feeder capacity, which
is taken to be the maximum demand current . Assuming
that 100% of this maximum demand current is covered by
distributed generation and that all distributed generators are
limited by the first row in 519-1992 current table, the maximum
substation current at each harmonic can be determined as
. It should be noted that a value of the
greater than 1 means that 100% of the load
ratio
can be covered by distributed generation resources without
encountering the 3% individual voltage harmonic limit. These
ratio results are illustrated in Tables VI and VII. These ratios
indicate explicitly how much of the feeder demand can be
supplied using distributed resources without encountering the
3% individual voltage harmonic limit.
When interpreting the results in Tables VI and VII, it is important to recognize that the values indicate a percentage value. Because each feeder has a different maximum load current, it could
appear that, for example, the 10-mile Feeder-3 can support less
distributed generation than the 10-mile Feeder-2. Because the
base values for Feeders 2 and 3 are different (200 A versus 300
A), the actual amperes of distributed generation supply current
that can be allowed is in fact greater for Feeder-3 even though
the reported percentage is smaller.
Conventional harmonic analysis simulations were performed
with linear load models on the feeder circuit. The results of these
simulation-based predictions compared to those obtained from
the closed-form techniques was also studied. With linear load
models included in these simulations, the voltage distortion decreases as the load increases (due to resistive damping). From
this result it can be concluded that the closed-form equations,

623

TABLE VII
ALLOWABLE PERCENTAGE OF DISTRIBUTED GENERATION
SEQUENCE HARMONICS

FOR

ZERO

TABLE VIII
ALLOWABLE PERCENTAGE OF DISTRIBUTED GENERATION FOR POSITIVE
SEQUENCE HARMONICS WITH LINEAR LOADS

TABLE IX
ALLOWABLE PERCENTAGE OF DISTRIBUTED GENERATION
SEQUENCE HARMONICS WITH LINEAR LOADS

FOR

ZERO

which do not take linear loads into account, yield conservative

results.
To predict the allowable distributed generation (including
were obtained through
linear load models), new values for
computer simulation of the given distribution circuits including
linear load models. The ratios of the substation currents with
and without linear loads are calculated and multiplied with
the percentage numbers obtained in Tables VI and VII. The
predictions obtained with linear loads using this approach
are shown in Tables VIII and IX. It is to be noted that there
is no reason to increase the allowable distributed generation
penetrations further for levels already equal to 100%.
It is appropriate to compare the results obtained without and
with linear load models (see Tables VIIX). The differences obtained with the two approachs are within 510%, which is well
within accuracy expectations for planning-level studies. For this

624

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 2, APRIL 2003

reason, the additional effort to include linear loads does not appear to be justifiable. Furthermore, neglecting linear loads leads
to conservative results.
As a concluding comment regarding the results in Tables VIIX, consider the situation where the distributed
generation does not produce the most limiting harmonics
(seventh or ninth). Because the 519-1992 limits are assumed to
(as
be in force, similar analyses to determine the allowable
shown in Tables IV and V) for other harmonics would lead to
greater permissible currents. Tables VIIX, therefore, provide
worst case estimates of allowable penetration percentages.
VI. CONCLUSION
Closed-form equations for determining allowable penetration
levels of distributed generation resources have been presented in
this paper. The methods are suitable for many typical radial distribution feeders with uniform, linearly increasing, or linearly
decreasing load patterns. The equations have been developed
according to acceptable harmonic analysis assumptions wherever possible; atypical assumptions have been carefully shown
to be either technically or economically justifiable.
With distributed generation fast becoming a reality, the
methods proposed will help provide guidelines to estimate
the allowable amount of distributed resources in typical
distribution circuits. These results can help shape numerous
economic decisions including capacity credits and rate structures. Furthermore, the intentional use of only readily-available
distribution system data allows the methods presented here to
be applied to other (more specific) cases in the planning stages
so that potential problems can be avoided.
It is also important to recognize that the results presented
here, while focused on distributed generation, also apply directly to the determination of allowable penetration levels of
other harmonic-producing equipment. Of course, the conclusions drawn would be different because the harmonic current
content would not be limited to the first row of the 519-1992
current table.
REFERENCES
[1] Recommended Practices and Requirements for Harmonic Control in
Electric Power Systems, IEEE Standard 519-1992, Apr. 1993.
[2] A. Mansoor, W. M. Grady, P. T. Staats, R. S. Thallam, M. T. Doyle, and
Samotyj, Predicting the net harmonic currents from large numbers of
distributed single-phase computer loads, IEEE Trans. Power Delivery,
vol. 10, pp. 20012006, Oct. 1995.
[3] A. Mansoor, W. M. Grady, R. S. Thallam, M. T. Doyle, S. D. Krein,
and M. J. Samotyj, Effect of supply voltage harmonics on the input
current of single-phase diode bridge rectifier loads, IEEE Trans. Power
Delivery, vol. 10, pp. 14161422, July 1995.

[4] R. Dwyer, A. K. Khan, M. McGranaghan, L. Tang, R. K. McCluskey,

R. Sung, and T. Houy, Evaluation of harmonic impacts from compact
fluorescent lights on distribution systems, IEEE Trans. Power Systems,
vol. 10, pp. 17721780, Nov. 1995.
[5] A. E. Emanuel, J. Janczak, D. G. Pillegi, E. M. Gulachenski, M. Breen,
T. J. Gentile, and D. Sorensen, Distribution feeders with nonlinear loads
in the northeast USA: Part IVoltage distortion forecast, IEEE Trans.
Power Delivery, vol. 10, pp. 340347, Jan. 1995.
[6] A. E. Emanuel, J. Janczak, Pillegi, and T. J. Gentile, Voltage distortion in distributed feeders with nonlinear loads, IEEE Trans. Power Delivery, vol. 9, pp. 7987, Jan. 1994.
[7] Electrical Transmission and Distribution Reference Book: Westinghouse Electric Corporation, 1950.
[8] T. Hiyama, M. S. A. A. Hammam, T. H. Ortmeyer, J. Janczak, Pillegi,
and T. J. Gentile, Distribution system modeling with distributed harmonic sources, IEEE Trans. Power Delivery, vol. 4, pp. 12971304,
Apr. 1989.
[9] Gonen and Turan, Electric Power Distribution System Engineering. New York: McGraw-Hill, 1986.

Arijit Bhowmik (M98) received the B.E.E. degree from Nagpur University,
India, in 1997, and the M.S. degree from Mississippi State University, Mississippi State, MS, in 2001.
Currently, he is an Engineer with Open Systems International, Inc., Minneapolis, MN. His research interests are in the areas of modeling and simulation
techniques for power systems, power system transients, and harmonics.

Arindam Maitra (M95) received the B.E.E. from Nagpur University, India, in
1995, and the M.S. and Ph.D. degrees from Mississippi State University, Mississippi State, MS, in 1997 and 2002, respectively.
Currently, he is a Senior Engineer at EPRI PEAC Corporation, Knoxville,
TN. At EPRI PEAC, he is responsible for conducting and managing research
activities associated with power quality. His specific areas of expertise are modeling and simulation techniques for power system harmonics, power system
transients, computer applications in power systems, and power system control
and protection.

S. Mark Halpin (M93SM02) received the B.E.E., M.S., and Ph.D. degrees
from Auburn University, Auburn University, AL, in 1988, 1989, and 1993, respectively.
Currently, he is a Professor in the Department of Electrical and Computer
Engineering at Auburn University. His teaching interests include power systems,
control systems, and network analysis. His research interests are in the areas
of modeling and simulation techniques for large-scale power systems, power
system transients, and computer algorithms.
Dr. Halpin is with the IEEE Power Engineering Society and serves as Chair
of the Task Force to revise IEEE 519-1992 , and is with the Industry Application
Society, serving as Chairman of the IAS Working Group on Harmonics.

Joe E. Schatz, photograph and biography not available at the time of publication.