You are on page 1of 6

# HARMONIC COMPENSATION OF

## COMMERCIAL AND INDUSTRIAL POWER SYSTEMS

S.J. Dodroe, Student Member D.T. Radmer, Student Member R. Montgomery, Student Member
J. Chen, Student Member J.S. Foertsch

## Department of Electrical Engineering

Gonzaga University
Spokane, WA 99258

Abstmcf;The harmonic currents drawn by many types of Transformers that compensate for the effects of nonlinear
nonlinear loads have been described in the literature. loads are now available [3].
However, there is still a need for a method for predicting The number and variety of nonlinear loads are different for
the effects of combinations of three-phase and single-phase virtually every electrical installation. Hence, reduction and
nonlinear loads. This paper presents a general method for compensation techniques for harmonic currents are specified
estimating the root-mean-square (rms) values of electric on a case-by-case basis. A method for predicting harmonic
currents and total harmonic distortion (THD) due to any currents would be helpful in retrofitting existing electrical
number of selected types of nonlinear loads on three-phase distribution systems, or designing new ones.
four-wire and single-phase threewire distribution systems. A general method for estimating the rms values and THD
Data was gathered from experimentationand was used to of electric currents in three-phase four-wire and single-phase
validate the proposed method. Results of simulations of three-wire distribution systems is presented in this paper.
various harmonic reduction and compensation techniques The proposed method was validated by actual experiments
are also presented. performed at the computer facility of the School of
Engineering at Gonzaga University. The method was used to
Key words: Harmonics,, overloaded neutral conductors, simulate various harmonic reduction techniques, and
harmonic reduction, harmonic compensation, power quality. numerical results are presented.

## I. INTRODUCTION 11. SOLUTION METHODOLOGY

Nonsinusoidal currents drawn by nonlinear loads were The currents drawn by a nonlinear device may be
recognized early in the analysis of electrical devices [l]. At characterized by a fundamental and harmonic components in
that time, the effects of nonlinear loads were seldom regarded the frequency domain. The frequency spectra of the currents
as a serious problem because of their relatively minor impact drawn by some nonlinear devices are available from the
on power distribution systems. However, present-day literature. The magnitude of a harmonic component is
nonlinear loads represent an ever-increasing and substantial usually expressed in per unit of the magnitude of the
portion of all electrical loads and have come under fundamental component. The phase angle of the harmonic
investigation [2-141. component is referred to that of the supply voltage.
The nonsinusoidal currents drawn by nonlinear devices have Each harmonic component of the current is represented by
been found to contribute to the overloading of neutral conduc- an ideal current source with a specified magnitude and phase
tors, and it has been suggested that harmonic currents be angle 141. All of these current sources are connected in
accounted for in conductor sizing [2]. Harmonic currents have parallel as shown in Fig. 1, and they can be combined into
also been found to reduce the service life of transformers. an equivalent current source.

+-
I I I
I
Fig. 1. Load represented as parallel current sources.

0-7803-1883-8\$04.00019941EEE
97
0 a
1-9
n
1 4
1-*
Lood
c
C r r I

## Fig. 3. Single-phase three-wire distribution system

Fig. 2. Three-phase four-wire distribution system. Having found all the harmonic components, the current
waveform is derived from the frequency data as follows:
A three-phase four-wire distribution system is shown in Fig. n
2. Any single-phase load is assumed to be connected between
a phase conductor and the neutral.
When two or more single-phase nonlinear loads are
where Zh is the rms value and 8, is the phase angle of the h"
c o ~ t ~ t to
e dphase a, the corresponding harmonic components
harmonic component. Equation (6) may be used to fmd the
waveforms of the phase currents, as well as the neutral
follows:
current. The rms value and the THD of the current are
calculated as follows:
i=l
I,, =
where Z,,,i is the nns value of the h* harmonic component from k=l (7)
load i , and N, represents all devices connected to phase a.
When nonlinear loads are connected to phase b, or phase c,
the corresponding harmonic components of the current from
respectively [5]:

## where Z, is the rms value of the fundamental component.

Equations (7) and (8) may be used to estimate the rms value
and the THD of the phase currents and the neutral current.
IcA = IkJI e,, + 120"h (3)
Harmonic reduction and compensation techniques may be
I=1
employed when the rms currents or THD are considered
Nb and N, represent all devices connected to phases b and c, high. These techniques may be implemented by multiplying
respectively. Equations (1)-(3) may be used to calculate all the transfer function of the technique by the Fourier spectral
harmonic components, h, of interest. The harmonic component density of the current for each phase. The result may be
of the neutral current is found by adding the corresponding expressed as a Fourier series, and (7) and (8) are solved to
harmonic components in the three phases as follows: obtain the desired outputs.
A measure of the distortion of the current seen by a
distribution transformer is given by the k-factor [3]. This
(4)
parameter is used in the design and manufacture of
Two single-phase loads are shown in the single-phase three- transformers. The k-factor is calculated as follows:
wire distribution system of Fig. 3. For a load Connected to
H
phase b, the harmonic components of the current are shifted
k-factor = h2 (9)
180" with respect to the components of the current, had the h=l
load been connected to phase a. The corresponding harmonic
components for phases a and b are added vectorially to find where I,, is expressed in per unit with respect to the total
the harmonic component of the neutral current as follows: true-rms current. The values of k-factors range from 1 to
4.h = Iah + SJI 50. A k-factor of 1 represents a purely sinusoidal load
current. Typical k-factors for harmonically compensated
This is repeated for all values, h, of interest. transformers are 4, 9, 13, 20, and 30.

98
111. DATA ACQUISITION
DRANETZ
@
120/208V Feeder
Experiments were performed to collect data that was used to
validate the proposed method. A DmetzRModel 658 Power Single-phase
computer and
Quality Analyzer was used to measure currents for fluorescent printer loads
light fixtures, and several types of personal computers and
pMters. Measurements were taken at the locations shown in
Fig. 4.
The lighting load consisted of fifteen fluorescent light
fixtures, each with a two-lamp ballast and a one-lamp ballast. Single-phase
Twelve of the two-lamp ballasts were connected to phase a. lighting loads
Phase b supplied twelve of the one-lamp ballasts, and phase c
supplied the three remaining fixtures.
The computer laboratory equipment is comprised of four
types of computers and four types of printers. Nine computers a b c n

of various types and two printers are on phase a. Phase b Fig. 4. Location of current measurements.
serves six computers and thee printers, and there are seven
computers and three printers on phase c. same for each device. Randomness of magnitudes and phase
The DranetzR analyzer measures currents and voltages, and angles of harmonic components are neglected. Harmonic
stores the data which can be later retrieved as either time- components above the twentieth are considered to be
domain waveforms or frequencydomain spectra. Frequency- negligible, and are not entered into the device library.
domain data consists of the magnitudes of harmonic For simplicity, the source voltage is assumed to be sinu-
components as a percentage of the fundamental, and their soidal, and all devices are assumed to have been supplied by
phase angles referred to the: supply voltage. a source of the same voltage and intemal impedance.
All the loads were turned on for some time to allow them to The program calculates the harmonic components of the
reach steady state prior to the measurements. Phase and currents in the phase and neutral conductors by using (1)-(5).
neutral currents were measured and recorded by the DranetzR The current waveform is found by using (6). Equations (7)
analyzer with all the loads connected, and this procedure was and (8) are used to compute the rms values of the current for
repeated as a device or group of devices was switched off. all phases and the neutral, and the percent THD for the
The nns values of the currents for the various combinations of phase conductors, and (9) is used to compute the k-factors.
loads were also measured using a Fluke' Model 87-A true-rms To determine the validity of the proposed method, various
ammeter. Seventeen different experiments were conducted loading conditions were simulated and the results were
with various numbers and types of single-phase devices on the compared with those gathered from the experiments. Figure
system. 5 shows the waveforms of the computed and actual currents
flowing in the feeder neutral with all the loads connected in
IV. ALGORITHM AND PROGRAM DEVELOPMENT the test system of Fig. 4. Experimental and computed
currents for the three-phase test system were shown to be in
A spreadsheet program was developed using Microsoft close agreement. The phase and neutral currents predicted by
ExcelR. To facilitate program use, characteristic data for the proposed method exhibited a mean percent difference of
various nonlinear devices was stored in a library. The data 4.52 and a standard deviation of 3.91 when compared with
consists of harmonic rms magnitudes and phase angles for the experimental data. The predicted currents were found to be
currents drawn by each device. The devices in the library on the high side compared to those measured in the experi-
include computers, printtas, and discharge lighting. The ment. This is due to cancellations of harmonics arising from
characteristic data for each device was acquired from the randomness of the phase angles and magnitudes of some
experiments and from the literature. components. Cancellations were observed for all even har-
The program is set up in an ExcelR notebook format. Data monics, and odd-order harmonics greater than the thirteenth.
may be retrieved from the library, or additional data may be The program may be used to simulate a variety of
entered into the library for newly characterized devices. Both harmonic compensation and reduction techniques. Some
three-phase and single-phase distribution systems with any assumptions are made in simulating them. The input
number and type of single-phase devices may be simulated. impedance of the active filter is taken to be equal to its
Combinations of three-phase and single-phase loads may also output impedance. The delta-wye transformer is considered
be simulated on the three-phase system. ideal. When simulating the reduction techniques, the
When multiple devices of the same type are modelled on a distribution system is considered to be isolated from that of
distribution system, the characteristic data is assumed to be the the utility.

99
harmonics may also provide a current path for the same
40 harmonics to the utility's power grid [7]. To remedy this,
I rms (measured) = 16.76A
30 I rms (simulated) = 17.82A series line reactors can be used to increase the impedance
20 between the filter and the power supply.
h
The rectification process performed by three-phase
a 10
Y variable-speed controllers for motors results in harmonic-
l o laden line currents. Tuned filters with series line reactors
-10 may be used to attenuate higher-order harmonic currents.
U
-20 Resonance is also a potential problem that must be
-30
considered. Shunt capacitors used for power-factor
correction are potential sources of resonance due to
-40 harmonic currents, and capacitors should be sized carefully
0.00 2.08 4.17 6.26 8.33 10.42 12.50 14.68 16.67 to avoid this problem [4].
Time(ms) Electrical distribution systems which contain large
fluorescent lighting loads may be supplied by transformers
Fig. 5. Comparison between computed and measured values with secondary and tertiary windings which are both wye
of neutral current. connected. The transformers are connected in antiphase, as
shown in Fig. 6., with a shared neutral, and with loads
The program has been designed to be interactive. The user divided equally between the secondary and tertiary windings.
can simulate various combinations of loads from the input The neutral current of the loads being supplied by the
page, and can later return to simulate reduction techniques. secondary are 180" out of phase with the neutral currents of
Any part or all of the results may be printed, and used to the loads supplied by the tertiary. Thus, cancellation of
compare the effectiveness of various harmonic reduction neutral currents occurs at the common neutral bus.
techniques. Simulations were made to show the effects of some of
these harmonic reduction and compensation techniques by
V. COMPENSATION AND REDUCTION TECHNIQUES using the computer program described in Section IV. Models
of filters and a transformer were created within the program.
Common methods employed to compensate for harmonic These include a third-order Butterworth and a second-onler
Currents include increased size of neutral conductors, and the active low-pass filters; a tuned, shunt LC filter; and a delta-
use of k-factor, derated, and delta-wye transformers. wye transformer. The reduction techniques were simulated
There are many methods to reduce the harmonic content of on the test system shown in Fig. 4. The numerical results of
nonsinusoidal currents. Active and passive filters may be these simulations are presented in Table I.
implemented to attenuate selected harmonics. A technique to The results show varying degrees of reduction of the
cancel harmonic currents in neutral conductors employs percent THD and rms values for the neutral and phase
tertiary-winding transformers connected in antiphase [6].The currents. For the delta-wye transformer, the turns ratio is
replacement of devices to those with lower THD values is as one-to-one, and the primary line currents are shown in the
well a valid reduction technique. table. All the reduction methods lowered both percent THD
K-factor transformers both reduce and compensate for the and rms values of the current. The low-pass filters provided
heating due to harmonic currents. Methods by which this is comparable harmonic reduction. They reduced the neutral
done include constructing windings that are built as parallel current to between 39 and 57 percent of the base case.

E
conductors rather than a single conductor to lower skin-effect
losses, and increasing core size to lower flux density and core
Secondory
saturation levels. Using derated (or oversized) transformers
will lower flux density and winding resistance. Core saturation

r*
may be eliminated, but possibly at the expense of higher load I, 1 b2
c2
currents [3]. Delta-wye transformers are often mistakenly
viewed as triplen harmonic traps. This will only be the case if O 1 q m

## a particular triplen harmonic carried by each phase conductor c1

Primary
is equal in magnitude and in phase. Otherwise, the imbalance
of each respective triplen harmonic will be carried by the line
Tertiary
conductors.
Low-pass and tuned filters that are used to shunt harmonic c3

currents must be implemented with caution. Tuned filters that Fig. 6. Tertiary-winding transformer connected in antiphase.
are used as low-impedance paths to ground for specific

100
TABLE I insights proved invaluable, and his ability to keep the team
Numerical Results of Harmonic Reduction Simulations focused was remarkable. The generosity of Washington
Water Power in allowing the use of the Dranetz' Power
Quality Analyzer is greatly appreciated. A special thanks is
given to Mr. Greg Smith, Power Quality Engineer at
Washington Water Power, for his assistance and instructions
on the use of the DranetzP analyzer.

VIII. REFERENCES

## A.F. Puchstein, T.C. Lloyd, and A.G. Conrad,

Alternating-Current Machines, New York: John Wiley
& Sons; London:Chapman and Hall, Ltd, 1954, pp.
Legend: 0 = Base Case, no reduction method 113-116, 305.
1 = 3"-order passive filter A. Freund, "Double the Neutral and Derate the
2 = P-order active filter
3 = Tuned LC filter Transformer-or Else!, " Electrical Construction and
4 = Delta-Wye transformer Maintenance, December, 1988, pp. 81-85.
R.D. Arthur, "Testing Reveals Surprising k-Factor
Diversity, "Electrical Construcdion and Maintenance,
VI. CONCLUSIONS April, 1993, pp. 51-55.
IEEE Standard 519-1992, "IEEE Recommended
A general method for estimating the rms values and THD of Practices and Requirements for Harmonic Control in
electric currents in three-phase and single-phase distribution Electrical Power Systems. 'I
systems has been described in this paper. It may be used to G.T. Heydt, W.M. Grady, and D. Xia, Harmonic
predict values of currents in branch circuits or feeders to Power Flow Studies, Vol. 1: Theoretical Bmis,
determine whether excessive harmonic currents are present at Prepared by Purdue University for Electric Power
any point in the system. ' n e proposed method will show Research Institute, Research Project 1764-7, Final
whether any circuits are lightly loaded and able to safely Report, November 1983, pp. 125-130.
supply more loads. It provides an alternative to a A. Liew, "Excessive Neutral Currents in Three Phase
comprehensive and expensive testing and characterization of Fluorescent Lighting Circuits, " ZEEE Transactions on
an electrical distribution system. Industry Applications, vol. 25, no. 4, JulyIAugust,
The proposed method has accurately predicted the rms 1989, pp. 776-782.
values of the phase and neutral currents, which exhibited a M.Z. Lowenstein, "Improving Power Factor in the
mean percent difference of less than 5 R when compared to the Presence of Harmonics Using Low-Voltage Tuned
experimental values of the three-phase system. Numerically Filters, " IEEE Transactions on Industry Applications,
derived and experimental values of rms currents were within vol. 29, no. 3, MayIJune, 1993, pp. 528-535.
ten percent difference for add harmonics for the single-phase G.T. Heydt, Electric Power Quality, West LaFayette:
system. The even harmonics from the experimental data were Stars in a Circle Publications, 1991, pp. 339-399.
deemed negligible in magnitude and were simply rejected for N. Mohan, T.M. Undeland, and W.P. Robbins, Power
the single-phase system. Electronics: Converters, Applications, and Design,
Filters and delta-wye transformers are effective in improving New York: John Wiley & Sons, 1989.
power quality, as shown. by the computer simulations. T.M. Gruzs, "A Survey of Neutral Currents in Three-
However, it is recommended that the implementation of these Phase Computer Power Systems, " IEEE Transactions
techniques be preceded by a detailed analysis of the electrical on Industry Applications, vol. 26, no. 4, July/August,
system. 1990, pp. 719-725.
Finally, several techniques are available for improving H.O. Aintablian and H.W. Hill, Jr., "Effects of
power quality, but the resulting benefits should be weighed Personal Computer Harmonics on the Distribution
against cost of implementation, possible decrease in overall transformer," in Proceedings of the I992 24th Annual
system efficiency, and other potential drawbacks. Meeting of the IEEE North American Power
Symposium, pp. 196-204
VII. ACKNOWLEDGEMENTS L.J, Bohmann and C.T. Plante, "A Harmonic Survey
of Switched-Mode Power Supply Loads and Their
The authors extend their thanks and appreciation to their Potential Problems in Commercial Buildings, "
faculty advisor, Dr. Juan L. Bala, Jr. His guidance and

101
Proceedings of the 1992 24th Annual Meeting of the and power electronics. He intends to pursue graduate studies
B E E North American Power Symposium, pp. 180-188. leading to a doctoral degree in electrical engineering with
[13] National Electrical Code, NFPA 70-1993. National Fire specialization in power engineering.
Protection Association, Quincy, MA.
[14] R.C. Lentz, F.J. Mercede, and J.N. Mercede, Jr., "A Robert Montgomery was born on February 15, 1961 in
Student Design Project to Improve Power Quality for a Washington, D.C. He served eight years in the United States
Commercial Facility," Paper 94 W M 237-8 PWRS, Air Force as an Autopilot and Instrument Systems
Presented at IEEEIPES Winter Meeting 1994, New Technician. He is pursuing his BSEE at Gonzaga University.
York, New York, January 30 - February 3, 1994. His interests include power distribution, power system
analysis, and controls.
BIOGRAPHIES
Jimmv Chen was born on September 17, 1968 in Taipei,
Steven J. Dodroe was born on May 29, 1959 in Portland, Taiwan. He worked at Logue McDonald Automation,
Oregon. He received his BSChE from the University of Idaho Spokane, Wa, as an assistant design engineer. He is studying
in 1984. He has been an electrician for eight years. He is for his BSEE at Gonzaga University. His interests include
currently a graduate student in electrical engineering at power engineering, control systems, and distribution and
Gonzaga University, and teaches the Electrical Power transmission systems.
Engineering Laboratory. His interests include electrical power
engineering, power system analysis and design, transmission John S. Foertsch was born on September 3, 1972 in Fargo,
and distribution systems, and optimization techniques. North Dakota. He is currently pursuing his BSEE at
Gonzaga University, and his interests include computer
Duane T. Radmer was bom on September 27, 1971 in Las engineering and control system design. He currently works
Vegas, Nevada. He is studying for his BSEE at Gonzaga for System Technology, Spokane, WA, performing
University. His interests include power systems engineering computer-aided control systems design.

102