Survey of Harmonics Measurements in Electrical Distribution Systems

Leon M. Tolbert, Member, IEEE Harold D. Hollis and Peyton S. Hale, Jr.
Oak Ridge National Laboratory* U.S. Army Center for Public Works, CECPW-K
P.O. Box 2008,131dg 1000 7701 Telegraph Drive
Oak Ridge, TN 37831-6334 Alexandria, VA 223 15-3862
Phone: (423) 576-8175 Phone: (703) 428-8191
FAX: (423) 576-8189 FAX: (703) 428-8297
e-mail: [email protected] e-mail: [email protected]

Abstruct- An assessment of the harmonics found in the from any of three causes: 1. nonsinusoidal generation of
electrical distribution system (of several buildings in the three voltage, 2. nonlinear devices used in the transportation of
Department of Energy (DOE) Oak Ridge plants in Tennessee electrical energy, 3. nonlinear load devices. The voltage
has yielded several conclusions useful in the design of electrical waveforms generated by utilities and then stepped up to a
distribution systems. A preliminary survey to determine where
transmission voltage level generally are very close to
significant amounts of harmonic currents or voltages existed in
the distribution system was performed at several buildings by sinusoidal and have little distortion. Transmission lines,
comparing readings taken with true root-mean-square (RMS) cables, and transformers are transportation devices which are
multimeters and averaging multimeters (non-true RMS meters). quite linear in nature and cause little distortion to voltage or
From these measurements and subsequent calculations, current waveforms. However, variable frequency drives and
facilities that appeared to have significant levels of harmonics unintemptible power supplies which use electronic devices to
present were analyzed with a power analyzer that could record rectify ac to dc and then invert back to ac are nonlinear
waveforms and give the spectrum and magnitude of harmonics. transportation devices. Several loads are nonlinear such as
Among the non-linear loads for which sample waveforms are switch mode power supplies [3,4] and fluorescent light
illustrated are fluorescent lighting (both magnetic and
ballasts [5, 61.
electronic ballasts), variable frequency drives, switch mode
power supplies, and uninterruptible power supplies. A
discussion as to how various harmonic waveforms come about as 11. BACKGROUND
a result of these nonlinear loads is also outlined in this paper.
Through Fourier analysis, any nonsinusoidal periodic
waveform can be represented as the sum of a dc component
I. INTRODUCTION and sine waves of various amplitudes and phase displacement
from some relative angle. The sine waves all have frequencies
A harmonic survey was done for all of the 120/208 V and which are an integer multiple of some fundamental
120/240 V branch circuit panels in 31 buildings at the three frequency, which in the United States is 60 Hz. Any periodic
Department of Energy (DOE) Oak Ridge plants. The voltage waveform V(g can be expressed as the sum of a dc
buildings ranged in age from 5 to 50 years old. Some of the component VOand sine waves with a fundamental frequency
building uses include ofice space, computer and control w (60 Hz):
centers, laboratory instrumentation, robotic manipulators, X- N
ray equipment, furnaces, and maintenance shops. This paper V ( t ) = V,, + c V hsin(hwt + eh) (1)
h= I
outlines a quick method of identifying harmonic currents in
an electrical distribution system and assessing where they A similar equation could be written for a current
originate and to what extent the harmonic currents should be waveform l(s). Because these individual sine waves are all
suppressed. integer multiples of a fundamental frequency, they are called
Because most electrical systems were designed for linear harmonics. The harmonic order is defined as the integer
voltage and current waveforms (i.e. nearly sinusoidal), multiple of the fundamental. For example, a 180 Hz harmonic
excessive non-linear loads can cause serious problems such as frequency wave on a 60 Hz source would be called the 3rd
overheating conductors or transformers, capacitor failures, harmonic. Equation (2) shows how to find the root-mean
inadvertent circuit breaker tripping, or malfunction of square (RMS) value of a current waveform where the R M S
electronic equipment [ 1,2]. Nonlinear currents can originate value of each of the harmonics, I,, is known.

* Prepared by the Oak Ridge National Laboratory, Oak Ridge, Tennessee, 37831-6334, managed by Lockheed Martin Energy Research, Inc. for the U.S.
Department of Energy under contract DE-AC05-960R22464. The submitted manuscript has been authored by a contractor of the U.S. Government.
Accordingly, the U.S. Government rehins a nonexclusive, royalty-free license to publish or reproduce the published from ofthis contribution, or allow others to do
so, for U.S. Government purposes.

0-7803-3544-9196 $5.00 0 1996 IEEE 2333

 ~

(2) Waveform I Description

I

One measure of the nonlinearity, or distortion, in a

waveform is given by (3) and called total harmonic distortion
(THD). THD is the ratio of the RMS value of the total
harmonic currents (nonfundamental part of the waveform)
and the RMS value of the fundamental portion, 11, of the
waveform. This value is usually expressed as a percentage of
the fundamental current.

THD =
[8 (Ih IZ
(3)
Sine wave with
exponentially

w
I, damued imuulse
Two other measures of distortion are the crest factor and
Pulsed Current
the form factor. The crest factor is the ratio of the peak of a
waveform to its RMS value. For a linear sinusoidal waveform,
the crest factor would be the square root of 2, or 1.414. I I3-phase neutral

Crest- Factor = -
IPd
(4)
Jvvvvv\ harmonic
current

I
IRMS Pulsed current
The form factor, or distortion factor, is the ratio of the RMS with linear
value of a waveform to the RMS of the waveform's lload
fundamental value, I,. For a linear sinusoidal waveform, the
form factor would be 1.0 [7].
Fig. 1. Comparison of conversion techniques for AC current measurements. All
r e a d h p are normalized to read 100 A on an average, converting RMS-
Form- Factor = (5) calibrated meter - the most common type of meter.
1,
Symmetrical components is a mathematical tool used to
analyze power systems. Harmonic orders can be divided into Two measurements of voltage, true-RMS and peak (or
positive, negative, and zero sequence components [SI. amplitude), and three measurements of current, true RMS,
Positive sequence components are the given phase order and average, and peak with a true-RMS meter, were taken at
include the following harmonic orders: 1st (fundamental), several panelboards in the DOE facilities. The averaging
4th, 7th, loth, etc. Negative sequence components are reverse meters used considered the fundamental portion of the
phase order and include the following harmonic orders: 2nd, waveform only and ignored harmonics. For nonlinear loads
5th, 8th, llth, etc. Zero sequence components have all three with substantial harmonic distortion present, the difference in
components in phase and include the following harmonic the measurements between the averaging and true-RMS
orders: 3rd, 6th, 9th, 12th, etc. Phase and amplitude balanced meters was notable, and the form factor computed from (5)
positive and negative sequence components sum to zero in the was on the order of 1.2 to 1.5 instead of 1.0 as is the case for
neutral or ground. Balanced zero sequence components, linear loads. Additionally, the crest factor computed from (4),
however, add in the neutral or ground. Because the zero which used the true-RMS and peak measurements from the
sequence harmonics are divisible by 3, they are referred to as RMS-calibrated meter, was on the order of 2 or 3 for
triplens. waveforms with s i m c a n t harmonic content instead of close
to 1.4 as would be the case for a sinusoidal wave.
111. RMS vs. NON-RMSMEASUREMENTS
B. Crest and Distortion Factors
A. Different Meter Types
From the measurements made on these panels, the
Fig. 1 is a chart comparing the measured values for various distortion factor and crest factor were calculated for each of
waveforms using peak, averaging, and true-RMS calibrated the phase and neutral voltages and currents. Of the 200
meters [9]. For the preliminary harmonic survey to determine panels surveyed, 72 had phase current crest factors greater
where significant amounts of harmonic current existed in the than 2.0 or current distortion factors greater than 1.5.
120/208 V and 120/240 V electrical distribution systems at The distribution of voltage crest factors is shown in Fig. 2.
the three Department of Energy (DOE) Oak Ridge plants, Seventy-five percent of the voltage crest factors were between
measurements were taken for branch circuit panels in 31 1.36 and 1.39, with an average crest factor of 1.376 for the
buildings with true-RMS calibrated multimeters and 506 different calculations. In fact, only 3% of the voltage
averaging (non-true RMS) calibrated multimeters [101, crest factors did not fall within the range of 1.34 to 1.43. The

2334
 1.48 1.42-1.49

1.46 1.41
1.44
1.#
1.42
9
0 b 1.39
1.40 Y
m
0
.
U

1.38 LL 1.38
F
0
a, 1.36 s!
o 1.37
U)
(II
= 1.34
8 8 1.36
1.32 8
1.35
1.30
1.28 1.34

1.26 1.26-1.33
0 50 100 150
0.00 1.oo 2.00 3.00 4.00
Number of Occurrences
Fig. 5. Average currentdistortion and crest factor for each voltage crest factor.
Fig. 2. Mathematical distribution of phase voltage crest factors.
the voltage at the panel to have more of a flat-topped voltage
b waveform.
3.00-4.00
U
H The distribution of current distortion factors and current
K 2.00-2.99 crest factors are shown in Figs. 3 and 4, respectively. The
.-r0 current distortion factor was minimal (less than 1.1) for more
0
44 1 50-1.99
U) than 54% of the 459 current measurements made at the power
E panels. However, 15% of the conductors checked had current
1.10-1.49
distortion factors which exceeded 1.50. Current crest factors
1 00-1.09 250 exceeded 2.00 for 36% of the conductors surveyed, and 15
instances were found where the crest factor exceeded 5.00.
0 50 100 150 200 250
Number of Occurrences To help illlustrate the interrelation between the voltage
distortion andl the current distortion, a bar chart is shown in
Fig. 3. Mathematical distribution of phase current distortion factors. Fig. 5 whiclh plots the average values for the current
distortion factor and current crest factor for each of the
voltage crest factors calculated, From Fig. 5 , one can see that
Above 5.00
the average current distortion factor and voltage crest factor
8 4.00-4.99 really have no interrelationship. However, the average current
Y
0 crest factor is generally higher for the lower voltage crest
2
44
3.00-3.99
factor. A high current crest factor load usually draws a large
U)
current pulse near the peak of the voltage waveform. The
-9
0
2.00-2.99

1.60-1.99
subsequent voltage drop causes a flat-topping of the voltage
5
waveform, hence a lower voltage crest factor.
0 1.30-1.59
C. Power Analyzer Measurements
1.oo-1.30

0 50 100 150 200 250 Several panels in the facilities that had high crest or
distortion factors and had a load of more than 30 A were
Number (of Occurrences
chosen to have measurements made with an 8 channel power
Fig. 4. Mathematical distribution of phase current crest factors. analyzer that takes simultaneous snapshots of the voltage and
current wavefforms for all three phases and neutral of a 3-
minimum voltage crest factor and the maximum voltage crest phase system. The analyzer also could compute the spectrum
factor calculated were 1.26 and 1.49, respectively. For panels and magnitude of harmonics for these various waveforms.
that had current crest factors greater than 2.0, the voltage Using multinneters to screen candidate panels first saved
crest factor averaged 1.371, and for the remaining panels the several days of work because they are much more easily and
crest factor averaged 1.380. These calculations show that quickly connected and disconnected from a power panel’s
panels with high amounts of current harmonics tend to cause main conductors. In addition, the power analyzer required an

2,335
 ~

external source of 120 V ac power which sometimes was not A. Fluorescent Lighting
readily available.
The voltage THD for all of the panels surveyed with the Measurements were taken at an office building at the Oak
power analyzer was between 2.0% and 4.5%. This meets the Ridge National Laboratory to get representative data samples
5.0% voltage THD limit for end use loads recommended by for traditional magnetic fluorescent ballasts and T-12 lamps.
E E E 5 19 [ 111. Most of the distortion was due to 3rd and 5th These ballasts were then removed and replaced with new
harmonics. The voltage distortion was greater at the panels electronic ballasts and T-8 lamps and new measurements then
that had the greatest current distortions (greater than 50% taken.
THD). This data lends credence to the fact that current Fig. 6 depicts the phase voltage and current and the neutral
harmonics do effect voltage THD. The voltage produced by current for a lighting panel which had a load consisting
current harmonics can be characterized as entirely of fluorescent lights with magnetic ballasts. The
N harmonics in the current waveform are caused by the
v = h=2 Ih(Rh -k jhw Lh) . (6) nonlinearity of the lamp arc itself in series with the ballast
However, this effect is usually slight when compared to [5]. The THD for the phase currents was 13.9% and for the
electrical equipment which generate or synthesize voltage neutral currents was 141.3%. The spectrum consisted of only
waveforms. odd harmonics and ranged from the 3rd to the 17th. In
addition, the amplitude of the individual harmonics for the
Iv. IDENTIFICATION OF HARMONICS
9th through the 17th were all less than 1% of the amplitude
of the fundamental (60 Hz).
From the more than 1000 sample waveforms collected Fig. 7 depicts the phase voltage and current and the neutral
throughout the buildings, several different sources of current for the same lighting panel after all of the fluorescent
harmonic currents and voltages were identified in the DOE lights were converted to electronic ballasts. With the new
complex. A sample of the different waveforms found during ballasts, the panel had a phase current THD of 17.2% and
the survey along with an identification of the source of the 85.3% THD for the neutral current. Unlike the magnetic
harmonic current or voltage follows. References [ l 1-13]
I
ballasts, the electronic ballasts had individual harmonics at
illustrate waveforms for some additional non-linear loads. frequencies up to the 33rd harmonic (1980 Hz) which had

200, I 60 200 , , 40

2
a,
100 5
cn
m
100
cn
m
r
-
c
0

8
c
0 2
3 0
a,

z
a,
S

.c
a
-100
5
:
LL
-100

-200 -200 ' ' -40

I I 40

I L -60 I -40
@) @)

Fig. 6. Waveforms for magnetic ballasts. (a) Phase voltage and current. Fig. 7. Waveforms for electronic ballasts. (a) Phase voltage and current.
@) Neutral current for a three phase, four wire system. @) Neutral current for a three phase, four wire system.

2336
 TABLE I.
HARMONIC
SPECnrUM FOR FLUORESCENT
LIGHT BALLASIX
I Magnetic Ballast I Electronic Ballast
I
a,
400
w
m
-
c

3 0
2
3
-400
0.8 12.7 20.1 4
111 lth I 0.2 2.3 I 3.5 i 2.0 I -800 L L -10
13th 0.3
15th 0.1 3.4
17th
19th
21st
0.1
0.0
0.0
0.0
0.7
0.5
2.2
2.0 9.6
800
-__ - I 10

23rd
25th
0.0
0.0
0.0
0.4
1.7
1.9
1.5
3.7
c
m
400 5 3
D
r 0
1127th I 0.0 I 0.0 I 1.7 I 8.2 11 3 0 O f
a
2
17
z
h

-400 -5 2
a
-800 -10

@)

Fig. 8. Variable fieqpency drive (a) line and @) load side waveforms.

zero during each half cycle, because this particular VFD does
not have any line inductors.
magnitudes of between 1.4 and 3 percent of the fundamental. This particular VFD used a pulse width modulation (PWM)
Table I is a compilation of the harmonic components of switching scheme in order to invert the voltage from its dc
current and voltage for the magnetic and electronic ballasts bus into a sinusoidal output. The PWM output of the VFD
surveyed. The voltage THD for the magnetic ballasts was a voltage ?waveformthat was quite noisy. The power
averaged 2.67 % THD with almost all of the distortion due to analyzer captured this phenomenon for the voltage waveform
a fifth harmonic component. With the electronic ballasts, the only because the VFD surveyed had a sufficiently low
voltage THD at the lighting panel averaged 3.70% THD with switching frequency (only about 500 Hz).State of the art low
most of the distortion due to a fifth harmonic component. The power VFDs have switching frequencies between 10 and 20
increase in voltage distortion is directly correlated to the kHz, and most power analyzers would not capture the
increase in fifth harmonic current distortion. The fifth switching noise: from these VFDs because they sample only
harmonic content of the current increased by 73.4 % when up to between 2 and 3 kHz.
the magnetic ballasts were replaced by electronic ballasts.
C. Switch Mode Power Supplies
B. Variable Frequency Drives
A switch mode power supply has a large capacitor in it
Sample waveforms for the volltage and current on the line (Fig. 9) which maintains a near-constant voltage for the DC
and load side of a variable frequency drive (VFD) at a bus in the power supply. The capacitor is charged (and draws
building in the Y-12 plant appear in Fig. 8. Note in Fig. 8a, current) only dluring the peak of the voltage waveform as
that the VFD draws two current pulses for each half cycle of shown in Fig. 10; hence the large current pulse at the peak of
the voltage waveform. This is typical of most VFD’s which the voltage waveform [ 101. In Fig. loa, the current waveform
employ six pulse conversion to commutate the three-phase for a power panel at ORNL which fed receptacles in offices
line voltage to a DC voltage [7]. Six pulse converters have that had large amounts of equipment that used switch mode
two switches per phase; this accounts for the two current power supplies i s depicted.
pulses during each half cycle of the voltage waveform as In three-phase systems that have a neutral conductor and a
shown in Fig. Sa. The current is discontinuous, i.e. goes to large number of single phase switch mode power supply

2337
 ~

800 , r 40

3
AC Input
LS

C 1
T-
DC-to-DC
Converter Load
2400
a)
m
WJ
-
c
20 ;
0
0

u I9
Fig. 9. Ordinary switch mode power supply circuit schematic.
8
E
i
z
0

-400
0

-20 2
5
5
h

-800 I I -40
200 , , loo 30%
3 25%
z loo E
c

20%
%
m
0
-g 15%
" 0 0 E7 S
8 5 2 10%
0 a
=!. h 5%
a -100 -50 2 0%
1 3 5 7 9 11 13 15 17 19
-200 -100 Current Harmonic Order

(4

80

40
0
0
O f
a
a
-40'i;

I ' -100 -200

I
I
\T
V
\T
\I
I
I -80

@)
- 50%
Fig. 10. Switch mode power supply waveforms. (a) Voltage and current for one 40%
a)
phase of a three phase 120/208 V system. @) Neutral current.
5 30%

loads, the neutral conductor will carry a large percentage of

; 20%
s 10%
each of the three phase conductors' current even under
0%
balanced load conditions. This is because switch mode power
supplies draw a current which has a significant amount of 1 3 5 7 9 11 13 15 17 19
triplen harmonics (3rd, 9th, 15th, etc.) Triplen harmonics are Current Harmonic Order
zero sequence harmonics which add in the neutral instead of
canceling like positive sequence (lst, 7th, 13th, etc.) and @)
negative sequence (5th, llth, 17th, etc.). The sample
Fig. 11. Uninterruptible power supply (a) line and (b) load waveforms and
waveform shown in Fig. lob indicates that the neutral current current harmonic spectrum.
is mostly 180 Hz (3rd harmonic).

D. Uninterruptible Power Supplies Although a U P S provides protection from loss of a voltage

source, many on-line U P S ' S demand current that has
Sample waveforms for the voltage and current on the line considerable harmonic content and/or supply a voltage to the
and load side of an uninterruptible power supply (UPS) load that has distortion.
located in a building at the Y-12 plant appear in Fig. 11. The UPS for which the waveforms are shown in Fig. 11

2338
was 3-phase, 37.5 kVA, 480 V input, and 120/208 V output. problem. Howwer, if one of these 11 panels had been
This particular UPS drew a line current that averaged 19.0 % carrying its rated load, the neutral conductor could have
THD for the three phases andl had an absence of triplen easily overheated because it was sized the same as the phase
harmonics (Fig. 1la). The UPS input consisted of a six-pulse conductors instead of the 200% recommended for panels with
rectifier and a smoothing inductor which drew a line current a substantial nonlinear load [14, 151.
whose harmonic spectrum consisted mostly of the 5th, 7th,
llth, 13th, 17th, and 19th harmonics. VI. CONCLUSIONS
The load current had a crest factor above 2.0 which caused
the flat-topping of the voltage output as shown in Fig. llb. Harmonics is an issue that has only come to the forefront
The UPS output voltage waveform for each of the three in general electrical distribution design in the last few years.
phases had a distortion ranging from 8.0 to 16.8 % THD. Because harmonics are present in all distribution systems, the
IEEE 5 19, Recommended Practices and Requirements for quick method of measurement with two different types of
Harmonic Control in Electrical Power Systems, states that multimeters anld simple calculations outlined in this paper
voltage distortion should not exceed 5% at the end use loads help the engineer to identify where high levels of harmonic
[ll]. Obviously, this UPS did not meet these requirements. voltages or curirents may be present. The engineer can then
use a more expensive and time-consuming power analyzer to
HARMONICS
V. FINDING BEFORE ARISE
PROBLEMS examine the spectrum of harmonics present and to determine
their source. The engineer must then decide whether the
The intent of the survey of the DOE buildings was first to harmonics pose a potential hazard to the electrical system and
determine the extent of harmonics in the facilities and what steps should be taken to suppress the harmonics.
secondly to apprise whether any of the locations that had a
significant level of harmonics could lead to an electrical REFERENCES
problem in the near future. IEEE Task Force, ‘Effects of harmonics on equipment”, IEEE Trans.
Voltage THD should be less tlhan 5% throughout a plant’s PowerDelivery, vol. 8, Apr. 1993, pp. 672-680.
electrical distribution system. Values above 10% should be IEEE Task Force, ‘The effects of power system harmonics on power
system equipment and loads”, IEEE Trans. Power Apparatus and
investigated thoroughly, and some harmonic mitigation Systems, vol PAS-104, Sept. 1985, pp. 2555-2563.
equipment will be necessary to bring the value to less than A. Mansoor et. al, ‘Fredicting the net harmonic currents produced by large
5%. The voltage THD for all of the panels surveyed at the numbers of distributed single-phase computer loads”, IEEE Trans. Power
Oak Ridge DOE plants was between 2.0% and 4.5%. Delivery, vol. ]LO, Oct. 1995, pp. 2001-2006.
J. Lai, D. Hurst, and T. Key, ‘Switch-mode power supply power factor
Simply having a high THD for a current waveform does not improvement via harmonic elimination methods”, Con$ Rec. Applied
necessarily pose a problem to the electrical system or other PowerElectron. Con$ (Dallas, TX), Mar. 1991, pp. 415-422.
nearby connected loads. The RMS magnitude and the R. Dwyer et. al, ‘Evaluation of harmonic impacts from compact
fluorescent lights on distribution systems”, IEEE Trans. Power Systems,
distribution of the harmonic spectrum are more important vol. 10, NOV.1995, pp. 1772-1779.
indicators for determining an electrical system’s susceptibility F. V. Topalis, ‘Efficiency of energy saving lamps and harmonic distortion
to harmonics. Higher frequency harmonics are more likely to in distribution systems,” IEEE Trans. Power Delivety, vol. 8, Oct. 1993,
interfere with communication lines or electronically pp. 2038-2042
N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics:
controlled equipment than harmonics which have a frequency Converters, Applicahons, andDesign, John Wiley & Sons, 1989.
closer to that of the line frequency. Harmonic currents also W.D. Stevenson, Jr., Elements of Power System Analysis, McGraw-Hill
cause increased heating effects im power handling equipment Book Company, 1980.
C. Melhom and M. McGranaghan, ‘Pnterpretationand analysis of power
such as cables, motors, and capacitors. Some of the heating quality measurements’: IEEE Trans. Industry Applications, vol. 3 1, Nov.
effects such as skin effect are directly proportional to the 1995, pp. 1363-1370.
frequency, and others such as eddy current losses in a L. M. Tolbert, “Harmonic Analysis of Electrical Distribution Systems,”
ORNL-6887, Oak Ridge National Laboratory Report, March 1996.
transformer are proportional to frequency squared. These IEEE Recommended Practices and Requirements for Harmonic Control
increased heating effects can shorten equipment operating life in Electric Power Systems, IEEE Std 5 19-1992.
[ 13. Circuit breakers, capacitors, and transformers which A McEacherrb Handbook of Power Signatures, Basic Measuring
carry current that has a substantial THD (greater than 25%) Instruments, 1989.
The Dranetz Field Handbook for Power QualifyAnalysis, 1991.
and are loaded to above 75% of their capacity should be A Freund, ‘Double the Neutral and Derate the Transformer - or Else!”,
checked to see that they are within their operating E C W , Dec. 1988, pp. 81-85.
temperature limit. Load derating of the equipment may then R. G. Ellis, “Harmonic analysis of industrial power systems,” IEEE Trans.
IndustryApplirations, vol. 32, Mar. 1996, pp. 417-421.
be necessary to maintain a normal operating life.
Only one of the 200 panels surveyed at the Oak Ridge DOE
plants was carrying a current greater than 75% of its listed
ampacity, and only 10 were loaded greater than 50% of their
rating, Of the 200 panels surveyed, 11 had current higher in
the neutral than any of the phase conductors. Because the
panels were not carrying their rated load, this did not pose a

2339