Sunteți pe pagina 1din 8

IEEE Transactions on Power Delivery, Vol. PWRD-1, No.

4, October 1986 23
DESIGN OF A SYSTEM FOR AUJTOMATED MEASUREMENT AND STATISTICS CALCULATION
OF VOLTAGE AND CURRENT HARMIONICS

J. A. Orr D. Cyganski A. F. Emanuel R. T. Saleh


Member, IEEE Member, IEEE Senior Member, IEEE

WORCESTER POLYTECHNIC INSTITUTE, WORCESTER MA 01609

ABSTRAcT
The design and operation of a portable, personal accurate representation of the harmonic state of a bus
computer-based system for the measurement and process- requires acquisition of much data, as well as an
ing of data on 60 Hz. and harmonic current and organized means of processing and presenting the
voltage amplitudes and phases, and for the production acquired data.
of statistics and time trends on those quantities is
described. Over the data acquisition interval, which This paper describes the design and operation of
may be several days in length, the system calculates a portable, personal computer-based system for the
and stores for each selected harmonic frequency of measurement and processing of data on 60 Hz and har-
each selected voltage or current channel, statistics monic current and voltage amplitudes and phases, and
chosen from among: Mean, Mean Square Value, Variance, for the production of statistics and time trends on
Maximum, Minimum. Trends in composite quantities such those quantities. The system was designed for use in
as Telephone Influence Factor and Total Harmonic Dis- troubleshooting cases of harmonics problems and for
tortion may also be calculated over the interval. gathering data on the present level of harmonic
Samples of the graphical presentation of statistical "pollution" on the power network. It also would be of
results are included. use in non-harmonic situations such as in monitoring
power factor variations, daily load profiles, and bal-
ance among phases. The system described here is
INTRODUCrION intended to be more flexible in application than con-
ventional line voltage condition monitors and
As harmonic producing loads become more common on recorders, and to process data in real time so that
the power network and as they represent larger frac- useful results are immediately available in the field.
tions of the total load, the consequences of the Significant characteristics of this system include the
resulting current and voltage distortions are becoming following: extensive use of software rather than
more significant. In order to understand the harmonic hardware to implement signal processing functions;
situation on a given bus and to predict the effects of applicability to a wide range of measurements includ-
that situation, it is necessary to acquire and process ing harmonic and 60 Hz voltage and current, and
data on the harmonics, and to present that data in an daily trends; on-line data processing and graphics
easily comprehensible form. Previous work (1-51 has display eliminating the need for centralized, off-line
been primarily theoretical or has reported results for processing. A relatively low cost system has been
particular harmonic-producing loads using relatively achieved by use of commercially-available personal
expensive laboratory instrumentation adapted to the computer equipment and a minimum of custom-built
purpose. An ongoing study [61 is gathering data on hardware.
radiated interference in the frequency range from 60
Hz to 30 MHz. Some beginning work [7-9] has been Precautions have been taken in the design of this
reported on techniques for data reduction and statist- system to avoid sources of error common to digital (as
ical analysis to process the raw data into convenient opposed to conventional analog) frequency analysis
and immediately usable form. techniques. These include use of four-pole
anti-aliasing filters, a high resolution (12 bit) A/D
Except in unusual cases, the current and voltage converter, and phase locked loop to synchronize sam-
harmonics on a bus will vary in amplitude and phase pling to the input fundamental frequency (nominally 60
angle over time. These variations are analogous to Hz).
the 60 Hz load current and voltage variations over a
daily period, but typically contain more rapid varia-
tions because of the controlled nature of many SYSTEM CAPABILITIES
harmonic sources. Further, individual harmonics can-
not be assumed to vary in synchronism. Therefore, a Following is a summary of the variables and
large number of harmonics (possibly as many as 50) statistics capable of being measured and calculated
must occasionally be studied. It is clear that an during data runs. Additional types of information
(such as effects on specific types of equipment) may
be derived off-line using these basic results.

Over a period of N hours, divided into subinter-


86 WN 198-6 A paper recommended and approved vals of length T seconds, the system calculates and
by the IEEE Power System Instrumentation & Measure- stores for each selected harmonic frequency of each
ments Committee of the IEEE Power Engineering Society selected voltage or current channel, statistics
for presentation at the IEEE/PES 1986 Winter Meeting, selected from among:
New York, New York, February 2 - 7, 1986. Manuscript Mean,
submitted August 29, 1985; made available for printing Mean Square Value,
November 27, 1985 Variance,
Maximum,
Minimum.

0885-8977/86/1000-0023$01.00O©1986 IEEE

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.
24

Composite quantities such as Telephone Influence Hardware Design


Factor and Total hIarmonic Distortion may also be cal-
culated over the entire interval. The system may also The design and construction of this
calculate probability distributions of magnitude instrumentation has been guided by several factors:
and/or phase angle as histograms, and joint probabili-
ty distributions relating voltage or current 1) The need for portability and ease of use so that
magnitudes or angles on different phases, as the system may be set up at a harmonic trouble site
3-dimensional histograms. easily and quickly.
The above options define the range of data col- 2) The need for unattended operation for periods of
lection possibilities. More details of the data up to several days.
processing capabilities and the theoretical basis for
the statistics are presented in the System Test 3) The requirement of moderate cost.
Results section and in the Appendix. Practical imple-
mentation of these possibilities will be illustrated 4) The need for on-site, real-time data reduction to
below. provide immediately useful results in the field, and
to avoid the necessity for collecting unwieldy amounts
The measurement system is almost completely of raw data for later reduction.
software controlled, and a menu-driven setup algorithm
interacts with the operator to establish the measure- 5) The need for adaptibility to new harmonic measure-
ment conditions for a data run. It is necessary to ment tasks to avoid obsolescence.
carefully organize the setup procedure because of the
flexibility of the system and the large number of data These factors resulted in the selection of a
acquisition options. Options to be selected include: general-purpose 16 bit personal computer as the cen-
transducer type (CT or PT) and ratio, input channels tral processor. Significant characteristics of this
used, input gain, current shunt value, harmonies class of devices include: low cost, ready availabili-
desired on each channel, statistics desired for each ty, standardized interfaces, and software (language
harmonic on each channel, total measurement time, and compilers and graphics output) availability. An
time resolution for each data point in the measurement 8086-based, IBM PC-compatible computer was selected.
interval. This setup data is stored in a disk file The system used here supports full 16 bit data
for later re-use or modification. transfers, has an 8 MHz clock speed, a numeric
co-processor, and 640 X 400 graphics resolution. All
In addition to the above harmonic statistics, a of these characteristics are important to meeting the
time domain waveform distortion classification and goals of real-time signal processing and immediate
storage system is provided. This produces a data base display of the resulting statistics.
library of measured voltage or current waveforms,
together with their probabilities of occurrence, for Fig. 1 shows the functional interconnection of
uses requiring knowledge of actual current or voltage the major system components. Only the analog signal
wave shapes, such as partial discharge analysis. conditioning interface and the phase-locked loop sys-
tem required circuit design and construction. These
are described below. Potentially dangerous voltages
INSTRUMENTATION DESIGN are isolated by current and potential transformers
from the analog and digital electronics. Potential
Measurement Capabilities transformers with stepdown ratios of 40:1 are used in
the prototype. These are appropriate for 120 volt
The data acquisition hardware and its associated circuits, and may easily be replaced by other PT's for
software is designed to have the following charac- higher voltages. Clamp-on current transformers for
teristics: the range of 0 to 100 A are used. These must be
shunted using either the switchable shunts in the
1) One to seven analog input channels, representing interface unit or external shunts. The overall line
any combination of currents and voltages (typically 3 current to input voltage ratio produced by the CT and
phase voltages, 3 line currents, and 1 neutral cur- shunt must be supplied to the data acquisition program
rent); for proper calibration.

2) Current and voltage magnitude ranges determined by The resolution of the A/D converter was selected
transducers (current and potential transformers); to maintain acceptable performance on low amplitude
harmonics. The uantization noise power of an A/D
3) Analog voltage gains of 1, 2.5, 5, 10, 25, 50 converter is S /12 where S is the voltage step size.
switch-selectable to accommodate differing transducer Since this noise power is uniformly distributed over
outputs. Full scale transducer output (including all frequencies in the frequency domain output, the
effect of current shunt) must lie in the range of 0.1 noise power per harmonic may be found by dividing by
to 10 volts to achieve desired accuracy. the number of discrete frequencies in the output
transform. This is equal to N/2 where N is the number
4) Maximum amplitude measurement error of one petcent of samples taken by the A/D converter. Hence the
of full scale. quantization noise power at any harmonic is S2/6N.
The 12 bit A/D converter used here maintains at least
5) Maximnum phase measurement error of i 2 degrees for a 20 dB signal to quantization noise ratio on harmonic
fundamental frequency through the fifteenth harmonic. measurements over an input dynamic range of 2000:1 on
the low frequency range and 4000:1 on the high fre-
6) Two frequency ranges with corresponding sampling quency range.
rates are available: a low frequency range (60 to
1000 Hz) using 1 to 7 channels, and a high frequency Seven independent analog input channels are
range using 1 or 2 channels. provided to accommodate measurement of voltage harmon-
ics on 3 phases and current harmonics on 4 lines (3
7) On the low frequency range, 32 samples per cycle phases and neutral). Also provided on each channel
of the 60 Hz waveforms are taken; on the high frequen- are four-pole active lowpass filters to limit the sig-
cy range, 128 samples per cycle are taken. nal bandwidth before sampling. Plug-in capacitor

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.
25
troublesome in attempting to measure low-amplitude
harmonics.

Any PLL system consists of three major


components: a voltage controlled oscillator (VCO), a
two-input phase comparator, and a lowpass filter.
These are connected in a closed-loop feedback arrange-
ment so that an error signal generated in the phase
comparator by any mismatch of phase (or frequency)
between the incoming and VCO signals is filtered and
used to adjust the VCO frequency. In the locked con-
dition then, the VCO phase exactly matches the phase
of the incoming signal (except possibly for a known
and constant phase offset). Note that since frequency
is the time derivative of phase, phase lock implies
frequency lock. PLL's may be implemented on either an
analog or a digital basis. In the digital case, the
input signal is shaped (often by clipping) into a
logic-valued square wave, and the VC0 generates a sim-
ilar square wave rather than a sine wave.
The PLL used here is of the digital form and
extends the basic concept presented above by raising
the VCO center frequency to 256 times the nominal line
frequency, and adding a divide by 256 circuit between
the VCO output and phase comparator input. The system
operates as before, but now we derive a clock frequen-
cy of 256 times the line frequency, but exactly
synchronized with it. This clock may be used directly
to trigger an A/D converter resulting in 256 samples
per cycle of the line frequency, or the clock output
may be divided by any integer if fewer samples per
cycle are desired. In fact this system requires
either 32 samples per cycle in the narrowband mode
(harmonics 1-15) or 128 samples per cycle in the wide-
band mode (harmonics 1-60). The system software
adjusts the multiplexer addressing to provide the cor-
rect number of samples in each case with no hardware
adj ustments.
The pull-in range of the PLL is from 58 to 62 Hz,
while the tracking range is from 52 to 68 Hz. This is
ample for commercial AC power systems.
Input data is taken during 1/60 sec.
time win-
dows, sampled at either 32 or 128 samples per 1/60
sec., and quantized to 4096 levels. The repetition
rate of these windows is a function of the amount of
statistical processing requested in the particular
data run. A typical value is approximately one data
window per second.
Fig. 1 Functional and interconnection diagram of
major system components.
Software Design

Software for the system consists of three primary


programs: SETUP which assists the user in configuring
packs are provided so that the filters may be operated the system for a data acquisition run, ACQUIRE which
in awideband (0-3600 Hz) or narrowband (0-1000 Hz) actually acquires the current and voltage data and
mode. The narrowband mode is appropriate for taking performs real-time processing, and DISPLAY which
detailed magnitude and phase measurements on the first accesses the stored statistics from ACQUIRE, formats
through fifteenth harmonics of 3-phase currents and and displays them in convenient form. At the heart of
voltages while the wideband mode is appropriate for the data acquisition software is the efficient split
analysis of high frequency harmonics (and other radix Fast Fourier Transform algorithm which produces
interference) across the entire voice frequency spec- harmonic amplitudes and phase angles from the input
trum on 1 or 2 current or voltage channels. time domain data.
The purpose of the phase lock loop (PLL) circuit For normal system use, the operator is not
is to synchronize the sampling rate of the A/D con- required to have any knowledge of the internal details
verter to the fundamental frequency of the power of any of the programs. However, the system is
system being measured. Any deviation from an exact designed for adaptibility to new measurement tasks, in
integer multiple relation between sampling rate and which case some reprogramming will be needed. All
power line frequency results in errors in the calcu- software is written in the PASCAL language. Fig. 2
lated harmonic spectrum. These spectrum errors shows a flow diagram of the major elements of the
(referred to as the Gibb's phenomenon and the "picket ACQUIRE program. It is simple in concept, but fairly
fence effect") can be quite significant even for small complex in reality because of the large number of
deviations in sampling frequency, and are particularly options which must be accomodated.

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.
26

Harmorn i c Charnrnel rutriiber


SETUP cord er
4
S ELE CT __ 1 la a a
CHANNELS
& GAINS 3 Iaa a a
4
.~ a ,a 'a
6
7
SETUP
ISC)
SELECT
STATISTICS 12 IV
13
14
15
v o,r I IV
PT r-at ic:0'O.02700 e0. 02700 0). 0)2700( II

CT rat io,c
Gain 1. 00000 1. 00000 1. (:)0:C)00t

Strike Retorn, to ccorit ir,ue :


Harmon 1 c Char,riell Lriiber
order-
1 4
1 H r
rl H ri± ril H ril±rii

2
H ri± rii H rl±+ri1 H ril rii
4
H rgi±rii H r,i+r H ril+r,i
6
7

1 C)

11
12
13
14
15
THO N N

EPS
IPS/P (re
o-nri h r' ri

Fig. 2 Flow diagram of data acquisition and process- Fig. 3 Example of CRT output displaying types of sta-
ing program. tistical data available from a data run for
specific channels and harmonic frequencies.
Key to abbreviations: a: amplitude (p would
Fig.
3 shows a screen printout illustrating the indicate phase angle); V: voltage; I:
manner in which the program assists the user. This current; m: mean; ±m: min. and max.
display summarizes the available data produced in the values; THD: total harmonic distortion; EPS:
data acquisition program with respect to analog chan- envelope power spectrum; IPS/P: interphase
nels used, harmonics studied, and statistics gathered. statistics on phase angle.

SYSTEM TEST RESULTS


ance. Ten independently variable channels are
To demonstrate and verify the operation of the available for control of firing angle (three channels
hardware and software while maintaining control of the are needed per three-phase load). Tests reported here
measured voltages and currents, a network of were performed in the single-phase mode.
time-varying, harmonic-producing loads was constructed
in the laboratory. It employs SCR's whose firing The equipment arrangement of the harmonic test
angles are controlled by a mini-computer so that they generator is shown in Fig. 4. The comparator and
vary in a pseudo-random, but repeatable pattern. step-down transformer supply to the mini-computer a
Currents and voltages are measured using the standard digital square wave signal synchronized with the AC
transducers so that a complete simulation of actual line voltage waveform. Using this input, the output
field conditions, but with known "randomness" is SCR trigger signal is synchronized to zero crossings
created. The simulator may be operated with any com- of the AC voltage. The output is delivered at TTL
bination of one and three-phase loads, with logic levels via a 16-bit parallel I/O port, and buf-
three-phase loads operating with any amount of unbal- fered by inverters.

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.
t-lu
27
culations, whereas the digital system may
simultaneously measure and calculate statistics on
many harmonics for several input channels.

6.93
6.98
S. S8
mmax
6.23
9.87
5.62
5.17
f 4.82
Ea 4.4? (a)
c 4.12
o 3.76
: 3.41
4.,
2
0.
3.96
E 2. 7i
0
2.36
£c4,o
time
2.34
2.18
Fig. 4 Equipment arrangement of pseudo-random harmon- 2.93
ics test generator. 1.87
1.71
1. 96
D.E 1.49
o
The PASCAL code for the simulator contains five go 1.24
major functions: random number generator; digital
filter to add any desired amount of correlation among
C 1.9s
channels; generation of firing times with respect to o 4.9
AC voltage zero crossings from the filtered random D o.
numbers; storage to RAM for fast recall; output via a _ 9.62
o-
digital parallel I/O port. E 9.46
0

Actual current waveforms resulting from the sum-


9.31
+4
mation of the 10 individual currents are illustrated timne in hours 0.99 I1.8s 2.27
in the section on waveshape data. Examples of the
measurement and data reduction and presentation capa- Fig. 5 Examples of time trend plots of current
bilities are illustrated in the following figures. amplitude; (a): fundamental current
Data were produced using the pseudo-random harmonics amplitude; (b): third harmonic current ampli-
generator described above. tude.

Harmonic Time Trends


Time trend plots of harmonic amplitude provide 3.72 rnean
the most basic data display capability. Each stored 3.46
and plotted point represents an average over the sam- 3.19
- -z (b)
pled harmonic amplitudes in the corresponding time
interval together with the maximum and minimum values 2.92
in that interval. For example, Fig. 5 illustrates 2.66
fundamental and third harmonic current amplitudes for 2.39
a 3 hour period with each point representing 1.8 2. 13
minute averages Each 1.8 minute average is formed from 1.86 (a)
measurement of 200 individual cycles. The only basic C 1. 69
limit on time resolution is the available data sto-
rage. For most purposes it is neither practical nor
c 1.33
necessary to provide very high time resolution (less
3 1.6G
than 1 min.) on very long (on the order of days) data
runs. E 0.53
+4
-4 -I -S I

X s . 27
To validate the A/D conversion, FFT, and data -1 - i---I-
processing hardware and software, a time trend plot time in minutes i.39 9s.9
from this system is compared to a plot produced on a
strip chart recorder using an analog wave analyzer in Fig. 6 Comparison of digital and conventional
Fig. 6. The only discernible differences appear to measurement results; fundamental current
be due to the physical response time limitations of amplitude vs. time as measured with: (a) an
the analog chart recorder. It should be noted that analog wave analyzer and strip chart recorder,
the analog system is only capable of producing one and (b) the digital measurement system.
type of output, and performs no statistical data cal- Scales in the two cases correspond.

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.
28
Harmonic Probabilities Fast Fourier Transform (FFT) method of frequency ana-
lysis. In single phase situations the measurement
Time trend data are useful in correlating system produces phase angle histograms of the same
observed problems with harmonic levels, and for aid in form as the harmonic amplitude histograms previously
understanding the sources of harmonics. However, described. However, such a two-dimensional form is
other representations such as histograms or cumulative not convenient to depict phase angle relations among
probability distributions more directly relate to har- all phases of a conventional three-phase system. For
monic effects such as telephone interference, this situation a three-dimensional graphics display
capacitor failures, or transformer overheating. The was developed. Choosing phase A as the reference, the
data of Fig. 5 are repeated as histograms in Fig. 7. x and y directions of this 3-D coordinate system
Often the percentage of time during which a harmonic represent phase angles of phases B and C, respective-
is above some particular value is signficant; this ly, and the z direction represents the number of
information may easily be produced by numerical occurences of each pair of phase angles of phases B
integration of the histogram data to produce cumula- and C. For example, the origin represents 00 of phase
tive distribution plots. shift of phases B and C with respect to A. The column
of height z=n located at x=120, y=-120 represents n
occurences of the standard fundamental frequency phase
HARMONIC HISTOGRAM relations in a three phase system.
22Z1 . c
channel 4 Fig. 8 illustrates this graphical display scheme
harmonic 1 for the fundamental and third harmonic on a three
phase system. In this laboratory situation the vol-
0

v (a)
L

Li
0
(a)
.0

amps

bin overflow:
HARMONIC HISTOGRA
1845. W
channel 4
harmrtonic 3 90

c
FUNDAMENTAL, PHASE REFERENCE: 0 A
a
L (b)
3
L
RELATIVE PHASE, FREQUENCY OF OCCURENCE
0
u
c4

1li.fl dl
0

ihIhhiI}hAIII1
(b)

."LLi FL
)-J-LLJ4LA4J-LA~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ - -----4-
-

amps (99.59 02 1. 1. 46 '1 .389 '2. 33


bin overflow 0

Fig. 7 Probability information on harmonic amplitudes


based on same data as Fig. 5, plotted as his-
tograms; (a): fundamental current; (b):
third harmonic current.

Phase Angle Statistics


HARMONIC 3, PHASE REFERENCE: 0 A
Knowledge of the time variation and statistics of
harmonic phase angles, on either a single-phase or
three-phase basis, may be desirable in several situa- RELATIVE PHASE, FREQUENCY OF OCCURENCE
tions. These include identification of type and
location of harmonic sources, investigation of harmon- Fig. 8 Three-dimensional histogram plots of phase
ic cancellation with multiple harmonic sources, and angle measured with respect to phase A; (a):
studies of imbalance of harmonic loads among phases. fundamental phase angles; (b): third harmonic
Phase angle information is inherently provided by the phase angles.

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.
29

tages were quite stable, so that all of the measured CONCLUSIONS


phases of the fundamental frequency fell at the nomi-
nal (120,-120) phase angle relation. For the third The capabilities, design, and operation of a per-
harmonic in a balanced system, the voltages would be sonal computer-based system for acquisition,
expected to be in phase. The figure indicates that processing, storage, and display of deterministic and
this is very nearly the case for this measurement, statistical data on current and voltage harmonics has
with some samples indicating small deviations. A been described. By use of modern methods of signal
broader distribution of phase angles would be charac- processing and powerful but relatively inexpensive
teristic of a random summation of single phase commercially-available computer equipment, it has been
harmonic loads. possible to develop a flexible prototype harmonics
measurement system. It is hoped that this or similar
systems will be used to gather needed data on the
Waveshape Data present harmonic situation in the US, as well as for
engineering and troubleshooting of harmonics problems.
The above data are based on frequency domain ana-
lysis wherein the voltage or current waveshape is
decomposed into a sum of harmonically related sinusoi- ACKNOWLEDGEMENTS
dal waves. This harmonic analysis is well suited to
situations such as analysis of resonance conditions or This work was conducted in the Electrical
calculation of Telephone Influence Factor. However, Engineering Department of Worcester Polytechnic Insti-
for effects relating to peak voltages such as partial tute under a research contract funded by Eastern
discharge or capacitor failure, it may be more Utilities Associates.
appropriate to deal with actual current or voltage
time waveshapes. In order to implement this systemat-
ically, a mechanism for storing, cataloging, and REFERENCES
calculating statistics on these waveforms is needed.
Such a mechanism has been implemented in this measure- [1] R. Kniel, "Studies of Voltage Distortion caused
ment system. It creates a database of one-cycle
waveshapes together with frequency of occurence infor-
by Multitudes of Electronically Controlled Equip-
ment with Aleatoric Phase Control", Rev. Gen.
mation for each shape. The system uses basic concepts D. Elect., Vol. 84, No. 11, Nov. 1975, pp.
of artificial intelligence in its choice of waveshapes 794-801.
to include in the database, and in its method of reor-
ganizing the data to make best use of finite memory. [2] J. Blommaert, R. DeVre, R. Kniel, "Studies of
Essentially, the system uses the available system Harmonic Distortion Produced in Low Voltage Net-
memory to keep the largest set of waveforms which are works by TV Receivers", Soc. Royale. Belge. D.
"significantly" different from each other. As new Elect., Vol. 95, No. 1, 1979, pp. 7-19.
waveforms arrive, the definition of "significance" is
changed as necessary and the database is reorganized [3] J.A. Orr, A.E. Emanuel, K.W.
so that no "unusual" waveforms are lost due to lack of Oberg, "Current
Harmonics Generated by a Cluster of Battery
memory. The data are organized as doubly-linked Chargers", IEEE Trans. on PAS, Vol. 101, No.
lists, and a lexicographic ordering is used to facili- 3, March 1982, pp. 691-700.
tate rapid searching for differences between the new
waveform and all previously-stored waveforms. [4] J.A. Orr, A.E. Emanuel, D.J. Pileggi, "Current
These time domain waveforms may be used directly, Harmonics, Voltage Distortion, and Powers Associ-
ated with Battery Charters, Comparisons Among
or their Fourier Transforms may be taken if a frequen- Different Types of Chargers", IEEE PES Winter
cy domain description is needed. Fig. 9 illustrates
the result of a laboratory data run with the Meeting, N.Y., 1982, Paper 82 WM 030-5.
pseudo-random harmonics generator. All significantly [5] EPRI Project RP 1024-1, "Distribution System
different current waveforms are shown superimposed.
Surge and Harmonic Characteristics", January
1980.

8$9 [6] EPRI Project RP 2017-1, "Measurement and Analysis


of Radio Noise Associated with Electric Distribu-
tion Lines," September, 1981.

[7] A. Kloss, "A Probability View on the Power


Converter Harmonics," International Conference on
Harmonics in Power Systems, Worcester Polytechnic
Institute, Worcester, MA, Oct. 22-23, 1984, pp.
65-68.

VS -20

-4$
t -- (8] D. A.
Power
Bradley,
"Graphical Representation of
System
Harmonic Data," International
Conference on Harmonics in Power Systems, Worces-
ter Polytechnic Institute, Worcester, MA, Oct.
-6$

22-23, 1984, pp. 69-72.


[9] J. A. Orr, D. Cyganski, "Data Collection and
Statistical Analysis Techniques for Power System
Time sample (W..31) Harmonics," International Conference on Harmonics
in Power Systems, Worcester Polytechnic Insti-
Fig. 9 Superimposition of current waveforms from tute, Worcester, MA, Oct. 22-23, 1984, pp.
the 73-77.
harmonic test generator cataloged by the
waveshape analysis program.

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.
30
APPENDIX which is expected to occur, and i=L is associated with
the largest amplitude (Xmax). The amplitude range
MATHEMATICAL DEFINITION OF STATISTICS associated with each element is then (Xmax - Xmin)/L =
DX and the amplitude of the center of the range at a
Following are definitions of the basic statistics particular index may be found from
which are collected and displayed by the system.

Time-Dependent Statistics
Xi =
Xmin + DX(i 1/2)
The histogram software operates by comparing each
following statistics are calculated
The and measured value to the amplitudes corresponding to each
presented versus time over the data acquisition inter- index, and incrementing the value stored in that
val: mean, mean square value, extremes, variance. indexed location by one. On completion of the obser-
Each statistic is calculated for the specified subin- vation interval, the sum over all array locations is
terval length (referred to as T seconds here) and equal to the number of measurements, and each location
stored. The complete data acquisition period is contains the number of occurrences of the correspond-
divided into M subintervals, so that the total obser- ing amplitude range. Hence a probability distribution
vation time is equal to M*T seconds. Storing the data may be obtained by dividing each location value by the
in this way enables time variations over the observa- total number of measurements.
tion period to be presented.
Note that the maximum and minimum histogram
The mean for each subinterval is defined as: amplitudes must be established at setup time. If
larger of smaller actual values occur, their values
K cannot be included in the histogram. To monitor these
Xn = 1/K
i-1
Xn events, a counter is incremented for each measurement
which exceeds the histogram range. The histogram is
plotted as a bar graph with bar height representing
where K measurements are taken during each subinterval relative frequency of occurence of each amplitude.
and Xn represents the rms value of the current or vol-
tage harmonic n from each measurement. The mean The histogram concept may be extended to more
square value is calculated similarly: than one dimension to indicate the probability rela-
tion between two (or more) variables. In particular,
it is often desirable to compare quantities among the
X- l/K three phases of a power bus. The direct approach
i=l
(Xn) would require a three-dimensional array, with each
element representing a particular set of amplitudes
From the above values the variance and standard devia- (or phase angles) among the three phases. However, by
tion may be found: normalization of the measurments to one of the phases,
the important information on differences from that
phase may be stored in a two-dimensional array which
]1/ 2 is more compact and more easily displayed. Relative
[x[ Xg - (Xn)]
(-
2
phase angle information among phases A, B, and C is
The maximum and minimum values are easily found: important in engineering studies and diagnosis of har-
monics problems; the generation of these histograms
Xn max Max[ Xn ] over K measurements will be described.
The exactly analogous
procedure is to the
single-variable case. Phase A will be chosen as the
Xn min = Min[ Xn ] over K measurements
phase to which the other phase angles are normalized.
An array Hn(i,j) is used, where index i refers to the
Each of the above values is stored with respect angle of phase B normalized by the angle of phase A
to anindex representing time from the beginning of (0B - OA) and index j refers to the angle of phase C
the observation interval so that actual clock time may normalized by the angle of phase A (OC - 0A). In the
be associated with each data point. case of phase angles (which range between -180 and
+180 degrees) there is no overflow problem. For each
measurement of the selected quantity (current or vol-
Time-Independent Statistics tage of a particular harmonic) the phase angle on
phase A is subtracted from the phase angles on phases
Sample probability distributions (often referred B and C, and the indexes i and j corresponding to this
to ashistograms) may be produced for any current or pair of phases is found. The contents of the indexed
location is then incremented by one. The resulting
voltage harmonic over the entire observation period.
They are calculated by creating an initially empty histogram is three-dimensional, with two dimensions
array Hn(i) of elements indexed by i over the range representing phases B and C, and the third dimension
from i=l to i=L where each element indexed by i is representing the number of occurrences of each normal-
associated with a range of harmonic amplitudes. Index ized phase angle pair.
i=1 is associated with the smallest amplitude (Xmin)

Authorized licensed use limited to: University Colleges of Engineering. Downloaded on October 19, 2009 at 06:31 from IEEE Xplore. Restrictions apply.

S-ar putea să vă placă și