Escolar Documentos
Profissional Documentos
Cultura Documentos
1, January 1989
569
I. Introduction
There has been renewed interest in the analysis of harmonics in electric power systems. Perhaps the proliferation
of electronic loads explains much of this interest. Also the
utilization of high voltage DC systems, many alternative
energy sources, and systems with a high percentage of
fluorescent lighting loads may require an analysis of the generation and propagation of harmonic signals in the power
system. References 111 and [2] contain extensive discussions
of these matters. While attention to power systems has
focused on the propagation of harmonic signals for given
sources, relatively little attention has been given to the
reverse problem: to identify the source of harmonic injection
given certain (and perhaps limited) bus and line measurements. Under some practical circumstances, the question of
identifying the source of harmonics is more pertinent than
the calculation of the propagation of harmonic signals for
given sources.
This paper addresses, in effect, the reverse power flow
study problem. A harmonic power flow study is a technique
whereby line currents and bus voltages a t harmonic frequencies are calculated. This procedure may be a NewtonRaphson based technique [3-51 or a technique based on the
injection current response of the network i5-8,17]. In each of
these methods, the signal level a t harmonic sources (e.g.,
large rectifiers) must be known, estimated, or calculated.
Subsequently, the line currents and bus voltages a t integer
multiplies of the power frequency are calculated. However, if
the source of the harmonic signals is not known, a practical
engineering question may be to identify the location and type
of the harmonic sources; this is the case when operating
problems are encountered which suggest the presence of harmonics. These operating anomalies include inordinate
failures of shunt capacitors, telephone and other communications interference, evidence of timing signal interference
(when digital timing signals are derived from the bus voltage), high negative or zero sequence levels and direct evidence in the form of oscillograms. Thus, attention is turned
toward the use of line and bus measurements to identify the
source of harmonics: a state estimation technique is chosen
here t o estimate or identify the spectrum bus injection
current, I(w). In a conventional power flow study, the bus
voltage magnitudes and phase angles are system states; similarly, in a harmonic power flow study, the bus voltage harmonic components are states. The bus injection currents
functions of the bus voltages, and identification of the bus
injections may be viewed as a state estimation problem. If
the estimation procedure is of sufficient accuracy, I(w) may
be used t o identify the type of harmonic source by the frequency characteristics of the spectrum. Even in instances
where I(w) is of insufficient accuracy to firmly identify the
type of harmonic source, it is usually nonetheless possible t o
identify the busses a t which the sources are located.
The concept of identification of harmonic sources may
have special significance in locations where harmonic standards are adopted. The general tennets of such standards
usually require that the electric utility company maintains
the sinusoidal waveshape of the bus voltage whereas the entities served must maintain certain harmonic limits in the load
current. Identification procedures might be used to locate
loads which exceed specified harmonic current limits.
Lct V,,,(t)
and Ibus(t) be the instantaneous time funclion bus voltages (referenced to ground) and bus injection
currents (finding return path through ground) for an N bus
power system. The Fourier transforms of these two N- vecand Ibus(W). Under conventional cperation,
tors are vbuS(u)
only signals a t the power frequency, wo, occur and thus vec-
570
Pa,
= -
j- v(t)i(t)dt .
T -TI2
This is readily shown to be related to Fourier series
coefficients a i and b; of v(t) and ci and di of i(t) as follows,
2 TI2
ap = -
-T/2
TI2
cp = -
2 TI2
v(t)sin(Pw,t)dt
v(t)cos(Qwot)dt bp = T -TI2
i(t)cos(Pw,t)dt
T -TI2
Pay =
2 TI2
i(t)sin(Pw,t)dt
dp = T -T/2
a p c ~ bpdp
(1)
V(W)
= 7r
c[ao6(wtPwo)+apS(w-PW0)+jh & ( w + P ~ o )
9
- jbpS(w-!w0)1
I(w) = 7r
P
m. Estimation of Ibus(w)
[ ~ ~ ~ ( ~ ~ ~ ~ ) + ~ ~ ~ ( w - P ~ ~ by
) +a ~state
d p estimation
~ ( ~ P w method
~ )
based on Equation (2)
If Eq. ( 2 ) is written for a lossless system with negligible
reactance,
- jdp6(w-Pwo)l
and again one finds Eq. (1) gives the average power, Pa,.
Note that Pa, contains terms associated with all frequencies.
It is possible to single out the average power associated with
one frequency, Pw,, by examining apcp
bpdp. A harmonic
source is said to exist when a t bus k, the bus voltage vk(t)
and ;lljected current ik(t) are such that
agcp
+ bodp > 0
for P # 1. In other words, the average injected power associated with harmonic frequency P wo (other than the power frequency) is positive.
The frequency response of the network may be viewed in
two ways: from the point of view of line incidence, one finds
'bus = Ltlline
In Eq.
+I
-1
0
I:,s
is not enough information to specify both cy and dp. Partition vectors Sbus and Sllnein (za),
Pa)
Sbus = LtSlme
sbk =
at
busses
S,, =
Spu =
spk =
(3)
571
TussELTWHICHINJECTION
- LINES
NsTRUMENTEDAND BUSSES
I
-&SANDBUSSESWITHOUT
INSTWMENTATION
INFORMATION IS
INTERCONNECTED NETWORK
INSTRUMENTED LINES
AND BUSSES
SOURCES
Figure 1.
The measurements here are harmonic spectrum measurements and are measured a t each frequency of interest, hw,,
as
A B
sbu
sbk
bn
[c
E
[E:;]*
(4)
Under this partition, the rows of A and B agree with sb,,; the
rows of c and D agree with s b k ; the rows of E and F agree
with Sbn; the columns of A, C, and E agree with the number
of rows of Syu;and the number of columns of B, D, and F
agree with the number of rows of SPk. The notation (-)+ is
used to denote the pseudoinverse [lo] and Eq. (4) is "solved"
in the least squares sense for s b , as
,s,
= AC'Sbk
+ (B-AC+D)Spk
(5)
572
ph<ol
BUS 3
ESTIMATION
-Ph
00
500
1 1 ~ 1300
0
BUS 4
ESTIMATION
P"0
ph
ARE INSTRUMENTED
Figure 2.
BUSSES INSTRUMENTED
FOR Ph. a" SHOWN
THUS. 1 1 BUSSES
INSTRUMENTED
-4.5db
BUS 6
ESTIMATION
P(kf2) =
-ph
N-1
p(iAT)e-JR'knT
I=o
with the number of sample points, N, being chosen large
enough to get high resolution in the frequency domain, 12. In
order to capture 25th harmonic data (1500 Hz), sampling a t
a minimum of 3 K H z is required. A sampling interval of 0.1
ms was chosen (n = 10 KHz = 62.83 Kr/s). Data above the
19th harmonic were deemed unuseable because the signal
strengths were very low.
The results of this test are partially depicted in Figure
2. When injection spectra a t busses 3, 4, and 6 were
estimated, the results shown a t the right of the figure were
obtained. Note that the injection power a t the fundamental
allcl1
+ blld,,
573
500
Figure 3.
i"
0 db reference
7w0
lloo1300
17% 19wo
rO
db reference
Figure 4.
where Ibk and Vb, are the bus harmonic carrent and voltage
a t instrumented (known) busses.
The estimator (7) could not be verified by field d a t a for
the reasons cited earlier. Perhaps using better instrumentation (for example, better than than in [13]), using high pass
filters to extract harmonic signals, and under circumstances
of very high harmonic source levels, Eq. (7) might be applied.
It is clear that the current approach is superior in the majority of cases.
574
VI. I n a c c u r a c i e s i n t h e identification p r o c e d u r e
In this section, the harmonic source identification procedure (5) is considered for its inaccuracies. The principal
inaccuracies are:
i. Reactive voltampere losses. The system line series inductive reactance causes decreased values of apdp - bpcp due to
"losses". Harmonic current levels are low, typically below
lo%, and one would expect the loss effect to be corresponding low. However, apdp - bpcp may also be very small (e.g.,
1% or less a t ! = 13). In most applications, the reactive
losses probably degrade the estimation more than any other
phenomenon.
where 11*112 denotes the 2-norm and a, and a, are the largest
and smallest singular values of matrix A of rank r. The 2norm of a n m by n matrix A is
(9)
where x is a n n-vector,
lbll2 =
6Xi2Y2
i=l
W.Conclusions
It is concluded t h a t least squares estimators may be
used t o identify the location of harmonic sources in electric
power systems. The identification is based on the fact that
in a linear transmission network, active power associated
with voltages and currents of the same frequency is conserved. Inaccuracies occur due to losses, estimation errors,
and modelling errors. These inaccuracies result in error in
the spectrum of the injected current, I(w). Nontheless, it is
possible to identify harmonic sources. Also, in many cases,
the spectrum I(w) may be used to identify the type of harmonic source.
References
J. Arrilaga, D. Bradley, D. Bodger, 'power System Harmonics," J. Wiley and Sons, Chichester, UK, 1985.
575
[12]M. Merow, "A Non-Iterative Power Flow Study Technique Based on the Method of Minimal Least Squares,"
MS Thesis, Purdue University, W. Lafayette, IN, May,
1987.
(131 D. Carlson, L. Crane, C. Fedora, G. Heydt, A. Lucero,
M. Marz, B. McEnany, R. Parenteau, J. Peterson, A.
Tredeau, "The Analysis and Measurement of Harmonics
in the Vicinity of a n HVDC Inverter," J. Electric
Machines and Power System, v. 11, No. 6, pp. 499-510.
576
Discussion
JOHN E. HARDER, Westinghouse E l e c t r i c Corporation,
Bloomington, Indiana:
The author has presented an
i n t e r e s t i n g , elegant method f o r t h e i d e n t i f i c a t i o n
o f harmonic sources on a dispersed power system. I
would i n v i t e t h e author's
assessment o f t h e
practical
implementation
of
the
technique,
considering t h e f o l l o w i n g :
1. I f several s m a l l sources e x c i t e a h i g h Q
resonance on t h e system, i t would seem t h a t t h e
measurements would be dominated by t h e resonance,
making t h e i d e n t i f i c a t i o n o f t h e l o c a t i o n o f t h e
sources d i f f i c u l t .
2.
3.