You are on page 1of 4

Electrical Power and Energy Systems 26 (2004) 545–548

www.elsevier.com/locate/ijepes

Identification of individual types of harmonic loads


in an electric power system bus
R.A. Alammaria,*, S.A. Solimana, M.E. El-Hawaryb
a
Electrical Engineering Department, University of Qatar, P.O. Box 2713, Doha, Qatar
b
Electrical and Computer Engineering Department, Dalhousie University, P.O. Box 10000, Halifax, NS, Canada

Abstract
Some buses in a power system feed power to loads that are sources of harmonics. This type of bus is called a complex bus. This paper
provides a new application of the constrained least absolute value algorithm (LAV) to identify the type of harmonic loads on a complex bus
and their harmonics contribution to the main bus current. The problem is first formulated as least error squares constrained parameter
estimation problem using LaGrange multiplier to adjoin the equality constraints. To this stage the equality constrained is used to sum the
contribution from each load at certain harmonic frequency. The LAV technique is used to solve the resulting problem. Computational results
are given for actual recorded data.
q 2004 Elsevier Ltd. All rights reserved.
Keywords: Harmonics identification; Constrained estimation; Least error squares; Least absolute value

1. Introduction harmonics. Mori et al. [6] present an algorithm based on


the ANN assuming that only one harmonic load is
Most of the work done on the area of harmonic load connected to a load bus [8], and identifies this load at any
identification assumes that a single load, linear or nonlinear, time. This approach requires long computing time.
is fed from a bus in a power system, and the techniques This paper offers a new application to the constrained
developed in this area focused basically to find an least absolute value algorithm (LAV) to identify the type of
equivalent circuit to this single load, so that it can be easily the loads on the complex bus as well as their harmonic
included in the load flow study. These techniques can be contribution to the main bus current. The problem is first
classified into either time domain technique [3,4] or formulated as an LES constrained parameter estimation
frequency domain technique [2,7]. Most of the proposed problem using LaGrange multipliers to adjoin the equality
techniques use the measurements of the bus voltage and constraints. An equality constrained is used to sum the
load current where the load is connected. contribution from each load. An approximate solution is
In most of the power system networks, there are buses obtained using the LES technique. The technique developed
that individually feed many loads. These loads may or may in Ref. [1] is used to solve the resulting problem based on
not be a source of harmonics and may interfere with each constrained LAV. Results are given for actual recorded data
other. Varadan and Makram [5] propose a simple method from the literature.
for identifying different types of loads on line and describes
a quantitative metric that gives the contribution of each 1.1. Background
individual load type to the total load at the complex bus. The
technique is based on least error squares (LES) and does not The harmonic loads are classified into three categories,
according to the dominant harmonics in each load. The three
account the constraints on the sum of the percentage
classes are [9]:
contribution of the three load types used in the study. In this
case, KCL could be violated for a certain order of
Class A: loads with dominant harmonics of order
* Corresponding author. {5,7,11,13}
E-mail addresses: ralammari@qu.edu.qa (R.A. Alammari); solimans@ Class B: loads with dominant harmonics of order {3,5,7}
qu.edu.qa (S.A. Soliman); elhawary@dal.ca (M.E. El-Hawary). Class C: loads with dominant harmonics of order {3,5}
0142-0615/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijepes.2004.01.004
546 R.A. Alammari et al. / Electrical Power and Energy Systems 26 (2004) 545–548

Table 1 cn # cn # c n ; n ¼ 1; 3; 5; 7; 11; 13 ð11Þ


Harmonic current characteristics of load types 
The values of an ; a n ; bn ; b n ; cn ; c n are given in Table 1.
Harmonic order % magnitude   
Eqs. (3) – (8) can be written in vector form as
Load type A Load type B Load type C Z ¼ HX þ 1 ð12Þ
1 100 100 100 where Z is m £ 1 harmonic components of the total current
3 4.4–4.7 54.8–65.8 6.4 –24.3 to be known in advance, H an m £ n matrix of harmonics
5 32.1–76.5 36.7–43.6 6.2 –14 per unit contents in each type of loads and 1 is the error to
7 16.2–62.1 18.3–20.3 0.9 –5.4
be minimized. In this stage, we consider that a’s and b’s in
11 6.5–24.8 0.5 –0.6 0.2 –6.0
13 5.7–12.7 2.6 –4.0 0.3 –2.3 the above equations are approximately known and are
equal to [9]

Refs. [9,10] gives the harmonic current characteristic of an ¼ ðan þ a n Þ=2 ð13Þ

each load type, as given in Table 1. The harmonic ranges in 
bn ¼ ðbn þ bn Þ=2 ð14Þ
the table indicate the maximum and minimum harmonic 
cn ¼ ðcn þ c n Þ=2 ð15Þ
contents of each load. When analyzing these ranges we use 
the average values of these ranges to avoid, for a while, the Otherwise, the problem will be a nonlinear estimation
nonlinearities resulting in the problem formulation [11]. problem. Now, the estimation problem can be stated
mathematically as
Minimizel1l ð16Þ
2. Problem description and formulation or
MinimizelZ 2 HXl ð17Þ
Given the total load current, I; at the complex bus we
define the parameters a; b; and g as subject to satisfying the equality constrained given by
Eq. (2). The algorithm developed in Ref. [1] is used to
Ia Ib Ic
a¼ ; b¼ ; and g¼ ; solve this problem. The steps in this algorithm can be
I I I summarized as follows:
where Ia ; Ib ; and Ic are the rms currents of loads A, B, and C
and I is the rms current of the complex bus. Here, we require Step 1. Calculate the LES estimates X p by using
aþbþg¼1 ð1Þ Xp ¼ ½H T H21 {H T Z 2 CT ½CðH T HÞ21 C T 21
This equation is simply the KCL equation, which states
 ½CðH T HÞ21 H T Z 2 f } ð18Þ
that the total current at any harmonic frequency should be
equal to the harmonic current at the same frequency Step 2. Calculate the LES residuals vector generated
contributed from each load. from this solution as
In vector form the above equation can be written as
rp ¼ Z 2 HXp
T
C X¼f ð2Þ
Step 3. Calculate the standard deviation for this
where C ¼ col: ð1; 1; 1Þ; X ¼ ½ a b g T per unit contri- residuals vector as
bution of each load to be estimated, and f ¼ ½1: " #1=2
The complex bus total current harmonic components can 1 X
m
2
s¼ ðr 2 rav Þ
be written as m 2 n 2 1 i¼1 i
I1 ¼ a1 Ia þ b1 Ib þ c1 Ic ð3Þ Step 4. Reject measurements having residual greater
I3 ¼ a3 Ia þ b3 Ib þ c3 Ic ð4Þ than s and form the new measurement vector Z and the
matrix H for the rest of measurements
I5 ¼ a5 Ia þ b5 Ib þ c5 Ic ð5Þ
Step 5. Recalculate the LES solution for the rest of
I7 ¼ a7 Ia þ b7 Ib þ c7 Ic ð6Þ measurements, using the same expression in Step 1
I11 ¼ a11 Ia þ b11 Ib þ c11 Ic ð7Þ Step 6. Calculate the new residuals vector generated
from this solution, and rank them in order starting from
I13 ¼ a13 Ia þ b13 Ib þ c13 Ic ð8Þ the smallest
where the parameters a; b and c of the above equations are Step 7. Choose the ðn 2 bÞ measurements with smallest
bounded by residuals, together with the constraints and solve n
equations in n unknowns to obtain the LAV solution as
an # an # a n ; n ¼ 1; 3; 5; 7; 11; 13 ð9Þ

bn # bn # b n ; n ¼ 1; 3; 5; 7; 11; 13 ð10Þ ^ 21 Z^
X^ ¼ ½H

R.A. Alammari et al. / Electrical Power and Energy Systems 26 (2004) 545–548 547

3. Testing of the algorithm Table 3


Contribution of each type load
The proposed algorithm is tested off line mode, where the Contribution AB BC CA ABC
data given in the above tables are used in the tests. These
tests are conducted as follows: a 0.6948 0.000 0.6221 0.54227
b 1.1057 0.8627 0.0000 0.95803
† A load type A is only used to be supplied from the g 20.8005 0.1373 0.37791 20.50030
complex bus of the system with harmonic contents as
Note that the dominant harmonic content in this load is
ZT ¼ ½ 1:0 0:045 0:5 0:5 0:2 0:1 
3rd, 5th, and 7th harmonics which mainly the load type B
† A load type B is only used to be supplied from the
complex bus of the system with harmonic contents as † The two load types B and C are fed from the complex bus
T
Z ¼ ½ 1:0 0:6 0:4 0:19 0:055 0:03  with harmonic contents as

† A load type C is only used to be supplied from the ZT ¼ ½ 1:0 0:85 0:50 0:15 0:009 0:006 
complex bus of the system with harmonic contents as † The two load types CA are fed from the complex bus
ZT ¼ ½ 1:0 0:15 0:14 0:002 0:001 0:003  with harmonic contents as
Table 2 gives the results obtained for each study. ZT ¼ ½ 1:0 0:15 0:75 0:45 0:1 0:075 
Examining this table reveals the following remarks: Table 3 gives the results obtained for the per unit
contribution of each type load. Examining this table reveals
† The proposed algorithm is succeeded to identify the type the following remarks:
of load; the largest values in each column (bold numbers)
indicate the most contribution load to the system. † The technique successfully identified the type of load
† For load type A, a in the first column has the largest that are fed from the complex bus, the parameter having
value, for load type B, b in the second column has the the largest value means that it has more influence on the
largest value, and for load type C, g in the third column complex bus, for example, in the first column b has the
has the largest value. largest value, which means the load type B has the largest
† The sum of the three parameters, a þ b þ g ¼ 1; influence on the harmonic contents of the complex bus.
satisfied KCL at the complex bus. † For the combination of the two loads KCL is still valid at
† The values of the per unit contributions in Table 2 are not the complex bus for the three tests.
unique values, since these values depend on the vector of † The negative sign in the table means that the load type C
the harmonic contents of the complex bus, Z. injects harmonics to the network.
† It has been found that during extensive runs for different
values of Z, the proposed algorithm identified exactly the Finally, the proposed algorithm is tested for the three
type of loads, even the values of per unit contributions types of loads fed from the complex bus at the same time;
are different at each run, and KCL is valid for all runs. the load is ABC with total load harmonic contents as

In the second test conducted, where two load types are ZT ¼ ½ 1:0 0:8 0:65 0:35 0:15 0:075 
fed from the complex bus, we simulate the harmonic
contents in a sequence of AB, BC, CA as The last column in Table 3 gives the results obtained for the
per unit contributions. Examining this table, we notice that
† The two load types A and B are fed from the complex bus the load type B has the largest influence on the complex bus,
with harmonic contents as then the load type A has the next influence on the total load;
finally, the load type C injects harmonics to the system.
ZT ¼ ½ 1:0 0:55 0:70 0:50 0:09 0:09  Note that, as we said earlier, the values of a; b; and g
depend on the harmonic contents of the total load Z. Thus,
the values of these parameters given above are not unique,
Table 2 but they are unique for the given Z.
Harmonic contributions for each load type

Contribution Load type


4. Conclusions
A B C
In this paper, a new technique based on constrained least
a 1.1060 20.0585 20.00432 absolute value is presented for the identification of load type
b 20.0185 1.0810 0.00432
in the complex buses in power systems. The proposed
g 20.0875 20.0226 1.00000
technique is succeeded to identify the load harmonic type
548 R.A. Alammari et al. / Electrical Power and Energy Systems 26 (2004) 545–548

contribution to the total bus current, as well as the load type. [3] Soliman SA, Al-Kandari AM, El-Hawary ME. Time domain
The identification of the contribution of each harmonic load estimation techniques for harmonic load models. Electric Machines
Power Syst 1997;25:885–96.
type to the complex bus is of great importance to the
[4] Varadan S, Makram EB. Generalized load modeling in the
distribution utilities so that they can easily advise the presence of harmonics and distortion. Proceedings of the 25th
customer having the largest influence on the total bus Southeastern Symposium on System Theory. IEEE catalog #
harmonic current to install a filter across its terminal. 93Th0530-6; 1993.
In this paper the average value for each harmonic [5] Varadan S, Makram EB. Harmonic load identification and determi-
component is used to form the matrix H; which may be true nation of load composition using a least squares method. Electric
Power Syst Res 1996;37:203– 8.
to some extent; an accurate mean is needed to solve this
[6] Mori H, Suga S. Power system harmonics prediction with an artificial
problem. As a matter of fact, the fuzzy set theory that solved neural network. IEEE International Symposium on Circuits and
this problem easily is our current research. In addition, the Systems 1991;2:1129– 32.
phase angle for each harmonic component for each load [7] Najjar M, Heydt GT. A hybrid nonlinear-least squares estimation of
type as well as the total current should be used, when harmonic signal levels in power systems. IEEE Trans Power Delivery
dealing with such technique. Furthermore, an accurate 1991;16(1):282– 8.
[8] Heydt GT, Gunther E. Post-measurement processing of electric power
method is required for the harmonic identification of the
quality data. IEEE Trans Power Delivery 1996;11(4):1853–9.
total load harmonic content, such as FFT, LES, Kalman [9] Emanuel AE, Janczak J, Pileggi DJ, Gulachenski EM, Root CE,
filter algorithm, LAV, etc. Breen M, Gentile TJ. Voltage distribution in distribution feeders
with nonlinear loads. IEEE Trans Power Delivery 1994;9(1):
79– 87.
References [10] Emanuel AE, Janczak J, Pileggi DJ, Gulachenski EM, Root CE, Breen
M, Gentile TJ, Sorensen D. Distribution feeders with nonlinear loads
[1] Christensen GS, Rouhi AH, Soliman SA. A new technique for in the northeast USA. Part I—voltage distortion forecast. IEEE Trans
unconstrained and constrained linear LAV parameter estimation. Can Power Delivery 1995;10(1):340 –7.
J Elect Comp Engng 1998;14(1):24–30. [11] Pileggi DJ, Gulachenski EM, Breen M, Gentile TJ, Sorensen D,
[2] Soliman SA, Abbasy NH, El-Hawary ME. Frequency domain Emanuel AE, Janczak J. Distribution feeders with nonlinear loads in
modeling and identification of nonlinear loads using a least error the northeast USA. Part II—economic evaluation of harmonic effects.
squares algorithm. Electric Power Res J 1997;40(1):1– 6. IEEE Trans Power Delivery 1995;10(1):348–56.

You might also like