1

Impacts Over the Distribution Grid from the

Adoption of Distributed Harmonic Filters on
Low Voltage Customers
Maria Emília de Lima Tostes, Ubiratan H. Bezerra, Rogério D. S. Silva, Jorge A. L. Valente,
Carminda C. M. de Moura, Tadeu M. M. Branco

Abstract--This paper presents a study of distributed passive

harmonic filter design to minimize harmonic distortions caused
by non-linear loads in residential customers. The main objectives
in this study are: (i) to improve the power factor, (ii) to reduce
current and voltage distortions to standard limits, and (iii) to
reduce electrical losses. In this sense, several measurements that
were carried out in some domestic customers of the metropolitan
distribution grid of Belém City, in the North of Brazil, have
characterized as main harmonic distribution current sources the
3rd, 5th, 7th harmonic components. According to the results
obtained, and considering the impacts caused in the distribution
grid, a distributed filtering strategy is proposed using passive
tuned filters of low costs to be installed in the customers.
Fig. 1 – Typical harmonic current generation pattern of low-income
Index Terms--Harmonics, measurement, passive filters, power customers measured on Sunday afternoon.
distribution systems, power quality.

I. INTRODUCTION

T HE low voltage customers of a distribution grid may be

considered as harmonic current sources that typically
produce the 3rd, 5th and 7th harmonic components with
significant distortion level, and the 9th and 11th current
harmonics with less significance [1]. In order to better evaluate
the impacts that those distributed harmonic sources have in the
power quality a set of measurements were carried out in the
metropolitan grid of Belém City (the capital of the state of Pará
– Brazil). Figures 1, 2 and 3 were selected among all
measurements as typical harmonic current profiles representing Fig. 2 – Typical harmonic current generation pattern of average-income
respectively a low, an average and a high-income customer. customers measured on Thursday night.
Those measurements were carried out directly at each
individual customer’s energy installation [1]. Based on results
so obtained, prototypes of low cost tuned passive filters were
designed, manufactured and installed in some customer’s
facilities. The projected filters are series RLC filters that are
tuned in the 3rd, 5th, and 7th harmonic components.

Fig 3 – Typical high-income customer’s harmonic current generation pattern,

Department of Electrical and Computer Engineering. measured on Tuesday night.
Federal University of Pará.
P.O. Box 8619. 66.075-900 – Belém – Pará – Brazil.
Phone/ fax: +55 91 211-1634.
In this paper, results comparing the impacts over the
e-mail: tostes@ufpa.br; bira@ufpa.br; carminda@ufpa.br. distribution grid considering the adoption of this proposed
 2

strategy of harmonic filtering at the low voltage customers are Zf – is the filter’s impedance at frequency wn, expressed in
presented. The main objectives in this study are: Ohm (Ω).
(i) To improve the power factor of the customer’s electrical
installation, Considering the resistance R equal to 0.5 (Ω) in (2) and the
(ii) To reduce current and voltage distortions to standard capacitance C varying according to the interval of 2µF ≤ C ≤
limits, 50µF due to commercial reasons, one can obtain the values
(iii) To reduce active and reactive losses. that represent the locus of minimum Zf, for the frequencies of
60 Hz, 180 Hz, 300 Hz, and 420 Hz, with the inductance
Section II presents some considerations about the filter varying according to the interval of 20 mH ≤ L ≤ 200 mH. A
design; section III shows computer results used to determine good solution, which simultaneously minimizes the filter
the filter parameters. The results obtained with the application impedance to some specific harmonic frequency and
of the filter at domestic customers are in section IV; section V maximizes the filter impedance to 60 Hz in order to limit the
presents the conclusions of the paper. losses caused by current at fundamental frequency, may be
obtained.
II. THE FILTER DESIGN At the same frequency the capacitance value is inversely
This work proposes the use of a shunt passive R–L–C filter proportional to the inductance value, and one may choose a
per phase tuned at frequencies of 180 Hz, 300 Hz, and 420 Hz, combination that has a capacitor value that may be easily
which correspond respectively to the 3rd, 5th, and 7th harmonic available in the market. It is also recommended to give
components. The project consists in the determination of the preference to pairs of inductor and capacitor with smaller
R, L, and C (resistance, inductance and capacitance, dimensions and low weight.
respectively) values for each branch of the filter that is to be
connected in parallel with the non-linear customer’s load in III. COMPUTER SIMULATION
order to obtain the minimal impedance at the desired tuning The proposed strategy is to design a passive tuned filter at
frequencies. the 3rd, 5th and 7th harmonic components and to apply that filter
According to the type of customer, one-phase, two-phase or to all customers of the energy distribution grid that present the
three-phase, the filter will have three, six or nine branches, and same energy consumption characteristic, mainly current
may be installed at the customer’s energy entrance, which is capacities and voltage levels. To the filter design, typical
connected at the low voltage side of the distribution single phase and three-phase customers fed by a distribution
transformer, as illustrated in fig. 4, for a single-phase transformer of 45 kVA, 13.8 kV / 220/127 V, with a leakage
customer. reactance of 3.5%, was chosen. The values of the fundamental
current (I1) and of the harmonic currents I3, I5, and I7, as
shown in table 1 are typical values that were obtained through
measurements carried out in many single phase customers.
These current values are used to design and manufacture
single-phase filters to be installed in single-phase customers
with similar harmonic characteristics.

Table 1 – Example of current measurement at a typical single-phase customer

Harmonic Order RMS Current

Fundamental (I1) 4,90 A

Fig. 4 – Single phase shunt passive RLC filter tuned at 3rd, 5th, and 7th 3rd (I3) 1,08 A
harmonics installed at a low voltage customer. th
5 (I5) 0,92 A
At the tuning frequency (wn) the inductive and capacitive th
7 (I7) 0,50 A
reactances of the harmonic filter are equal, resulting in the
relationship given in (1) [3]:
For testing capacitance values varying from 2µF to 10µF
1 the corresponding inductance values were obtained aiming the
wn = (1)
LC maximization of filter’s impedance at fundamental frequency,
The filter’s impedance is given as: in order to assure that a minimum portion of the current at that
frequency will pass the filter, and, at the same time,
 1  minimizing the harmonic impedance under interest,
Z f = R + j  wL −  (2) considering the limitations of the R, L and C components:
 wC  resistance can neither be very large - because the filter’s
Where impedance can’t be higher than the system’s impedance - nor
 3

600mA
very small, because it wouldn’t represent a real value. The
capacitance and inductance values are constrained by size and
cost limitations, which would make the installation of the
400mA
filters at the consumers’ residences not feasible.
After several simulations using the software PSPICE, the
best performance obtained have provided the results shown in
table 2 for the branches of 3rd, 5th and 7th harmonics that form 200mA

the single phase filter adopted. In all cases, the values of the
branches’ resistances, due to project reasons, were specified as
0A
0.5 Ohm. This resistance corresponds to the project value to be 0Hz 100Hz 200Hz 300Hz 400Hz 500Hz
Frequency
attained when manufacturing the respective inductors. I(I7) I(L7)

Table 2 – Values for L and C of the passive single phase shunt filter Fig. 7 – Seventh harmonic current (420Hz) that is filtered [I(I3) in black] and
seventh harmonic current that is generated by the load [I(L3) in gray].
Frequency (Hz) L (mH) µF)
C (µ
It should be pointed out that in the PSPICE simulations the
180 78 10
equivalent impedance of the system was considered as being
300 47 6 the impedance of the transformer, and not the Thevenin's
equivalent impedance at the point of the consumer's
420 24 6
connection in the grid. In spite of that inaccuracy, the results
obtained with the filter have shown a filtering rate of 41.76%,
The results of the PSPICE simulations for the values of L, 34.78% and 54.00% for the 3rd, 5th and 7th harmonic currents
C and R chosen for the branches of the filter are shown in fig. respectively, when compared with the case without filter.
5, 6 and 7 respectively for the branches of 3rd, 5th and 7th It is worth mentioning that the grounding resistance used in
harmonics. In those graphs it can be observed that the these simulations is not the theoretical value of 0.5 Ohms
harmonic currents (of gray color) are being filtered in some assumed earlier, but the real value of 23 Ohms that is very
percentage, and that the currents in the fundamental frequency common to be found among the low voltage customers of the
that cross the filter’s branches are less than 0.8 A. metropolitan distribution grid of Belém. This is the main
1.2A
reason why the theoretical results presented in fig. 5, 6 and 7
don’t filter completely the harmonic currents as expected of a
good filter design.
0.8A
IV. APPLICATION OF THE FILTERS AT DOMESTIC CUSTOMERS
After the installation of the designed filter in some
0.4A customers of the low voltage distribution grid, its operation
was monitored continuously in order to evaluate its
performance according to the natural variation of the loads. It
0A
0Hz 100Hz
I(I3) I(L3)
200Hz 300Hz 400Hz should be noticed that the filter was projected for a certain
Frequency
load condition and that the configuration of the distribution
Fig. 5 – Third harmonic current (180 Hz) that is filtered [I(I3) in black] and grid and its parameters not necessarily correspond to the
third harmonic current that is generated by the load [I(L3) in gray]. projected conditions. The measurements performed in two
1.0A
single-phase customers, with different grounding resistances,
and in one three-phase customer are shown next. It was
observed in general that the filter presented a load depending
behavior. Its performance was better in some periods when the
load condition was next to that considered in the filter design.
0.5A
A. One-Phase Measurements
Figures 8, 9 and 10 show the performance of the filter when
the 3rd, 5th and 7th harmonic currents are considered separately
in a customer with 23 Ohms of grounding resistance. The same
0A
situation is shown in fig. 11, 12 and 13 to a customer with 8
0Hz 100Hz 200Hz 300Hz 400Hz
I(I5) I(L5) Frequency
Ohms of grounding resistance.

Fig. 6 – Fifth harmonic current (300Hz) that is filtered [I(I3) in black] and
fifth harmonic current that is generated by the load [I(L3) in gray].
 4

general, the filter exhibited a better performance when

installed at the costumer with the lower grounding resistance.

12 0
13 5
14 0
14 5
15 0
16 5
0
:5
:3
:2
:0
:5
:3
:2
11
3rd Harmonic

20
Current (%)

11:50
12:35
13:20
14:05
14:50
15:35
16:20
15
without filter
10
with filter

5th Harmonic
60

Current (%)
5
0 40 without filter
Time (h) 20 with filter

0
Fig. 8 – Third harmonic current of the customer with 23 Ohms grounding
resistance. Time (h)

Fig. 12 – Fifth harmonic current of the customer with 8 Ohms grounding

resistance.
12 0
13 5
14 0
14 5
15 0
16 5
0
:5
:3
:2
:0
:5
:3
:2
11

25
5th Harmonic
Current (%)

11:50
12:35
13:20
14:05
14:50
15:35
16:20
20
15 without filter
10 with filter 40

7th Harmonic
Current (%)
5 30
0 without filter
20
Time (h) with filter
10

Fig. 9 – Fifth harmonic current of the customer with 23 Ohms grounding 0

resistance. Time (h)

Fig. 13 – Seventh harmonic current of the customer with 8 Ohms grounding

resistance.
12 0
13 5
14 0
14 5
15 0
16 5
0
:5
:3
:2
:0
:5
:3
:2
11

12
7th Harmonic
Current (%)

10 3
8 without filter 2,5
6
4 with filter 2
THDv%

2 without filter
0 1,5
with filter
1
Time (h)
0,5
0
Fig. 10 – Seventh harmonic current of the customer with 23 Ohms grounding
00:00
02:30
05:00
07:30
11:00
13:30
16:00
18:30
21:00
23:30
resistance

Time (h)
12 0
13 5
14 0
14 5
15 0
16 5
0
:5
:3
:2
:0
:5
:3
:2

Fig. 14 – THDv% measured at a low voltage customer, with ground resistance

11

around 23 Ohm, showing the same period of hours in two different days: one
80
3rd Harmonic

with filter and the other without filter.

Current (%)

60
without filter
40
20:50:00

21:05:00

21:20:00

21:35:00

21:50:00

with filter
20
0
Time (h) 4
THDv (%)

3
without filter
Fig. 11 – Third harmonic current of the customer with 8 Ohms grounding 2
resistance with filter
1
0
Figures 14 and 15 show the THDV%(Total Harmonic
Distortion for voltage) variation for the two costumers with Time (h)
grounding resistances of 23 and 8 Ohms, respectively. In the
case the grounding resistance is around 23 Ohm, it was Fig. 15 – THDv% measured at a low voltage customer, with ground resistance
observed a reduction in the THDV% in some periods, around 8 Ohm, showing the same period of hours in two different days: one
corresponding to a reasonable performance of the filter. In with filter and the other without filter.
 5

Although the filtering performance wasn’t very regular

during all the periods considered, as showed in fig. 8-13, the
1,000

Power factor
THDv% (fig. 14 and 15) presented a more significant 0,800
reduction, especially at the customer with 8 Ohms grounding 0,600 without filter
0,400 with filter
resistance. 0,200
A reduction of the reactive power consumption was verified 0,000
in both costumers (with 23 and 8 Ohm of grounding

21 :00

21 :00

21 :00

21 :00
00
0:
resistances) when the filter is connected, as showed in fig. 16

5
:5

:0

:2

:3

:5
20
and 17.
Time (h)
02 0
05 5
08 0
12 5
14 0
17 5
20 0
23 5
0
:0
:4
:3
:1
:0
:4
:3
:1
:0
00

Fig. 19 – The customer’s load power factor variation, with ground resistance
0,5 around 8 Ohm, for two different periods: with filter and without filter.
0,4
0,3 The use of the filter introduced a significant improvement
Q (kVar)

0,2 without filter in the power factor, considering that in almost all measured
0,1 with filter cases the power factor increases from around 0.7 to 0.9.
0
-0,1 B. Three-Phase Measurements
-0,2
Figures 20, 21 and 22 show the THDV% performance when
Time (h)
the harmonic filter is installed in phases A, B and C,
Fig. 16 – Reactive power required by the customer’s load, with ground respectively, of a three-phase customer. The measurements
resistance around 23 Ohms. The measurements correspond to two different were performed in small periods of time in order to have as
periods of 24 hours: with filter and without filter.
small deviation of the system operation state as possible for
the two measured conditions: with filter and without filter.
Similarly to the one-phase case, it was observed a reduction in
21 0:00
21 5:00
21 0:00
21 5:00
00

THDV% in some periods, depending on the load variation.

0:
:5
:0
:2
:3
:5
20

0,8
THDv-Phase A-Interval 3
Q (kVar)

0,6
without filter
0,4
with filter
0,2 2,8 With filter
0 2,6
2,4
Time (h) 2,2 Without filter
4:00
7:00
01:00

10:00
13:00

16:00

19:00

Fig. 17 – Reactive power required by the customer’s load, with ground

resistance around 8 Ohms. The measurements correspond to two different Time (minutes)
periods: with filter and without filter.

It was observed in all measurements obtained that the Fig. 20 – THDv% measured in phase A at a three-phase voltage customer.
power factors of the customers´ electric installations were The same period of time in two different moments: one with filter and the
other without filter.
improved due to the filter’s capacitance. Figures 18 and 19
show this aspect for different measurement periods for a
customer with 8 Ohm grounding resistance. THDv-Phase B-Interval 3

3 With filter
11:50
12:30
13:10
13:50
14:30
15:10
15:50
16:30

2,9
2,8
2,7
1,2 2,6 Without filter
01:00

10:00

13:00

16:00

19:00
4:00

7:00
Power factor

1
0,8
without filter
0,6 Time (minutes)
0,4 with filter
0,2
Fig. 21 – THDv% in phase B at a three-phase voltage customer. The same
0
period of time in two different moments: one with filter and the other without
Time (h) filter.

Fig. 18 – The customer’s load power factor variation, with ground resistance
around 8 Ohm, for two different periods: with filter and without filter.
 6

when considering the set of customers in the distribution grid.

THDv-Phase C-Interval 3
This strategy is of low cost implementation and may
alleviate the reactive power required from the grid, but special
3,4 attention must be given to the problem of harmonic resonance
3,2 With filter
3 that may arise when using the distributed filtering strategy.
2,8
Without filter
01:00

10:00

13:00

16:00

19:00
4:00

7:00
VI. REFERENCES
[1] M. E. de Lima Tostes, U. H. Bezerra, J. N. Garcez, A. A. Tupiassu e A.
Time (minutes)
C. A. Sena, “Development of Experimental Models for Harmonic
Representation of Low Voltage Customers”, 2001 IEEE – Porto Power
Fig. 22 – THDv% in phase C at a three-phase voltage customer. The same Tech Proceedings, Portugal.
period of time in two different moments: one with filter and the other without [2] R. C. Dugan, M. F. McGranahan and H. W. Beaty, Electric Power
filter. Systems Quality, Mc Graw-Hill Companies, 1996.
[3] J. Arrillaga, D. A. Bradley and P.S. Bodger, Power System Harmonics,
John Willey & Sons,.1985.
Figure 23 shows the power factor improvement in phase C [4] IEEE Recommended Practice for Electric Power Quality, IEEE Std.
of the three-phase customer. The same was observed in phases 1159-1995, New York. IEEE, 1995.
A and B. [5] R. Lamerica, A.Prudenzi, E. Tiroli, D. Zaninelli ,“A model of large load
areas for harmonic studies in distribution networks”, IEEE Transactions
on Power Delivery, Vol. 2, Nr.1, January 1997.
Power Factor-Phase C-Interval 3 [6] Guide for Applying Harmonic Limits on Power System, P519A Task
Force of the Harmonics Working Group (IEEE PES T&D Committee)
and SCC22 - Power Quality, IEEE, May, 1996.
1 [7] M. Z. Lowenstein, “Improving Power Factor in the Presence of
0,9 Harmonics Using Low-Voltage Tuned Filters”, IEEE Transactions on
With filter Industry Applications, Vol. 29, Nr. 3, May/June 1993.
0,8
Without filter
0,7
VII. BIOGRAPHIES
4:00

7:00
01:00

10:00

13:00

16:00

19:00

Time (minutes) Ubiratan H. Bezerra was born in Pereiro – CE –

Brazil in 1950. He received the B. Sc. degree from
Federal University of Pará – Brazil, in 1976, the M.
Fig. 23 – The customer’s load power factor variation for two different Sc. degree from Federal University of Itajubá –
periods: with filter and without filter. Brazil, in 1980 and the D. Sc. degree from Federal
University of Rio de Janeiro – Brazil, in 1988, all in
V. CONCLUSIONS Electrical Engineering. Since 1977 he has been with
the Electrical and Computer Engineering Department
The harmonic impacts over the distribution grid produced at Federal University of Pará, being at present a
by currents injected by low voltage customers can be reduced titular professor. His research interests are focused on
power system static and dynamic security assessment, power quality, and
by the use of passive filters tuned to the frequencies of 180Hz, renewable energy generation.
300Hz and 420Hz, which correspond to the 3rd, 5th and 7th
harmonics, respectively. In this sense, this paper presented Maria Emília de Lima Tostes was born in
some preliminary results involving the installation of tuned Recife, PE, Brazil, on April 6, 1966. She received
the B. Sc. degree from Federal University of Pará in
passive filters in some customers of the low voltage 1988 and the M. Sc. degree in 1992 in the same
metropolitan distribution grid of Belém, in the North of Brazil. university.
Although it has not being observed a regular reduction in She is Professor of the Electrical and Computer
Department of the Federal University of Pará, where
the harmonic current distortion for all load conditions, as was is attending a Ph.D. program. Her main research
expected with passive filter, the use of the filter has areas are Power Quality, Distribution Systems and
contributed, in the average, to a better performance of the Industrial Process.
domestic electrical plant concerning:
• Power factor improvement; Carminda Célia Moura de Moura was born in
• THDV% reduction; Belém, PA, Brazil, on November 27, 1970. She
received the B. Sc. degree in Electrical Engineering
• Reduction of reactive power demand from the grid. from Federal University of Pará – Brazil, in 1994
and the M. Sc. degree in 1996 in the same
If this distributed filtering strategy is to be applied to a university.
She is Professor of the Electrical and Computer
considerable number of customers in the low voltage Department of the Federal University of Pará,
distribution grid it is expected that the system will exhibit in where is attending a Ph.D. program. Her areas of
general a better performance with respect to power quality interest are Power Quality, Generation and
Distribution Systems and Industrial Process.
indicators concerning voltage and current distortions. Also,
due mainly to the capacitance presented by the passive RLC
proposed filter, there is a local reactive power compensation at
each customer individually that may sums up to a large amount