Distribution network modeling for Power Line

Communication applications

T.Tran-Anh, P.Auriol T.Tran-Quoc

Centre de Génie Électrique de Lyon (CNRS n° 5005) IDEA
École Centrale de Lyon BP 46, Lab. Electrotechnique de Grenoble
Écully, France Saint Martin d’Hères, France
E-mail: [email protected] E-mail: [email protected]

Abstract—Power Line Communication (PLC) is a rapidly which assumes a complete decoupling between the 50Hz
evolving technology, aiming to use electrical power lines for the power transmission and the PLC signal transmission.
transmission of data. The used signal is high frequency and in the
range of 1 to 30 MHz. For PLC applications, this paper presents Therefore, this paper presents a method that permit to
the modeling of the high frequency distribution network that evaluate the propagation of PLC signals. The advantage of the
includes overhead or underground cables and the MV/LV proposed method is:
transformer by ATP/EMTP software. Impacts of several factors
on the PLC attenuation such as frequency and line length are
• Its simplicity and its accuracy
investigated. A method, that permits to evaluate the influence of • The use of a well-know software (ATP/EMTP)
50 Hz power signal on the attenuation of PLC transmission, is
proposed as well. • The possibility to study the impact of several factors on
the PLC attenuation.
Keywords- PLC, Distribution Network, Cable and Line,
Transformer, Modeling • The consideration of possible coupling of the PLC
signal with the 50Hz voltage waveform.
I. INTRODUCTION
II. CABLE AND LINE MODELING
Power Line Communications is known for many years as
Power Line Carrier. It uses the low bandwidth analog and The cable and line modeling is carried out with
digital information to communicate over the residential, ATP/EMTP. This is one of the most popular software in the
commercial, and high voltage power lines for AMR (Automatic electro-technical world for the power network transient
Metering Reading), home automation, and protective relay… modeling. The software is able to model the network in time
The fast development of new communication services domain or frequency domain with accuracy.
(Internet, voice over IP, video...) and the deregulation of the The signal with the frequency in range of 1 to 30 MHz
telecommunication market give both electricity and telecom corresponds to a wavelength from respectively 300m to 10m,
sectors a new significant business potential. while the length of cable and line in distribution network varies
The main idea of PLC is to use the electrical grid for the from few meters to few kilometers. Therefore, the length of
communication because it is an existing infrastructure and it line and cable becomes considerable in comparison with
covers a wider area than any other traditional communication wavelength. Thus the line and cable model have to be
networks (telephone, TV cable…). This permits to avoid established by the distributed parameters. In addition, in high
building a new network. frequency these parameters depend strongly on the frequency.
So the model for the PLC signal simulation in distribution
In order to be a high rate communication system, PLC network, must be distributed and frequency dependent.
system requires a data transfer speed, in general, greater than
1Mbps. Currently, there is no common standard to enforce the In this part, some of the most well-know models in this type
frequency band, but some papers suggest that the frequency of are investigated. They are the Semlyen, the Noda, and the
signal should be greater than 2-3 MHz [1] or in range of 1 to 30 J.Marti model. All of these models use Carson’s formula to
MHz [2]. In this paper, we consider that the PLC signal is calculate the parameters of lines or cables, but each of them has
sinusoidal in the frequency range of 1 to 30 MHz. the different method to integrate in the power network.
Actually the applications of PLC system remain essentially The Semlyen model [5] is a 2nd order recursive –
in indoor and low voltage (LV) network. There are several convolution line model. It simplifies the approximation of the
methods [3, 4] that permit to evaluate the attenuation of PLC parameter of line and cable by using only the 2nd order rational
signals but these methods are either complex or supposed to function, thus allows the usage of the recursive convolution in
use for the LV and indoor network. The whole MV/LV system time domain modeling.
is not considered yet. The 50 Hz signals are totally eliminated,

0-7803-8844-5/05/$20.00 c 2005 IEEE. 361
 The J. Marti model [6] approximates the frequency profile of attenuation along the cable or as a function of
dependent parameters of lines and cables into the rational frequency and line length.
functions. Then theses functions are replaced by an equivalent
circuit. This circuit can reproduce all the characteristics of lines III. TRANSFORMER MODELING
and cables.
The modeling of the MV/LV power transformer in high
The Noda [7] model represents the lines and the cables in frequency is very complex due to the frequency dependent and
phase domain rather than in modal domain. In this model, the nonlinear parameters. For our study, we can neglect the
time domain convolutions are replaced by an ARMA (Auto- nonlinear characteristics, because we consider small amplitude
Regressive Moving Average) model that helps to reduce the signals. A transformer model at high frequency needs taking
computation work. into account: self and mutual inductance between coils, skin
Among three above models, we decided to choose the effects and proximity effects in coil, eddy currents losses in
J.Marti model to realize the modeling because of its stability core, self- capacity, and capacity between coils.
and its convenience for our work. In fact, the transformer modeling can be divided in two
essential trends: detailed internal winding models and terminal
models. The first one represents the transformer by a complex
R, L, C circuit. The value of R, L, and C is the result of
complex field problem resolutions. This kind of model requires
the information on physical layout and construction details of
the transformer, and it always needs a huge computation work.
The second one considers the transformer as a black box. The
equivalent circuit is developed from the frequency
a) b) characteristics seen from the terminals. Thus this kind of model
does not need any knowledge about the internal construction of
Figure 1. Cross-section of MV underground (a) and MV overhead cable (b) the transformer. For us the second one is better in term of the
easy usage, of the computation time and of the interaction with
In France, as in many countries, the distribution network the other elements in power network. This is why some models
uses different kinds of normalized section cables, which are of second type are investigated here.
either overhead or underground. In this paper four types of
cable are used for studies: The principle of the first model [8] is to represent the
transformer by a complex-value admittance matrix which is
• Medium voltage (MV) underground cable symmetrical and frequency dependent. This matrix matches the
original of the nodal admittance matrix over the frequency
• Medium voltage overhead cable range of interest. The elements of the matrix are approximated
• Low voltage underground cable with the rational function that contains real as well as complex
poles and zeros. Then the equivalent circuit is established in
• Low voltage overhead cable order to be integrated in the simulation program.
The authors in [9] proposed a new approximate method that
TABLE I. PARAMETER OF CABLES replaces the admittance matrix of transformer by an equivalent
Parameters MV underground MV overhead circuit. This method called vector fitting shows the efficiency
but needs a complicated measurement.
Conductor radius 6,9 mm 4,2 mm
Inner radius of gain 15 mm _ A third high frequency transformer model [10] is also
developed from the frequency response measurement through
Outer radius of gain 16 mm _
its terminals. This model needs the preliminary knowledge
Total radius 20 mm 10 mm about the phenomena that happen in the transformer at high
Conductor resistivity 3.10-8 Ω.m 3.10-8 Ω.m frequency. The measurements are carried out with the different
Relative permeability of conductor 1 1 connections of the terminals in order to represent each
phenomenon by an equivalent circuit. And then the circuits are
Relative permeability of gain 1 1
gathered in a unique circuit that represents the whole
Relative permittivity of insulator 3 3 transformer. This model takes into accounts:
• The self-capacitances of winding.
Two among these above cable types, MV underground
cable and MV overhead line are used for the study. They are all • The capacitance between windings, between windings
tri-phase typical cables. The underground cable is insulated, and ground.
and is reinforced by a grounded sheath. This sheath is located
inside the insulator. For the overhead line, it has a conductor • The magnetic induction and the frequency dependent
enclosed by only an isolator. The cross-section of the cable and series impedance.
the overhead line is shown on Fig. 1. The parameters are • Multiple resonance
detailed on Tab. 1. The modeling is carried out to determine the

362
 Primary side Secondary side In this case a LV or MV distribution network is used for the
Connected in delta Connected in star studies of PLC attenuation. The attenuation is calculated by
using the following formula:

Phase A
 V 
att . = 20 lg m 
V 
 injected 

Phase B
where:
Vm: Modeled voltage along the line
Vinjected: Injected PLC voltage.

Phase C
Figure 2. Transformer model used for study

The fourth high frequency transformer model [11] uses the

same method as the third one in order to determine the
equivalent circuit. However, the equivalent circuits which
represent the phenomena are different. And these equivalent
circuits are added to the conventional transformer model which
consists of an ideal transformer, a winding resistance, a leakage
inductance, and a magnetizing conductance and inductance.
This model takes into account: Figure 3. 50Hz voltage without PLC signal.
• Winding to winding and winding to enclosure capacity Fig. 3 shows the 50Hz voltage without PLC signal. Fig. 4
• Skin effect of winding conductor and iron core shows the 50Hz voltage superposed with PLC signal. The
difference gives the PLC signal shown in Fig. 5.
• Multiple resonance
In this paper, the equivalent circuit of transformer model is
shown in Fig.2. This model is established for a MV/LV core-
type transformer.

IV. PROPOSED METHOD FOR PLC ATTENUATION

EVALUATION
This part presents a method that permit to evaluate the PLC
attenuations in the distribution network including both 50Hz
power and PLC signal. The simulation is carried out by two
steps:
1) The distribution network is simulated with 50Hz power Figure 4. Voltage with both 50Hz and PLC signal.
signal and without PLC signal (SIG1)
2) The distribution network is simulated with 50Hz power
signal and with PLC signal injection(SIG2)
The difference between SIG1 and SIG2 enables us to
extract PLC signal injected on the distribution network. This
method allows superposing the PLC signals on 50 Hz power
signal, thus giving the possibility to study the network in the
real condition of work.

V. SIMULATION RESULTS

A. Case 1: Simplified network

Figure 5. Extracted PLC signal.

363
 • Cables (underground, overhead, LV, MV, length…)
• Transformers (rated power, type, connection…)
• Loads
• Frequency of PLC signals
• Injection point of PLC signals (LV or MV…)
• Others…
The impact of these factors on PLC signal attenuation is
shown in this part.
LV2 LV3 LV4

LV1
Figure 6. Influence of frequency and line length on the attenuation of PLC
MV1 MV2 32m 31m 13m
signals for MV underground cable.

115m
50 Hz LV5 LV6 LV7
Signal
18m 27m 25m
MV Side LV Side

Figure 8. MV/LV network used for the study

1) Impact of position of the PLC signal injection :

The influence of PLC signal injection position is studied.
The study is carried out for 2 cases. In first case the PLC signal
is injected on three phases in LV side (at LV1 point in Fig. 8).
Figure 7. Influence of frequency and line length on the attenuation of PLC In the second case the PLC signal is injected on three phases
signals for MV overhead cable but in the MV side (at point MV1 in Fig. 8).

Fig. 6, 7 show the influence of frequency and line length on For the first case, a PLC signal of 10V in amplitude is used.
the attenuation of PLC signals for MV underground and MV It is about 2.5% of amplitude of 50Hz power voltage signal
overhead cables. Several comments are carried out: (0.4kV). In the second case, suppose that we also use a PLC
signal which represents 2.5% of 50Hz power voltage signal
• In both types of cable when the frequency increases the (10kV), and then a signal at 250V amplitude is selected to
number of peak is also increased. This is explained by inject in the MV side.
the one quarter of wavelength phenomena
Phase A Phase B Phase C
• With the same line length the number of peak is the
same but their amplitudes are different. This is due to 5
Voltage attenuation (dB)

their different structure.

0
• It shows that the attenuation of PLC signal depends on -5
its frequency and the length of line.
-10
B. Case 2: MV/LV network -15
In this case, we use a MV/LV distribution network shown
-20
in Fig. 8 to study. This network includes a MV underground
LV2 LV3 LV4 LV5 LV6 LV7
cable, a MV/LV transformer and a LV grid.
The feeder connected to the load LV2, LV3, and LV4 is an
underground cable. The section of conductor is normalized at Figure 9. Attenuation for different nodes when the PLC signal is injected in
240 mm2. The feeder connected to the load at LV5, LV6, and LV side (LV1 in Fig. 8)
LV7 has the same type but its conductor has 95 mm2
normalized section. The transformer is the MV/LV core-type The following formulas are used to calculate the attenuation
with the rated power 100 kVA. on the MV and LV side:
V  V 
In order to evaluate the PLC signal attenuation, the att . = 20 lg m  ; att . = 20 lg m 
following factors can be taken into account for the study: MV  250  LV  10 
   
• Network configuration (MV or LV…) where Vm is the measured PLC signal

364
 The attenuation of the PLC signal for different nodes is LV2, LV3, LV5, and LV6 are varied. The PLC signal is
shown in Fig. 9, 10. injected at MV1 or LV1. In this case, the attenuation remains
the same. This can be explained by the inductive characteristic
Phase A Phase B Phase C of load. With high frequency signal, its impedance amplitude
becomes very high, almost infinite, this is why the influence of
-10 load is very small for the PLC signal attenuation. In order to
Voltage attenuation (dB)

-20 obtain a higher precision for this study, a frequency dependent

-30 load model should be used.
-40
-50 VI. CONCLUSION
-60
Some methods in the aim to simulate a distribution network
-70 in high frequency range are investigated. A method that permits
MV2 LV1 LV2 LV3 LV4 LV5 LV6 LV7
to evaluate the attenuation of PLC signal has been proposed.
The results are reliable and help to estimate the propagation of
Figure 10. Attenuation for different nodes when the PLC signal is injected in PLC signal through an area which is wider than usual home
MV side (MV1 on Fig. 8)
network.
From this result, we can say that: Different factors and elements of a real network are taken
into account such as the 50Hz power signal, the network
• The propagation of PLC signal injected in MV side is configuration, the injection mode, the load, the transformer, the
more attenuated. It is caused by the strong attenuation line and the cable… In the future, we will extend this
along the MV cable and through the MV/LV simulation to a more complex and real system with the on-site
transformer. experience comparison.
• The attenuation of the PLC signals injected in MV side
(Fig.10) presents a slight difference between the REFERENCES
phases. This can be explained by the unbalance [1] Charles J. Kim, Mohamed F. Chouikha “Attenuation characteristics of
parameters between phases in the MV/LV transformer. high rate home-networking PLC signals”, IEEE Trans. On Power
Delivery, Vol. 17, No. 4, October 2002. pp. 945 – 949.
2) Impact of injection phase
[2] F. Issa, D. Chaffanjon, E.R. de la Bathie, A. Pacaud, “An efficient tool
In this case, the PLC signal is injected only on one phase at for modal analysis of multi-conductor transmission lines for PLC
network development”, Proc. International Symposium on PLC and its
point MV1 in Fig. 8. applications, Athens ( Greece ), 2002.
[3] T. Calliacoudas, F. Issa, “Multi-conductor transmission lines and cables
Phase A Phase B Phase C solver, an efficient simulation tool for PLC networks development”,
0 Proc. International Symposium on PLC and its applications, Athens (
Voltage attenuation (dB)

Greece ), 2002.
-10
-20 [4] L.T. Tang, P.L. So, E. Gunawan, S.Chen, and T.T. Lie “Characterization
and modeling of in-building power lines for high speed data
-30 transmission”, IEEE Trans. On Power Delivery, Vol. 18, No. 1, January
-40 2003.
-50 [5] ATP Rule Book. Available on www.emtp.org
-60 [6] José R.Marti, “Implementation at BPA of a new frequency – dependence
-70 model”, EMTP Newsletter, vol. 2, num. 3, February 1982, pp. 33 – 37.
MV2 LV1 LV2 LV3 LV4 LV5 LV6 LV7 [7] T. Noda, N. Nagaoka, A. Ametani, “Phase domain modeling of
frequency - dependent transmission lines by means of an ARMA
model”, IEEE Trans. On Power Delivery, Vol. 11, No. 1, January 1996.
Figure 11. Attenuation in case of the PLC signal injected on one phase
[8] A. Morched, L. Marti, J. Ottevangers, “A high frequency transformer
model for EMTP”, IEEE Trans. On Power Delivery, Vol. 8, No. 3, July
The attenuation of the PLC signal for different nodes is 1993.
shown in Fig. 11. This demonstrates: [9] B. Gustavsen, A. Semlyen, “Application of vector fitting to state
• a coupling between the phases, therefore the PLC equation representation of transformers for simulation of
electromagnetic transients”, IEEE Trans. On Power Delivery, Vol. 13,
signal injected on one phase (phase A) can propagate No. 3, July 1998.
on two other phases (phases B and C) [10] C. Andrieu, E. Dauphant, D. Boss “A frequency dependent model for
MV/LV transformer”, International Conference on Power System
• the attenuation is the same for phases B, and C at Transients, June 20-24, 1999, Budapest, Hungary.
MV2. But beyond the transformer, the signal is [11] T. Noda, H. Nakamoto, S. Yokoyama “Accurate modeling of core-type
unbalanced in three phases distribution transformers for electromagnetic transients studies”, IEEE
Trans. On Power Delivery, Vol. 17, No. 4, October 2002.
3) Impact of load
To investigate the impact of loads on the PLC signal
attenuation, the network in Fig. 8 is still used. The loads at

365