Proceedings of the 14th International Middle East Power Systems Conference (MEPCON10), Cairo University, Egypt, December 19-21,

2010, Paper ID 252.

Practical Issues of Power Line Communication for Automatic Meter Reading Systems
Yasser Fathi, MIEEE, Tamer A. Kawady, MIEEE, Ahmed Husein, and Mohamed El-Geziry
Abstract Power Line Communication (PLC) represents an economic, versatile and dependable tool for distribution automation purposes. It is expected to play a major role towards realizing "Smart Grids" in the future. This paper visualizes some practical issues of Automatic Meter Reading (AMR) based on Power Line Communication systems. It highlights different factors affecting system performance and its time and location dependent behavior. The relation of PLC system performance fluctuations with the average customer current demand in two mechanisms namely Access Impedance and Injected Noise is emphasized. In addition, it explains the performance variation across different system areas. It therefore helps to realize a better understanding of the PLC implementation for real applications which is essential for successful utilization of such tools in the future. Index terms: Automatic Meter Reading, Distribution Automation, , Power Line Communication, Spread Spectrum

Several AMR systems have been evaluated in Egypt in different governorates faced may problems especially at the beginning. Several systems utilizing different modulation techniques were proposed. One of the most robust systems that covered more than 30,000 customers is ARSEL AMR system developed by EOIP1 and utilized one of the Spread Spectrum techniques and will be taken as our case study. Section II, introduces different problems facing PLC systems concentrating on our case study. Section III is dedicated to the measurements setups and evaluation results. Finally, the conclusions are given in section IV. II. CASE STUDY AND PROBLEM CLARIFICATION In our case study of ARSEL AMR system shown in Figure 1, each customer KWH meter is equipped with a PLC transmitter that converts the meter disk movement into an equivalent reading. Subsequently, the meter reading is appended by a unique customer code and sent to a central receiver located in the distribution transformer[9]. The system deploys Direct Sequence Spread Spectrum DSSS modulation technique working in the frequency range defined by CENELEC standard Band A form 9KHz to 95KHz [5], [10].
Complex tree of Customer Loads

tilizing power network for data transmission is not a new idea [1]-[3]. It started as early as the beginning of the last century with the wave of full coverage electrification. Although power line communication systems are getting more and more interest, their performance especially in long distance AMR applications is sill subject of doubt[4]. The performance of any PLC system is found to be heavily depending on time, location and power network wiring topology. Distribution transformers heavily attenuate high frequency signals beyond several kilo hertz and hence form the communication point of view, the low tension network is divided into isolated islands or cells concentrated on each distribution transformer. Communication signaling can only be implemented within the area serviced by one transformer [10]. The first known patent on power line signaling was proposed in 1897 in the United Kingdom [8]. In such systems commands of load shading are broadcasted for a central generation point to meters. In Automatic Meter Reading systems the information emerges from different meters and required to be collected at central location in the reverse direction[9], [6]. Home appliance manufacturers are continuously developing and producing new instruments with less regard to communication signaling issues. On the other hand especially in the developing counties, Arial lines are still heavily used in the LV grid. This environment negatively affects the quality and reliability of any communication system and thus creating a harsh, hostile and horrible channel as described in several papers.

I. INTRODUCTION

PLC Receiver & Data Concentrator Dist. Transformer

1234 KWH Meter

PLC Transmitter Customer Entry Point

Main Feeder 1 1

Feeding other customers

Figure 1: ARSEL PLC based AMR System

The generated PLC signal is coupled to the line at the premises entry point through a special coupling circuit designed to block the line frequency meanwhile passing the higher fre1 EOIP Engineering Office for Integrated Projects an R&D house in Egypt involved in several communication projects serving the utility companies.

634

quency communication signals [10]. Since its generation within the transmitter circuit, the signal experience the coupling circuit power and cable network characteristics in addition it is contaminated by noise. Finally, the receiver has to extract the transmitted information carried by the weekend and distorted signal that is completely embedded in noise. So, we can deduce the three factors almost influencing any PLC system as follows A. Access Impedance PLC devices directly drive the mains terminals and are required to deliver sufficient signaling power depending on the required coverage range. The lumped element equivalent impedance of the power network seen across the coupling circuit terminals is known as Access Impedance and in general is a function of the frequency [10]. Access impedance shunts the power amplifier and its coupling circuit and hence it determines the injected signal power. Every load or appliance connected to the mains contributes to this access impedance. Unfortunately, these loads are not fixed but continuously changing in random patterns resulting in time varying channel response. B. PLC channel Along its path form Transmitter to Receiver, the signal encounters several impedance discontinuities[1]-[3]. It may travel across different cable types and may encounter branching or even couple across phases. Such discontinuities introduce impedance mismatches and dramatically degrade the signal strength by and dividing its power among branches. In addition, no guarantee, that different paths have the same channel characteristics, instead signal from different customers will in general experience different environment[9], [10]. C. Power Line Noise Every connected load running from the mains injects a load specific noise signature. In addition, the power network itself works as a huge antenna that captures different types of electromagnetic waves thus providing a rich cocktail of noise at the receiver front end [7]. III. EVALUATION TESTS A. Current demand and Access Impedance Measurements Customers loads are powered on and off arbitrarily resulting in time dependant current demands and introducing randomly fluctuating impedance. In order to study this time dependence behavior, we have to estimate the ensample average current demand over a large number of customers. Fortunately, the power network hierarchy can be utilized to estimate the average access impedance per customer. The total outgoing current of the distribution transformer is the sum of about 300 customer currents. Furthermore, the MV feeder is responsible for feeding about 6 to 7 or even more distribution transformers and thus providing a summation of very high number of customers.

The Supervisory Control and Data Acquisition SCADA system shown in Figure 2 is utilized to capture and record the current trends over several days through a complex communication network. MV currents are monitored by the outgoing protection digital relays meanwhile the LV side is monitored by a special Transformer Inelegant Module TIM built for this purpose.

Medium Voltage 11 Kv Protection and monitoring Relay

SCADA System

Low Voltage distribution Transformer Transformer Intelligent Module Low Voltage 0.4 Kv

LV Grid

Figure 2: SCADA System Block Diagram

70 60 50 40 30

Phase Currents (A)

FeederCurrentA
20 10 0 3/25 19:12

FeederCurrentB FeederCurrentC

3/26 0:00

3/26 4:48

3/26 9:36

3/26 14:24

3/26 19:12

3/27 0:00

3/27 4:48

(a)
80 70 60 50 40 30 20 10 0 3/25 19:12 3/26 0:00 3/26 4:48 3/26 9:36 3/26 14:24 3/26 19:12 3/27 0:00 3/27 4:48

Phase Currents (A)

FeederCurrentA FeederCurrentB FeederCurrentC

(b)

Figure 3: Current demand of MV Feeder 2 (25 to 27 March)

635

70 60 50 40 30 20 10 0 3/25 19:12 3/26 0:00 3/26 4:48 3/26 9:36 3/26 14:24 3/26 19:12 3/27 0:00 3/27 4:48

500.00 450.00

Phase Currents (A)

Phase Currents (A)

400.00 350.00 300.00 250.00 200.00 150.00 100.00 50.00 0.00 8/14 12:00 8/15 0:00 8/15 12:00 8/16 0:00 8/16 12:00 8/17 0:00
`

FeederCurrentA FeederCurrentB FeederCurrentC

Phase A Phase B Phase C

(a)
80 70 60 50 40 30

Phase Currents (A)

Figure 5: Current demand of LV side of single distribution transformer

180

Average Estimated Custormer Access Impedance(Ohm)

20 10 0 3/25 19:12 3/26 0:00 3/26 4:48 3/26 9:36 3/26 14:24

FeederCurrentA FeederCurrentB FeederCurrentC

160 140 120 100 80 60 40 20 0 3/25 19:12

(b)

Figure 4: Current demand of MV Feeder 4 (25 to 27 March)

B. Access Impedance near the Power Frequency Figures 3, 4 and 5 show the captured current demand of four MV feeders over the same time period. A closer look at the curves reveals several astonishing results. Although customers switch on and off loads arbitrarily, they tend to produce common trends related to the human normal activities. In general, the time from mid-night to noon is considered as low power demand period at which a large number of the customer loads are disconnected. This period corresponds to higher access impedance and lower noise level as a large portion of the home appliances are disconnected. Hence, this represents the best conditions for communication and collecting the meter readings. The recorded samples clearly indicate relatively large variations of the customer current demand. The recorded peak to minimum load demand reached 200% or even more. Turning our attention to the LV side of the network, the LV current demand shown in Figure 5, indicates that the three phases are not well balanced as expected because the most of loads are single phase. The sample shows that one of the phases is loaded by approximately double the current of another phase that represents another source of asymmetry in the PLC channels. Furthermore, the current demand shows the same fluctuating behavior as indicated by the feeding MV. By simple manipulation the average customer access impedance with time is derived and shown in Figure 6.

3/26 0:00

3/26 4:48

3/26 9:36

3/26 14:24

3/26 19:12

3/27 0:00

3/27 4:48

Figure 6: Average Customer Access Impedance

C. Access Impedance Measurement in the Frequency Range 10 - 100 KHz As the frequency rises, most of the components of the electric appliances gradually change their electrical characteristics resulting in frequency dependant impedance. For example, the iron used in the transformers and motors dramatically changes its magnetic characteristics and consequently introducing higher impedance levels. In addition, the bulk capacitors normally included in the motors as phase shift devices and in the switched mode power supplies experiences a resonance frequency in the range below 100 KHz at which they are equivalently behave as coils. In other words, the most of the electronic devices and components that have direct connection to the mains are optimized to operate in the low frequency range and lose their properties gradually as the frequency rises. As a result, the impedance of the home appliances exhibits rising trends with the frequency. Eventually, in the mega Hertz frequency ranges the load impedances takes sufficiently high values to be neglected. Hence, the power line channel characteristics are dominated by the wiring and cable characteristics [11], [12].

636

Sweep sinusoidal transmitter Sweep Generator

Signal conditioning and capturing

Line
Receive Coupling & BPF

Coupling & protection Circuit

Anti Aliasing LPF Filter (fc <100KHz)

Data Acquisition

PC (off line spectral analysis)

Current signal

Voltage signal

Figure 9: Power line noise measurement set-up.

True RMS Measurements

Figure 7: Access Impedance measurement in the frequency range 10 to 100 KHz

In the frequency range from 10 to 100 KHz, both the network architecture and the loads affect the resulting channel properties. A special setup has been developed to measure the impedance magnitude indirectly at several frequency points. The measurement set-up shown in Figure 7 consists of a sweep sinusoidal transmitter driving a capacitive coupling circuit in series with a 1-ohm current sensing resistance, and a band pass filter covering the frequency range from 1 to 100 KHz.

1) Measurement Setup To capture the noise bouncing over the power line a special coupling and protection circuit is used. As shown in Figure 9, this special circuit filters out the high line voltage while passing signal components above 1 KHz. In addition it provides the necessary protection and isolation for the following circuits. The captured signal is then passed through Anti-Alias filter before sampling by the Data Acquisition System. The acquired samples are then offline processed by Welch spectrum estimation algorithm implemented in Matlab software. 2) Measurement results The captured samples are analyzed in time and frequency domains. The noise sample shown in Figure 10 clearly contains a variety of components, particularly, the impulsive noise, the line frequency harmonics and the background high frequency hum with different weights.
1.5

12 10 8
Ohm

6 4 2 0 0 20
Frequency (KHz)

0.5

40

60

80

100

Voltage

-0.5

Figure 8: Access impedance (10-100 KHz)

Figure 8 shows one sample of the measurement indicating impedance values ranging from 2 to 10 ohms with frequencies up to 100 KHz. Such relatively low value must be taken into consideration in designing the PLC Power Amplifier, Power Supply and coupling circuits. D. Noise Measurements As the Signal to Noise Ratio SNR is the figure of merit for any communication system, we implemented several noise measurements in the system transmission band.

10

20

30

40

50

60

Time (ms)

Figure 10: Power line noise sample

A closer look at one of the large periodic impulses penetrated throughout the entire sample is shown in Figure 11. The impulses usually take the shape of a damped oscillation spanning about 1 ms and a peak voltage of 1 volt. The origin of these impulses is referred to different types of switching and sparking loads like motors dimmers and different types of home appliances.

637

1.5

20

Powe r S pe ctrum Magnitude (dB)

10 0 -1 0 -2 0 -3 0 -4 0 -5 0

Ampl i tude (vol t)

0.5

-0.5

-1 51

52

Ti me ( ms )

53

54

55

F re qu e n cy H z
(a)

10
4

Powe r S pe ctrum Magnitude (dB)

Figure 11: Common Shape of Power Line Pulses

10 0 -1 0 -2 0 -3 0 -4 0 -5 0

20

P ow er Spectrum Magni tude (dB )

10

Hi gh PSD porti on of the Col ored Background noi se Monotoni cal l y decreasi ng PSD porti on of theCocl red Background noi se Narrow Band i nterference

-10

-20

-30

F re qu e n cy H z
(b)

10
4

-40

-50

10 x 10
4

Figure 13: PSD of a power line noise sample, Cairo, August 2001, 3.00 PM, Phase 2,3

Frequency (Hz)

P ow er Spectrum Magni tude (dB )

Figure 12: PSD of the power line noise sample, Cairo, August 2001, 3.00 PM, Phase 1

-5

The power spectral density estimation (PSD) of theses samples are carried out using Welch Algorithm. Clearly, Figure 12 and Figure 13 distinguished the main noise components dominating in the low frequency band. Particularly, the narrow band needle shaped noise resulting from the monitors and switching power supplies included in the PC. Through this band, the PSD decreases gradually with the frequency to reach about 35 dB below its initial value at 1 KHz. Obviously, the noise power concentration below 30 KHz indicates the insignificance of using such band for signaling. Focusing on the low range below 10 KHz shown in Figure 14, reveals that, the major components contributing to the noise in this band is the line related harmonics.

-10

-15

-20

-25

5000

6000

7000

8000

9000

10000

Frequency (Hz)

Figure 14: Background noise details showing the line harmonics below 10 KHz

638

E. Field Measurements

Buried Cable Section Buried, Aerial Cable or Overhead Lines Virtual Zoning lines

Zone 1
Outgoing 380V HV/LV Distribution Transformer Incoming 11KV

Hence, the performance of any PLC regarding, coverage range and even the signaling reliability is expected to be dimensioned and vice versa. The network cable construction especially cable types, pillar boxes branching, mixing between different types of cables introduces rich impedance mismatch opportunities in addition to different number of customers resulting in different system performance for each LV Island.

Acknowledgement The authors are expressing their gratitude to Engineering Office for Integrated Projects EOIP the owner of ARSEL and the SCADA Systems. In addition to Middle East Electricity Company Egypt MEEDCO and Canal Company for Electricity Distribution, Egypt CCED.

Supply line (max. length. 1000m)

Zone 2
Supply radius 100-500m
Cable Box

Zone 2

Figure 15: Low Voltage Power Distribution network connection

V. REFERENCES
V. Guillet, G. Lamarque, P. Ravier and C. Leger, Improving the power line communication signal-to-noise ratio during a resistive load commutation ACADEMY PUBLISHER JOURNAL OF COMMUNICATIONS, VOL. 4, NO. 2, MARCH 2009 [2] Lars T. Berger, Gabriel Moreno-Rodrguez Power Line Communication Channel Modelling through Concatenated IIR-Filter Elements ACADEMY PUBLISHER JOURNAL OF COMMUNICATIONS, VOL. 4, NO. 1, FEBRUARY 2009 [3] Anatory J. , Theethayi N. , and Mvungi N. H Power Line Channel Models: Comparisons between different Modeling Adopted in BPLC Systems THIRD WORKSHOP ON POWER LINE COMMUNICATIONS, OCTOBER 1-2, 2009, UDINE, ITALY [4] N. Miura, H. Sato, H. Narita, M. Takaki Automatic meterreading system by power line carrier communications IEE PROCEEDINGS, Vol. 137, Pt. C, No. I , JANUARY 1990 [5] Stefan Ramseier, Michael Arzberger, Andreas Hauser MV and LV Powerline Communications: New Proposed IEC Standards, IEEE Transmission and Distribution Conference, 1999, IEEE, 1999, pp. 235-239. [6] Jrventausta, P., Mkinen, A., Kivikko, K., Verho P., Krelampi, M., Chrons, T., Vehvilinen, S., Trygg, P, Rinta-Opas, A,. Using advanced AMR system in low Voltage distribution network management, Proc. CIRED 2007 Conf., Vienna, Austria. 21-24 may 2007. [7] Edward P., Guillen, Julin J. Lpez,, Cesar Y. Barahona, Throughput Analysis over Power Line Communication Channel in an Electric Noisy Scenario, World Academy of Science, Engineering and Technology 43, 2008. [8] Klaus Dostert, "Power Line Communications", PRINTICE HALL, 2000. [9] El-Ghoneimy M., Kandil A.H., Harb K., "Automatic Meter Reading in Egypt", 6 'th International Middle East Power Systems Conference, MOPECON 1998, pp. 677-684 [10] Yasser Fathi Msc An Enhanced Direct Sequence Spread Spectrum Communication System over the Mains Msc Thesis Cairo University 2004. [11] T. Bostoen and O. V. de Wiel, Modeling the low-voltage power distribution network in the frequency band from 0.5 to 30 MHz for broadband power line communication, Communications, 2000, Proceedings. 2000 International Zurich Seminar on, February 2000, pp. 171-178. [12] G. Marubayashi, S.Tachikawa ; "Spread Spectrum Transmission on Residential Power Line", Spread Spectrum Techniques and Applications Proceedings 1996., IEEE 4th International Symposium on, 22-25 Sep. 1996.

While following the investigated system performance in different sites in Egypt, several notes about the LV grid that may influence the PLC performance have been recorded. As shown in Figure 15 the main output of the distribution transformer is split into several main feeders. Each feeder carries energy to about 100 customer premises. The feeder length can reach up to 1 km or even longer. Although the system construction is a triple phase, the majority of the loads are single phase ones. As a result, the load balance between different phases is not well guaranteed. Consequently, PLC performance differences among phases are expected. This tree configuration results in a minimum impedance at the transformer outlet functioning as break wall. Signals transmitted over one feeder are almost blocked upon reaching this point. When we put a receiver midway on one feeder, it receives only about 30% of other feeders corresponding to near customers. So, the customers fed form one distribution transformer are further divided into 3 or further nearly isolated groups. IV. CONCLUSIONS PLC systems are one of the strongest and most commercial solutions for AMR systems especially regarding the running cost. Transformers do not provide sufficient coupling beyond 10 KHz and thus crating isolated LV Islands requiring another means for communication to reach a central station. In addition, the transformer construction with several feeders further divides the LV service area into three or four divisions. Loads (Home Appliances) interacts with any PLC system in two fold action. First, introduce time dependant access impedance and second providing time and frequency dependant noise. During peak loading, in some area lower impedance is seen shunting transmitters and more noise injected interfering receivers.

639

VI. BIOGRAPHIES
was born in Fayoum Egypt on Mars, 20, 1971. He received his B.Sc M.Sc. degrees in Electrical Engineering, Cairo University, Egypt in 1993 and 2004 respectively. He is now Technical Manager of the hardware department and a technical consultant of several projects developed in Engineering Office for Integrated Projects EOIP Since his graduation in 1993 he designed several Communication, Control and Data Acquisition systems He implemented several modems utilizing several modulation techniques (DSSS, M-FSK, MSK and ASK) in different frequency bands and channels Through his work, he introduced the second generation of EOIP Automatic Meter Reading (AMR) system through the Power Line Channel that reached the mass production level and covered about 60,000 customer premises in several governorates in Egypt. Eventually, this system resulted in 3 patents and fulfilled his M.Sc. degree in 2004.

Yasser Fathi

Tamer A. Kawady (M02) was born in Shebin El-kom, Egypt on Sept. 30, 1972. He received his B.Sc. (honors) and M.Sc. degrees in Electrical Engineering, Menoufiya University, Egypt, Ph.D. degree (excellent) from Technical University Darmstadt, Germany in 1995, 1999 and 2005 respectively. Dr. Kawady is currently an assistant professor at Menoufiya University, Egypt since April 2005. His interests are in digital protection, Power system simulation using the Electromagnetic Transient Program (EMTP) and Artificial Intelligence applications to power system protection.

Mohammed Z.Elgeziry was born in Elbagour, Egypt on Mars, 13, 1981. He received his B.Sc. degrees in Electrical Engineering, Menoufiya University, Egypt in 2002. He is working currently towards M.Sc. degree. His interests are in digital protection, signal processing, and power system automation.

Ahmed Hussien was born in Giza, Egypt on 1960. He received his B.Sc., M.Sc., and PhD. degrees in Electrical Engineering, Cairo University, Egypt in 1983, 1987 and 2002 respectively Dr. Ahmed is currently an assistant professor at Ciro University, Egypt. His interests are in digital signal processing FPGA utilization in signal processing in addition to Special Interest in Analogue circuit design

640