Characterisation, Measurement and Modelling of Medium Voltage Power-Line Cables for High Data Rate Communication

Transcription

1 Characterisation, Measurement and Modelling of Medium Voltage Power-Line Cables for High Data Rate Communication Christian Hensen, Dr. Wolfgang Schulz, Sascha Schwane University of Paderborn ', : Department of Communications Engineering D Paderborn Enno Borchers, Dr. Georg Dickmann ke Kommunikations-Elektronik GmbH. R&D dep. TE 123, P.O. Box 3246 D Hannover Abstract Up to now high data rate power-line communication has been focused on the low voltage network []. This study examines the transmission line characteristics of medium voltage power-line cables with regard to high data rate communication. Several measurements were carried out in order to assess the characteristics of some cables. Both the primary and secondary transmission line parameters were determined. The measurement results were used to develop a passive model of a selected type of medium voltage power-line cable. Calculations showed good agreement between the model and the measurements. However, a different model had to be developed for hardware implementation. ntroduction Driven by both the liberalisation of the telecommunication markets and the availability of sophisticated modulation methods, power-line communication is becoming an increasingly popular field. Up to now most of the studies focused on the low voltage power-line network with the aim of covering the so-called 'last mile'. However low-speed data transmission over the medium or high voltage sections has a long tradition. Tlus paper deals with the transmission characteristics of medium voltage cables which appear to be promising for broadband communication. n Germany these cables cover the voltage range from 10 kv to 30 kv. The characterisation of the channel with regard to communication is one of the most important tasks. The network topology, the different types of cable and their basic characteristics have to be taken into account. Contrary to the low voltage power-line network the different types of cable as well as the topology are recorded by the power supply utilities. Tlus results in a potential advantage. A large variety of cables exists differing in the general structure, the number of cores, the material of the conductor and the used insulation [2]. Only single-core cables were examined in this study. The crucial 37

2 . _ -. properties attenuation coefficient, phase coefficient and c on the knowledge of these transmission line parameters the distribution of resistance, inductance, conductance and capacitance along the cable could be calculated. The examined cables have a coaxial structure. Therefore the measured results could be compared to those calculated with the known formulas for coaxial lines. The measurements yielded results allowing to assess certain types of cable as well as hinting at some basic properties the cables have in common. The measurements form a basis for the development of a pqsive model of a selected type of cable. The model consists of a chain of identical equivalent circuits each representing a short length of cable. The nmximum length which can be modelled with one equivalent circuit is determined by the maximuh'freq~enc~. The circuits were designed as extended pi-sections composed of resistors, capacitors and inductors. Special care was taken to reflect the skin-effect inherent in the examined cables. Calculated results showed that the developed model reflects all the important characteristics of the cable. Eventually it turned out that the hardware implementation of this passive model would require components that are hardly realisable. Nevertheless this restriction could be surmounted by reduced demands. Medium voltage power-line network Tlie transport of electrical energy from tlie power station to the consumer is commonly performed at three different voltage levels. The medium voltage power-line network represents the connection between the high voltage network and the low voltage network [3]. The high voltage network covers the voltage range from 110 kv to 380 kv. t is used for the transmission of energy over long distances between the power stations and the high voltage transformer substations close to urban areas. The medium voltage network is used for energy distribution between the high voltage transformer substations and tl~e medium voltage transformer substations located close to the consumers. Here the voltage is reduced to 0.4 kv and distributed to the households. Many consumers, most of tliern industrial, are directly connected to the medium voltage network. Both underground cables and overhead cables are used in the medium voltage network. The underground cables can be divided in several types of singlecore and threecore cables. The most common network topology is a radial net or an open ring net (Figure ). The former has tlie advantage of easier installation, the latter offers a higher reliability. n urban areas the typical distance between two medium voltage transformer substations lies between several hundred and one thousand metres. n this study only single-core cables were examined (Figure 2). Here transmission between conductor and screen is considered as opposed to other modes using conductors or screens of two or more cables in parallel. The cables possess a coaxial structure. Their general construction from the inside to the outside is as follows. The core of the cable consists of one or several copper or aluminium wires with cross-sectional areas between 35 mm2 and 500 mm2. The core is surrounded by a conductor screen of semi-conducting material, which is used to limit the electrical field, followed by the insulation used for separating the conductor from tlie screen. Tlie screen consists of semi-conducting material and copper wires. t is connected to ground and carries charging current, leakage current and ground fault cqrent. The outermost layer is a sheath for protection against mechanical stresses and corrosion.

3 sheath screen insulation conductor Figure 1: Open ring net Figure 2: Single-core cable Because of the coaxial structure of the single-core cables, the known formulas for coaxial cables can be applied to calculate the primary transmission line parameters. The resistance per unit length / R'= - + P2 n-d.t, n.d.tz (1) can be calculated using the diameter D of the outer conductor, the diameter d of the inner conductor and the respective specific resistivities pi and penetration depths ti which are calculated from with the angular frequency a, and tlie magnetic constant PO. The conductance per unit length equals O'E0 '6, G'=2.n. D ln- d tans with tl~e permittivity of free space EO, the relative permittivity 6,,and the dielectric loss angle tans.. The inductance per unit length equals and the capacitance per unit length is

4 The secondary transmission line parameters characteristic be derived from the primary transmission line parameters using 11 Can and A first estimation of the absolute value of the characteristic impedance is given by Measurements Several types of medium voltage power-line cables were measwed in order to determine their basic properties. All measurements were carried out with a network analyser with an input/output impedance of 59 n. The input impedances of open-ended and short-circuited cables were measured. The,theoretical input impedance of an open-ended transmission line with the length 1 is and that of a short-circuited is given by f both input impedances are known, the characteristic impedance can be calculated with and the propagation coefficient is derived from The primary transmission line parameters can be calculated with The real part of the propagation coefficient, namely the attenuation coefficient, and its imaginary part, the phase coefficient, can in general not be measured directly as a transfer function (or scattering parameter

5 pedance mistnatcll between the network analys inismatch results in reflections yielding distorted measurements. Thus problem can only be overcome if the cliaracteristic impedance is measured first and according impedance transformers are employed to match the inputloutput impedance of the network analyser to the characteristic impedance of the device under test. Measurement results The measurement results in this paper were obtained measuring the medium voltage cable. ABB NA2XS(l?L,)2Y 1*120RM/16 12/20 kv. Tlie~tiarking of the cable denotes a singlecore cable f0r.a nomind'voltage of 12 kv between conductor and ground and 20 kv between two conductors. Tlie centre conductor is concentric consisting of several wires. t is made of aluminium and offers a cross-sectional area of 120 mm2. Tlie cable is insulated wi& p6lyethylene and lias a screen made of copper with a cross-sectional area of 16 mm2. The outer sheath is again made of polyethylene. 'Figure 3,shows the primary transmission line parameters,.namely the resistance per unit length R', the conductance'per unit length G, the inductance per unit length L' and the capacitance per unit length C' in. the frequency range up to 10 MHz. \ _._----~-----* ' ' f/mhz -> f m -> 0.30 : : : :! :! : 0.30 : : : : : : : : :,,,,, a,,, V ,--.-:,,----:,..., L : : -----:... -:-...:------: -----: ----:...,, ',,,,,,,,,,, a ~,,, 0.26.k.i...;... i... iiiiiijjjjjjiiiiiiiiiiiiiiiiiiii i... i -----:... i...i... *,,,,, '.,,,.,.. a *, < ~, ,,,,..., ,...,----: i.----:...:.----: ---.-: -----; -.---:.----: ----;...,.,,,,., i -----;----- ;---:,- - --, : :---- -C ,,,,,,,,, : ;-..---;... i...-i----: -----; ---.-;... E, #,,,,,, m,.,,,,,, C ; :... L -----: -----:.-... L -----: ---.-:...,,,,,,,,, =. (,,,,,,, m,, V a,,,,,, yo 18...A----:-----: ; > :...:...:.----;... L ; -----:......, a, 8, &,,, 0 1 t-2-, ,...,,*...,...,...,...*...,...,,.,,,..., U., 1 t,, ' ----: -----; -----; -----'-----: -----; --.--; : :; ;.----; -----: -----; -----: : -.--;,,,...;...;...:...;...;.,...;... i 4...;...,,,,,,,,, : iiiiiii... { :-... L j...!.----; j..-..,,,,,, 1 ~,, ; -----; -----;...: ;...; -----; -----:... S, 1 0,, ;.----; -----: -----;... > ---.-: : : : : ; : 0.10 i i, : : : : : : :, ~,,,,, ' f/mhz -> f/mhz -> Figure 3: Primary transmission line parameters; solid: measured; dotted: calculated The measured results are compared to the calculated results using the aforementioned formulas for coaxial cables. The used parameters were d=12,5 mm, 0=26,5 mm, tan6 =0.0006, &,. =2.4, pl= mm2/m and pz = L2. mn2/m.

6 stance per unit length is proportional t rope@ matches the calculation for a coaxial cable. The measured values are slightly higher than those yielded by the formula for a coaxial cable. This is probably due to the sek-conducting zone around the conductor which starts carrying current with increasing skin-effect. The measured conductance grows, linearly with frequency as expected but tlle gradient is much higher than calculated. Since the dielectric material is ogtimised for the network frequency of 50 Hz, dielectric loss at higher frequencies is probably at the origin of this deviation. Both the measured inductance and capacitance are almost constant in tlie examined frequency range and offer fairly good agreement with the calculated parameters. The measurement results show that the primary transmission line parameters of single-core medium voltage power-line cables differ only marginally from those calculated for coaxial cables. The secondary transmission line parameters of the measured cable show characteristics that are typical of coaxial cables. The attenuation coefficient increases with frequency. Therefore, the cables exhibit a low pass characteristic. The phase coefficient is linear resulting in constant group delay. The characteristic impedance is almost constant over frequency with higher values at low frequencies. Development of a passive model A passive model of the above mentioned cable was developed employing the measurement results. The modelling is based on the chain matrices of a transmission line and a pi-section. The comparison of the coefficients yields expressions for the imp6dances which have to be used in the pi-section for it to have the sane behaviour as a transtnission line. The expressions contain hyperbolic functions whose argument is the propagation coefficient multiplied with the length of the transmission line. These hyperbolic functions are estimated by Taylor-series which are cut off after the first link. This fact requires that the absolute value of tlle argument of the hyperbolic function be smaller than one. Therefore, the maximum length of transmission line which can be modelled with one pi-section for a certain frequency is limited. Physically this is based on the representation of distributions as lumped elements. Greater lengths can be modelled by cascading identical pisections. The components used in the basic pi-section (Figure 4) are directly related to the measured primary transmission line parameters. The basic pi-section can only show the desired behaviour for a single frequency. Both the conductance and tlie resistance per unit length are frequency dependant. At relatively high frequencies the conductance per unit length grows linearly with frequency due to increasing dielectric loss. The resistance per urut lengtli grows with frequency due to the skin-effect. t is proportional to the square root of the frequency. The conductance distribution can often be neglected when compared to the capacitance distribution and need not be modelled in many cases. However, the impact of the skin-effect has to be modelled in most cases. One way of modelling the skin-effect was proposed by Yen et al.[4]. The resistance in the series leg of the pi-section is replaced by several legs composed of resistors and inductances. At low frequencies the current flows tluough all the resistors in parallel resulting in a low total resistance. At lugher frequencies the resistance of the inductors increases yielding a lugher total resistance. The values of the resistors and inductors depend on the geometry of the cable and on tlie number of legs. The higher the number of legs, the lugher the maximum frequency at which the model shows tlie desired behaviour. The resulting extended pi-section neglecting the conductance per unit length is shown in Figure 5.

7 Figure 4: Basic pi-section Figure 5: Extended pi-secti'on The component values obtained by tlie modelling technique described above were used to calculate the secondary transmission line parameters of the extended pi-section. n the frequency range up to 10 MHz the calculations showed good agreement with the measurement results. Nevertheless, certain aspects prevented a hardware implementation. The low DC resistance (0,23 Olunflun) of the cable resulted in extremely low values for the resistors. These values could not be implemented as lumped parts. Besides, the DC resistances of tlie necessary inductors are larger than the resistances of the resistors. Furthermore, the conductance per unit length of tlus cable is not to be neglected when compared to the capacitance per unit length. That is why the calculated attenuation coefficient is too low. The modelling of a conductance linearly dependent on frequency with lumped parts was not solved successfidly. However, a modification of the parts in the series leg would yield the desired result. A further drawback of this model is the necessary number of parts for a large cable length. Because of the lug11 propagation coefficient only very small lengths (2 m for a maximum frequency of 10 MHz) of cable can be modelled with one extended pi-section. A different approach was based on the attempt to model the secondary transmission line coefficients rather than starting witli the primary parameters. Therefore tlie values of the primary parameters are not needed and do not appear in the model. Because of the huge problems encountered in tlie first model, die new approach was intended to meet reduced demands. The major aim was to reflect the attenuation coefficient of the cable as well as possible in a reduced frequency range of up to 2 MHz. A different pi-section with a modelling of the skineffect distinct from the one used in the former circuit was developed (Figure 6). Figure 6: Extended pi-section with different modelling of skin-effect t proved to be realisable in hardware witli partsithat were easily obtainable. Besides, the length of cable that could be modelled with one pi-section was raised. This was not only due to tlie lower maximum frequency but also to the reduced demands. Eventually a circuit was devised that approximately reflected the attenuation coefficient of the cable. t offered a phase coefficient growing linearly with frequency. However, the phase increment of the model is significantly lower than that of the cable. The characteristic impedance was almost linear but much too lugli. Tlus problem could be solved by employing impedance transformers witli low loss. Figure 7 shows the resulting attenuation and the characteristic impedance in tlie frequency range up to 2 MHz compared to the measurement results for the above mentioned power-line cable.

8 Figure 7: Measurement results of medium voltage power-line cable (solid) and its passive model (dotted) As can be seen, the attenuation of the model offers good agreement with the measurement. The cliaracteristic impedance of the power-line cable is matched for frequencies higher than 400 khz but is modelled slightly to lug11 in the frequency range below. The measurement results show that the developed model meets the desired goals. Conclusions Tlus paper deals with the properties of medium voltage power-line cables with regard to high data rate communication. Measurement results for a selected type of single-core cable were presented. The measured primary transmission line parameters were employed in the development of a first passive model. This model should reflect all cable properties in the frequency range up to 10 MHz and proved to do so in calculations. As the first model could not be implemented in hardware, a different approach had to be taken. The second model was based on the measured secondary transmission line parameters and had to meet reduced demands. The new model could be realised in hardware and was able to reflect the attenuation coefficient and the characteristic impedance of the selected cable in the frequency range up to 2 MHz. References [l] Cllr. Hensen,,,Niederspannungsnetze als Medium zur Dateniibertragung", TG-Fachtagung Auf dem Weg zur modernen nformations-nfrastruktur, 1997, TG-Faclibericht 141, VDE-Verlag, S [2] E.W.G. Bungay, D. McAllister,,,Electric Cables Handbooku, BSP Professional Books, Oxford, 1990 [3] L. Heinhold,,,Kabel und Leitungen fiir Starkstrom", Siemens AG Berlin, 1987 [4] C. Yen et al..,,,time-domain Skin-Effect Model for Transient Analysis of Lossy Transmission ' Lines", Proceedings of the EEE, Vol. 70, No. 7, July 1982