550

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 2, APRIL 2002

Safety Aspects of GSM Systems on High-Voltage Towers: An Experimental Analysis

J. B. M. van Waes, A. P. J. van Deursen, M. J. M. van Riet, and F. Provoost
AbstractLow-voltage (LV) applications mounted on high-voltage (HV) towers may pose a safety problem in the low voltage network when a flashover occurs over the high-voltage insulators. To study the effects experimentally, a 150-kV system was made available. On a high voltage tower carrying a cellular phone base station, one phase insulator was shorted to the tower. A current of about 200 A was injected into that phase at a large distance. We measured the current distribution and the relevant voltages in the low voltage net. The experimental current data agree with EMTP calculations. Appropriate measures to guarantee safety for persons and electronic equipment are discussed. Index TermsDistribution networks, electric shocks, electrical faults, EMC, global system for mobile communication (GSM), grounding, overvoltage, safety, transmission lines.

I. INTRODUCTION HE RECENT growth of cellular phone base stations lead to a quest for elevated locations for the base stations, particularly in rural areas. Fig. 1 shows a common solution, a cellular phone antenna placed on top of a high-voltage (HV) tower. Further on we will use the European acronym GSM for global system for mobile communication, originally groupe speciale mobile. Lightning may induce flashover of the HV insulators and a subsequent phase to ground fault [1]. The resulting power frequency current distributes over the lightning protection wires and the soil near the line via the tower grounding electrodes. A large fraction of the fault current flows through the low voltage cable toward the distribution transformer feeding the GSM system. At other customers served by this transformer, this current fraction may cause dangerous touch voltages. It may also overstress the insulation of the power inputs of electric apparatus, certainly when assisted by the lightning proper [2]. Nevertheless, power faults often affect a larger region than the higher frequency lightning stroke at the tower. The configuration of the HV network determines the fault current. In the Netherlands, many HV and MV networks have a ring layout. Fault currents over 20 kA may occur in the 380 kV grid; switch off times are of the order of 0.1 s. Skywires interconnect all HV towers; often two such wires are installed. The ladder network formed by the skywires and the grounding
Manuscript received April 14, 2000. J. B. M. van Waes was with the Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven, The Netherlands. He is now with Holland Railconsult, Utrecht, The Netherlands. A. P. J. van Deursen is with the Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven, The Netherlands. M. J. M. van Riet and F. Provoost are with B.V. NUON Technical Projects and Consultancy, Duiven, The Netherlands. Publisher Item Identifier S 0885-8977(02)02213-6.

Fig. 1. GSM in HV Tower.

of neighboring towers strongly reduces the current flowing into the ground at the faulted tower [3], [4]. Proper grounding measures near the footing ensure safe touch voltages at all towers. However, the presence of LV appliances and the voltage transfer into the LV network was not considered in the tower grounding design. All networks below 50 kV are equipped with buried cables in The Netherlands. Many older medium and low voltage (MV, LV) cables have lead shields. The good contact between these shields and the often damp soil provides a low impedance grounding [5]. Because of environmental concerns, new or replacement cables for MV and LV are shielded XLPE type with an outer insulating PE jacket. This increases the grounding impedance. A generally applicable safety approach was then requested which would also be effective for LV applications in HV towers on dry soil and with large fault currents. In order to study the effects at power frequency and to establish appropriate measures, we deliberately made a short circuit with smaller currents on a tower with a GSM base station. We measured the current distribution in a major part of the system, and several relevant voltages in the LV network. The measurements were compared with simulations. The lightning transients proper are not considered. II. DESCRIPTION OF THE SITE A rural site has been selected; no conductors of other parties were buried nearby. The measurements have been carried out

0885-8977/02$17.00 2002 IEEE

VAN WAES et al.: SAFETY ASPECTS OF GSM SYSTEMS ON HIGH-VOLTAGE TOWERS

551

Fig. 2. Overview measurement locations.

on the 10.3 km long double 150-kV link between Doetinchem (DTC) and Ulft. Fig. 2 gives an overview. At 8.3 km distance from DTC, tower 33 carries a GSM base station. In the substation Ulft, tower 40 is directly connected to the local grounding grid. As mentioned before, two lightning protection wires interconnect all towers. One of the 150-kV circuits was taken out of operation for the measurements. The injection current originated from another 10-kV network connected to the DTC substation. The MV networks are floating; a single phase to ground short circuit produces a current limited by the capacitances in the cables concerned. One of the MV phases at DTC was connected to a phase lead of the freed HV line. This phase could be shorted to tower 33 by a switch, thus simulating a controlled flashover with smaller currents. An actual fault would also be fed from Ulft. A 10-kV/400-V transformer at about 300 m distance from tower 33 feeds the GSM base station. The MV voltage cables have lead shields over their full length, with good contact to the soil. A grounding electrode near the transformer provides a parallel 2 to 5 to ground. The LV XLPE cable toward the base station also feeds several customers; two other LV cables (one shown in Fig. 2) serve further customers. A part of the LV cables has lead shields, which are always connected to the XLPE cable shields. The LV system has a TT grounding configuration. Because of the better performance with respect to safety and EMC, a TN-C or TN-S system is preferred in the NUON region of service nowadays. A number of current probes determined the relevant part of the current distribution at three positions along the HV line: at DTC, at tower 33 and at Ulft. Long Rogowski coils measured the current for large diameter grounded objects. Digital oscilloscopes registered the waveforms. The overall measurement accuracy of the data is about 5%. We also measured the voltage

between the neutral and a local ground electrode at locations LV1 and LV2 in Fig. 2. These voltages are related to the touch voltages when a TN system would have been applied. At LV2 we also measured the voltage between the neutral and the telecom ground; this voltage appears, for instance, over the isolation of a modem. In the past, cables from different utilities (telecom, CATV) often had separate grounding at customers premises. This was the case at the site of our investigations. For better EMC, one now tends to interconnect these ground leads. III. MEASUREMENTS The current distribution is given in Fig. 3, where all currents and voltages are normalized to 100-A injection. The actual injection current was 228 A and 1.2 kA. The current distribution was similar at both amplitudes. The small tables represent the measured currents (top) and the calculated values (bottom) to be discussed in Section IV. The limited time for the measurements did not allow to determine the current distribution at the MV/LV transformer. The skywires at the tower carry about 64% of the fault current. The current toward the ground at tower 33 includes all telecom cables. A total of 36% descends; more than half of that current leaves the tower footing through the LV cable connected to the base station toward the LV transformer. The measured voltage transfers were all of the same order of magnitude. At LV1 and LV2 we found 5.7 V and 4.7 V between the neutral and local ground. In addition, between neutral and telecom ground at LV2 5.0 V was observed. All voltages again are scaled to 100-A injection current. Although the ground impedance of substation Ulft is very low, we measured only little current into the soil. The distance to the faulted tower is about twice the decay length of a 50-Hz current in the ladder network formed by the skywires.

552

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 2, APRIL 2002

Fig. 3.

Measurement results (top) versus EMTP simulations (bottom) per 100 A fault current. TABLE I EMTP CALCULATED CURRENT DISTRIBUTION FOR VARIOUS TOWER FOOTING RESISTANCES (TFR) OF THE FAULTED TOWER 33 (I PER 100 A FAULT CURRENT)

IV. SIMULATIONS We analyzed the situation at the DTC-Ulft link by EMTP prior to the measurements. The latest known tower footing resistances were implemented; most data were obtained immediately after the construction of the 150-kV line. Typical values were of the order of 5 , for instance, 3.6 at tower 33. The telecom companies installed their own grounding electrodes with 2.5 or less; this impedance is in parallel with the tower grounding. Both skywires were regarded as separate conductors in the simulation. The analytical results based on Endreyni [4] compare well with the EMTP results. The skywire parallel to the faulted phase lead carried the larger part of the current, as shown in Fig. 3. We varied the effective grounding resistance at tower 33, and calculated the current entering the ground at that tower. Table I gives the results normalagainst results in ized to 100-A injection. A plot of a straight line. In a simple picture, one neglects the mutual coupling of the buried conductors and the phase shift. The equivalent source impedance (skywires in combination with tower footing impedances) at the tower footing is then about 0.4 . The maxabout 60%, in good agreement with imum current ) in which is the coupling factor [4]. The ob( of 36% corresponds to an of about 0.32 and a served implies that more tower voltage of 11.5 V per 100 A. This flows into the LV cable. The order of magthan two thirds of nitude of this impedance value is quite acceptable.

The 300-m LV cable between tower and transformer contributes a series impedance of about 0.25 at . The MV cables provide each about 1 to ground [5]. The smaller LV lead shields further reduce the impedance, but this is difficult to estimate. The measured voltages at the customers are lower than the values at the tower from the EMTP calculations, which can be readily understood from the voltage reduction of about 5 V per 100 A due to the LV cable connected to the tower. A second simulation has been carried for the 20-km-long 150-kV line AmersfoortKattenberg, which mainly runs over sandy soil. The short circuit current is 10 kA. Tower footing resistances are typically 40 , which leads to a tower fault voltage of about 7 kV. The high soil resistivity also causes high MV and LV grounding impedances. A 20% transferred voltage at customers would already be harmful for living beings and equipment alike. V. MEASURES CONSIDERED For practical reasons a solution is sought which provides safety in all situations, also for larger fault currents and higher tower footing resistances. As shown above, the grounding impedance of the LV/MV network reduces the risk and hazard. Note that a TT LV grounding approach seems to protect man better than a TN approach. However, when the power frequency voltage at the faulted tower becomes larger than 1.5 kV, which is readily possible on dry sand, the insulation

VAN WAES et al.: SAFETY ASPECTS OF GSM SYSTEMS ON HIGH-VOLTAGE TOWERS

553

and surge protection devices in electric equipment insulation is likely to be damaged. We summarize possible approaches. A. Select the Location GSM locations could be chosen based on the fault current amplitude. Peterson coils limit that current in the NUON 110-kV grid to an acceptable value of 300 A or smaller. However, many GSM systems have already been placed, also at towers in the 380-kV system with larger fault currents of 20 kA or more. To remove these systems now would be difficult. B. Increase the Series Impedance Here, one places the LV cable in a high permeability iron tube. The skineffect concentrates the current on the outside of the tube, which increases the series impedance and decreases the transfer impedance. Considerable lengths are required for large reductions. Assume, for instance, a steel tube with an outer radius of 5 cm, a wall thickness of 1 cm, and a relative permeability of 500. The voltage decreases to 36% after 600 to 900 m [6], [7]. Magnetic saturation, which might occur at high currents, reduces the series impedance and thereby also the protection. C. Add Electrodes Along the LV Cable Install a number of grounding electrodes at the shield of the LV cable between the MV/LV transformer and the tower. This reduces the hazard for the customers connected to other LV cables; those connected to the same LV cable as the GSM will benefit less. Attention should be paid to the number, length, and position of the electrodes [9]. Again, a considerable length of LV cable is required. D. Add Ground Electrodes at Nearby Towers The total grounding impedance at faulted tower is strongly determined by the lightning protection wires interconnecting the towers. The simulations gave an effective grounding impedance between 0.5 and 0.2 . The effect of additional electrodes is limited. E. Install an Isolation Transformer A LV to LV transformer is installed to isolate the HV and the MV/LV networks [10]. The insulation between primary and secondary should withstand power frequency voltages up to 10 kV. A surge arrestor limits the larger voltages over the insulation during the lightning strike. This arrestor should pass the part of the lightning current proper and must recover immediately thereafter. A varistor is then required rather than a spark gap. A combination of options b), c), and d) could provide a reliable protection. As already mentioned, a low source impedance is found at the tower base due to the HV network and, in particular, the skywires. It is then difficult to really change the current distribution with these options. Since local parameters as the soil resistivity [11] and other buried cables are important, each site should be considered separately. The long distances involved in option b) up to d) are seldom practical in The Netherlands. The tower 33 of this investigation belongs to case a). The maximum fault current there is 5 kA, which is switched off within 100 ms. This would give an acceptably low touch voltage, also with a TN grounding system: about 250 V. The low tower

footing resistances, the lead shielded cables, and the low soil resistivity, reduces the voltages at other LV customers to acceptably low values in this network. The lead shields of the MV and LV cables determine the safety, which could be compromised if these cable are replaced by XLPE. The only general solution is then e). A 20 kVA transformer has been ordered with the following requirements: 20 kV, 50 Hz insulation and 70 kV, 1.2/50 s. A varistor with a rated voltage of 15 kV, and 20-kA, 8/20- s pulse capability is available commercially. Elements of this varistor have been tested with five surges of 6.6 kA, 10/350 s at short intervals; no degradation of the knee voltage at 1 mA has been observed. The transformer and the varistor mounted in their cabinet was also tested. NUON will measure the knee voltage in maintenance checks at regular intervals. The XLPE LV cable extends to minimally 30 m up to the nearest distribution cabinet; the outer PE jacket withstands 20 kV. As precaution, a plastic insulating tube is placed around the first 6 m from the tower. An additional grounding electrode with a length of 5 m will be placed at the distribution cabinet for the lightning current part through the varistor. VI. CONCLUSION Simulations indicated that safety could be compromised by LV applications mounted in HV towers, in particular, cellular phone base stations. The model has been verified by measurements on an existing installation. The observed rate of lightning induced flashover is about five per year in the NUON 150- kV grid. The large number of GSM base stations already installed and still planned, the dense population, and the serious consequences of a flashover justify extensive safety measures. Faults on neighboring towers are equally dangerous as a lightning strike on the GSM tower proper. Several solutions are proposed. An isolation transformer combined with a suitable varistor will be imposed. ACKNOWLEDGMENT The authors wish to thank Y. Rajotte at Hydro Quebec for his valuable comments, the Dutch telecom companies KPN and Dutchtone for their intensive assistance in the project, G. A. P. Jacobs and W. F. J. Kersten for their assistance in the EMTP simulations, and their colleagues at NUON and TUE for their contributions to the preparation and measurements. REFERENCES
[1] L. A. Choy and M. Darveniza, A sensitivity analysis of lightning performance calculations for transmission lines, IEEE Trans. Power. App. Syst., vol. PAS-90, pp. 14431451, July 1971. [2] Audio, Video and Similar Electronic ApparatusSafety Requirements, IEC Std 60 065, Sept. 1998. [3] Y. Rajotte, J. Fortin, and G. Raymond, Impedance of multigrounded neutrals on rural distribution systems, IEEE Trans. Power Delivery, vol. 10, pp. 14531459, July 1995. [4] J. Endreyni, Analysis of transmission tower potentials during ground faults, IEEE Trans. Power App. Syst., vol. PAS-86, pp. 12741283, Oct. 1967. [5] G. Buse, Der bleimantel als erder, Elektrizitaetswirtsch, pp. 124127, 1959. [6] G. E. Bridges, Fields generated by bare and isolated cables buried in a lossy half-space, IEEE Trans. Geosci. Remote Sensing, vol. 30, pp. 140146, Jan. 1992.

554

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 17, NO. 2, APRIL 2002

[7] W. Koch, Erdungen in Wechselstromanlagen ueber 1 kV, Berlin: Springer-Verlag, 1961. [8] A. P. Meliopoulis and J.-F. Masson, Modeling and analysis of URD cable systems, IEEE Trans. Power Delivery, vol. 5, pp. 806815, Apr. 1990. [9] B. Thapar, V. Gerez, H. Kejriwal, and T. J. Kendrew, Two efficient configurations of grounding electrodes for electric distribution systems, IEEE Trans. Power Delivery, vol. 9, pp. 11081114, Apr. 1994. [10] H. Zeuschel and P. Hoiss, Niederspannungsversorgung von mobilfunkstationen an hochspanningsmasten, EVU-Betriebspraxis 12/98, vol. 12, pp. 812, Dec. 1998. [11] R. J. Gustafson, R. Pursly, and V. D. Albertson, Seasonal grounding resistance variations on distribution systems, IEEE Trans. Power Delivery, vol. 5, pp. 10131018, Apr. 1990.

M. J. M. van Riet was born in Vessem, The Netherlands, in 1954. He received the M.Sc. degree in electrical engineering at the Eindhoven University of Technology, Eindhoven, The Netherlands, in 1979. In 1979, he started his career at PGEM, now called B.V. NUON Technical Projects and Consultancy, Duiven, The Netherlands. During this period, he was Manager of large projects such as SF6 150 kV/10 kV substations, 50- and 150-kV cable projects, and, recently, the new Command Centre at NUON. Currently, he is Head of the Research and Development Department at NUON. His research interests cover all aspects of power engineering, especially high-voltage cables, electrical components, transformers, power switches, and SF6-installations. Besides his technical interests, he is also responsible for the scouting and training of young engineers.

J. B. M. van Waes was born in Breda, The Netherlands, in 1973. He received the M.Sc. degree from the Eindhoven University of Technology (EUT), Eindhoven, The Netherlands, in 1996. He is currently pursuing the Ph.D. degree in electrical engineering at EUT. After graduation, he codeveloped a multimedia course on EMC, and investigated various grounding approaches for industry and power companies. Currently, he works for Holland Railconsult, Utrecht, The Netherlands, on the introduction of 25-kV traction in the Dutch railways. His thesis research at Eindhoven University also dealt with grounding of HV and MV substations, the consequences of MV faults for LV customers, and of the replacement of PILC cables by XLPE types.

F. Provoost was born in Domburg, The Netherlands, on March 4, 1957. He received the B.Sc. degree in electrical engineering from the University of Eindhoven, Eindhoven, The Netherlands, in 1982. From 1982 to 1985, he worked at ASEA, Ludvika, Sweden, on main circuit design of HVDC systems. In 1986, he joined PGEM, an energy utility in the Netherlands, which, after several merges, became B.V. NUON Technical Projects and Consultancy, Duiven, The Netherlands. Here, he has worked on many projects and studies on high- and low-voltage network systems, power quality, and grounding.

A. P. J. van Deursen received the B.S. degree in physics and the Ph.D. degree in weakly bound atomic and molecular complexes from Nijmegen Catholic University (NCU), Nijmegen, The Netherlands, in 1965 and 1976, respectively. In 1976, he investigated the electronic structure of metals and semiconductors by means of very high magnetic fields, continuous and pulsed, first at the Max-Planck Institute, Grenoble, France, and then at NCU. Since 1986, he has been a Lecturer on electromagnetic compatibility at Eindhoven University of Technology, Eindhoven, The Netherlands. His interests range from protection of microelectronics to high-voltage substations.