Sie sind auf Seite 1von 6

ID 63

Central or Local Compensation of Earth-Fault Currents in Non-Effectively Earthed Distribution Systems


A. Guldbrand, O. Samuelsson, Member, IEEE

AbstractCentralized Petersen coil earthing in networks with extensive use of underground cable involves flow of large quantities of reactive earth fault current. In networks with long radial cable feeders this creates problems involving active losses and active and reactive earth fault current. One possible solution to the problem is the use of local compensation. This paper aims at understanding the limitations of traditionally compensated networks and compares these to networks that are locally compensated. This is done through analytical calculations and PSCAD simulations that are combined with results from a project studying cables in Swedish distribution networks. In addition, basic PSCAD simulations are used to clarify how the rating and placing of the distributed compensation units affects active earth fault current and investment costs Index TermsFault current, Inductor compensated distribution lines, Power cables, Power distribution planning, Power transformers

I. INTRODUCTION

N January 2005 the storm Gudrun hit the south of Sweden. The storm reached wind speeds of about 45 meters per second and caused large outages in the Swedish power distribution systems. More than 20 000 kilometres of distribution lines were damaged during the storm. Networks consisting of a large part of underground cables were not as badly affected by the storm as distribution networks with a large amount of overhead lines [1]. To increase the incentive of network owners to build more reliable power systems, the Swedish parliament in December 2005 passed a law regulating costumer compensation at power outages. The law entitles consumers compensations up to 300 % of their estimated yearly network tariff. As a result of the Gudrun experience and forced by the new compensation regulations, several distribution network owners plan to, during a near future, replace large amounts of rural overhead lines by underground cables to increase the distribution network reliability. Before the storm Gudrun, the replacing of

distribution overhead lines by cable was a part of the long term planning. After the storm the pace in which the overhead lines are replaced has considerably increased. The resulting new rural networks will have longer underground cable radials than urban cable networks. This will bring new challenges to the distribution companies. The high pace in which the replacement of the overhead lines now takes place, creates an urgent demand for practical advise on how to use the, not yet well-established, distributed earth fault compensation technique. In this paper results from an ELFORSK project [2] are combined with analytical calculations and a few additional simulations to help understand the limitations of traditionally compensated networks and to supply more hands-on guidelines for when to consider distributed compensation. In addition basic PSCAD simulations are used to clarify how the rating and placing of the distributed compensation units affects active earth fault current, earth fault detection and investment costs. The Swedish distribution companies might be the first to experience this extensive change in rural distribution network design. Similar changes are however likely to take place in other European countries, since the authorities tend to be more and more restrictive in allowing overhead lines in distribution networks. II. NETWORKS WITH LONG CABLE FEEDERS Two factors limit the possible size of a cable network: First, the minimum active earth fault current must be properly directed and large enough to ensure correct function of the protection relays. This sometimes requires a neutral point resistor. Second, the maximum earth fault current may not cause the voltage across the earthing resistance to exceed the regulated value (100 V in Swedish networks). An extensive use of underground cable significantly increases the capacitive coupling of the network to earth, and as a consequence the capacitive earth fault current. In urban areas with high load density and short radial cable feeders the problem with extensive capacitive earth fault current is often solved by connecting a Petersen coil to the HV (High Voltage) to MV (Medium Voltage) transformer neutral. The coil creates a compensating inductive current. Centralised Petersen coil earthing in networks with

This work was financed in part by ELFORSK. Current project on the subject of system earthing of distribution networks is financed by ELEKTRA A. Guldbrand and O. Samuelsson are with the Department of Industrial Electrical Engineering and Automation, Lund University, P.O. Box 118, SE221 00 Lund, Sweden (e-mail: anna.guldbrand@iea.lth.se, olof.samuelsson@iea.lth.se)

ID 63 extensive use of underground cable involves flow of large quantities of reactive earth fault current. In networks with long radial cable feeders, consequences of this are active losses and active and reactive earth fault current. Fig. 1 illustrates how the active earth fault current contribution drastically increases with the length of the cable radial in centrally compensated 12 and 24 kV systems [2] [3].
LV

MV HV system

LV

Active Fault Current Contribution (A)

30 25 20 15 10 5 0 0 20 30 40 50 60 12 kV 24 kV

LV Fig. 3 System with central and local compensation

Radial Length (km) Fig. 1 Active earth fault contribution of 20, 30, 40, 50 and 60 km cable radials in a 12 and 24 kV system.

Neutral point reactors are not ideal but have losses that contribute to the active earth fault current. The active earth fault current due to losses in the Petersen coils are approximately 0.4 A and 0.6 A per 10 km of compensated cable, for 12 kV and 24 kV systems. In traditional overhead line systems, active earth fault current associated with Petersen coil losses and reactive current flow are small, regardless of network size. In contrast to this, in networks with long cable radials the earth fault current has an active part that will depend on the size of the network and on the fault location. III. DISTRIBUTED EARTH FAULT COMPENSATION A. Advantages of Distributed Compensation One possible solution to the problem with large active fault current is the use of distributed compensation i.e. use both central and local compensation coils. Fig. 2and Fig. 3 schematically illustrate a network with central and central and local compensation.
MV HV system LV

The distributed reactors are not any more efficient than those used for central compensation. The reactive current flow and the related losses are however reduced. Consequently the maximum size of a network using distributed compensation will not depend so much on the reactive current transportation but on the non-ideal central and distributed reactors. This opens for the possibility to use longer cable radials. There are additional advantages related to the use of distributed compensation. If the compensation coils are properly located with respect to the disconnectors, disconnection of part of the network will include disconnection of corresponding amount of compensation. Further, locally compensated networks can loose a compensation coil without it having devastating consequences to the network performance, as it would in a centrally compensated network. B. The ELFORSK Project Swedish distribution network owners have financed a project aiming at develop guidelines for how to use distributed Petersen compensation in distribution networks with significant amount of power cable [2]. The project was organized by ELFORSK, during the spring of 2006. The Elforsk project report is based on a large number of PSCAD simulations. Due to the complexity of the problem it has proven difficult to present any general recommendations concerning distributed compensation, based on the simulations. IV. DISTRIBUTED COMPENSATION COIL RATING One important aspect of distributed compensation is the choice of the most favorable rating and placing of the compensation units. A compensating unit is a MV to LV transformer with the MV side neutral point connected to earth via a compensation coil. The total amount of inductive current required in a specific system can be induced in an arbitrary number of distributed Petersen coils. It is reasonable to assume that the losses due to reactive current transportation increase with the distance between the units. On the other hand, a larger number of units gives higher investment and maintenance costs. Fig. 4 illustrates the constraints that decide the most advantageous distance between the compensation coils in a network with distributed earth fault compensation.

LV

LV Fig. 2 System with central compensation

ID 63 rating. This is necessary to verify in lab or tests on location.

Cost

Maximum distance due to maximum current through compensation coil

Total cost

Investment cost Losses Optimum

Fig. 5 Schematic illustration of three-legged ZNdyn transformer Distance between compensation coils Fig. 4. Constraints that decide the most advantageous distance between compensation coils in network with distributed earth fault compensation

To limit the investment costs, Petersen coil earthed transformers can be placed in already existing transformer stations. The price of one single Petersen coil earthed MV to LV transformer exceeds the price of two unearthed transformer. The price difference of the entire substation might, assuming a 1 to 4 conventional transformer to substation cost ratio, differ with a 5 to 4 ratio. To study how the losses in a power system, associated with reactive current transportation, change when the distance between the compensation coils increases, the active earth fault current has been simulated using PSCAD. The simulations show that the losses do not change much as long as the distance between the compensation coils does not exceed approximately 20 km. If only the costs of investments and losses would decide the size of the compensation coils, the distance between the coils should therefore be as long as 20 km. There is however a third constraint that further limits the largest possible compensation coil size. The use of distributed compensation is of most interest in rural distribution networks. In these networks the size of the MV to LV (Low Voltage) transformer is often limited to 100 kVA. The compensation current through the transformers is proportional to the distance between the coils. The transformer rating will therefore set a maximum limit for the distance between the compensation coils. Transfix ECOBLOC transformer is a new product designed to be used in locally compensated distribution systems [5]. To evaluate the performance of a ZNdyn transformer like ECOBLOC, a number of different simulations have been run. The three-legged ZNdyn transformer has four concentric windings on each leg. The windings are schematically illustrated in Fig. 5. Simulations using a ZNdyn transformer model with each core modelled individually indicate that the load on the LV side does not influence the zero sequence impedance of the transformer. Hence, the MV side zero sequence current will not under any load condition be transformed to the LV side. The function of the LV side fuses has therefore not to be taken into consideration when selecting the compensation coil

The MV side fuses are however a limiting factor for the compensation coil rating. To keep the MMF (Magnetomotive Force) balance of the transformers during a fault in the MV network, the induced compensation current will divide equally between the three MV phases. To ensure correct function of the fuses protecting the MV to LV transformer the sum of the compensation current and the load current must always be sufficiently much lower than the lowest short-circuit current that the fuses are dimensioned for. The maximum compensation current through the transformers will be reached with a shorter distance between the coils than 20 km, and is consequently the factor that limits the distance between the compensation units. V. ZNDYN TRANSFORMER AND EARTHING Table I gives the data of a ZNdyn transformer much like the Transfix ECOBLOC transformer bought by Swedish distribution network owners [5] [6].
TABLE I DATA OF ZNdyn TRANSFORMER Rated power Short circuit impedance Primary voltage Secondary voltage No load losses Load losses Zero sequence impedance, earthing reactor and transformer 100 kVA 0.04 pu 22 kV 0.4 kV 0.0023 pu 0.015 pu 0.005 + j 0.175 pu

A. Fault Current Calculations A transformer model based on the data in table I has been used to simulate the MV side fault currents in case of a bolted three, two or single phase LV fault. The simulated fault is located close to the transformer. The fault current will therefore depend solely on the impedance of the transformer. The simulated fault currents are listed in Table II. IA, IB and IC are the MV side currents. Ia, Ib and Ic are the LV side currents.

ID 63 phase angle.
TABLE II FAULT CURRENTS FOR THREE, TWO AND SINGLE PHASE FAULT Three-phase short circuit IA IB IC Ia Ib Ic 58.7 A 58.7 A 58.7 A 3228.5 A 3228.5 A 3228.5 A Two-phase short circuit 29.33 A 58.65 A 29.33 A 2793 A 2793 A 0 Single phase to earth 31.66 A 31.66 A 0 3015 A 0 0

I Max = I Max Load + I L 3 = S n I Max = 100 10


3 3

3U n + I L 3
(1)

3 22 10 + I L 3

I Max = 2.6 + I L 3
The capacitive earth fault current in a 22 kV network consisting of underground cable is approximately 3 A per km, i.e. the compensation current is 1 A per phase and km. If the fuses are selected to withstand currents up to 20 A the distance between the Petersen coils could be as long as 20-3 km. In case of an 11 kV distribution network the possible distance is even longer. Some major Swedish distribution network owners plan to use compensation coils of size 10 A in 12 kV systems and 15 A in 24 kV systems, which corresponds to compensation every 5 km in both 12 and 24 kV system. One possible reason these sizes have been decided on is reluctance from manufacturers to develop transformers with several different compensation coil sizes. The pressed time limit has probably prevented thorough evaluation of the most economical advantageous coil size. The network companies have had to decide for a reasonable size, which at least ensures that the technical constraints, as maximum compensation current through the MV to LV transformers, are fulfilled. The distribution network owners might find it advantageous to keep the same distance between the compensation units in both the 12 kV and 24 kV systems. In addition, with a distance of approximately 5 km between the units instead of 15 or 20, it might be possible to make more use of the fact that disconnection of part of the network will include disconnection of corresponding amount of compensation. VI. GUIDELINES A. Results of Analytical Calculations Approximate values of the possible size of a cable network can be analytically calculated. Despite being rough estimates, the calculations are helpful to understand the influence of local compensation compared to central compensation. Results of analytical calculations expressed as maximum radial lengths are shown in Fig. 8 and Fig. 9. The calculations are based on equipment efficiency and earth fault current contributions according to [2] and [4]. In case of only one single long cable radial a neutral resistance is necessary to ensure correct function of the directional residual over current relay. In case of three cable radials of the same lengths a neutral resistance is not considered necessary. The maximum earth fault current and hence the radial length is limited by the size of the earthing resistance. Considering Swedish conditions the earthing resistance is taken to be approximately 7 ohm, i.e. the maximal fault current is set to approximately 14 A.

The three-phase short circuit current is symmetrical and will be transformed according to the ratio of the transformer. A two-phase LV short circuit will induce flux in two of the transformer legs; consequently, because of the Z winding, all three MV side phases will carry fault current. One of the phases will however experience flux through both the Z windings and carry a current that is twice of the current carried by each of the other two phases. The current phasors are illustrated in Fig. 6.

Fig. 6 Current phasors illustrating the phase currents in case of a two-phase fault on the LV side.

In case of a single phase to earth fault the delta winding of the ZNdyn transformer will balance the one phase fault current of the LV side and hence induce flux in all three transformer legs. The per unit value of the delta current is onethird of the LV fault current. To keep MMF balance, two of the MV side phases carry fault current that, in pu value, equals the delta current. The current phasors are illustrated in Fig. 7.

Fig. 7 Current phasors illustrating the phase currents in case of a single phase to earth fault on the LV side.

A single phase to earth fault on the LV side will not result in any current through the earthing equipment on the MV side of the transformer. This is one of the main advantages of the ZNdyn transformer. The presence of a fault resistance will influence all the fault currents. B. Maximum Load and Compensation Current The maximum MV side combination of load current and compensation current in each of the three phases in a 22 kV network is calculated in (1). The maximum current is obtained if the load current and compensation current have the same

ID 63

140

80 120 110 70

Maximum Radial Length (km)

120 100

Fault Current (A)

60 50 40 30 20 10 0

Central Local

Central
80 60 40 40 20 0 12 24 30

Local

20

30

40

50

60

70

80

System Voltage (kV) Fig. 8 Analytically calculated maximum length of single radial, in only central and in central and local compensated systems
60

Radial Length (km) Fig. 10 Simulated earth fault current for different lengths of single radial, central and distributed compensation
80

Maximum Radial Length (km)

50 50 40 30 30 20 10 0 12

Central
35 25

70

Fault Current (A)

Local

60 50 40 30 20 10 0

Central Local

24

20

25

30

35

40

45

System Voltage (kV) Fig. 9 Analytically calculated maximum length of three radials of equal length, in only central and in central and local compensated systems

Radial Length (km) Fig. 11 Simulated earth fault current for different lengths of three radials of equal length, central and distributed compensation

B. Results of Simulations During a bolted single phase to earth fault there will be reactive current and hence real losses in the phase conductors regardless of perfectly tuned Petersen coils and solely active current at the fault location. During a single phase to earth fault in the MV network the capacitive currents of the healthy phases are shifted 120 degrees relative each other. The sum of the capacitive currents is of the same size as the inductive current. The inductive current will however divide equally between the three phases. Consequently, in the individual phases the inductive current will not fully compensate the capacitive current. The simulated earth fault current differs from the analytically calculated fault current, where the inductive and capacitive currents are assumed to compensate each other in all three phases. The earth fault currents simulated in only centrally and in locally and centrally compensated 22 kV systems are illustrated in Fig. 10 and Fig. 11. Fig. 10 shows the fault current in a system consisting of one long cable radial and one short overhead line. This system has a 10 A neutral resistor. Fig. 11 shows the fault current in a system consisting of three cables of equal length. This system does not have a neutral point resistor. The earth fault currents have been simulated with tuned central Petersen coils.

VII. FAULT DETECTION Distributed compensation is likely to influence the function of directional over current relays. The angle measured by the relays indicates where the fault is located. A 22 kV system consisting of a 50 km cable radial and a 10 km overhead line has been modelled to investigate how the distance between compensation units might influence the angle measured by the relays. Apart from the influence of the distance between the compensation units, relay protection is not studied in this paper. A thorough analysis of the relay protection will be necessary when planning distributed compensation. Simulated values of the angle between the zero sequence voltage and the zero sequence current of the cable radial close to the bus bar is used to evaluate the influence of the distance between the compensation units. Fig. 12 shows this angle, measured on the cable radial close to the bus bar as single phase to earth faults are simulated at different locations of the system. Faults are modelled to take place at the end of the cable radial 50 km from the bus bar, on the cable radial 20 from the bus bar and at the end of the overhead line 10 km from the bus bar.

ID 63 research have been most useful.


120 100

X. REFERENCE
[1] Johansson J., Lindahl S., Samuelsson O. & Ottosson H.: The storm Gudrun A seven-weeks power outage in Sweden CRIS, Third Conference on Critical Infrastructures, 2006 Ntkonsekvenser vid kablifiering av luftledningsnt, (Consequences to the distribution network when replacing overhead lines by underground cable), (In Swedish), Project report, ELFORSK AB, Stockholm, 2006 Norberg P.: Kompensering av lnga kablar, (Compensation of long cable radials), (In Swedish), Notes, 2006 Lindahl S., Messing L., stlund S., Olsson B. & Petterson A.: Knsliga jordfelsskydd. Bortkoppling av hgresistiva jordslutningar i icke direkt jordade distributionsoch transmissionssystem, Svenska Elverksfreningen, Stockholm, 1990. ECOBLOC-SVEA Drift och Instruktionsmanual broschyr, Operation and Instruction Manual brochure, (In Swedish), Helmer Verken, Gteborg, 2006 Akke M.: E.ON projekt Ic, (In Swedish), Test Report, Department of Industrial Electrical Engineering and Automation, Lund University, CODEN:LUTEDX/(TEIE-7213)/1-37/(2006), Lund, 2006

Measured Angle (Deg)

80 60

[2] Cable radial 50 km Cable radial 20 km Overhead line 10 km [3] [4]

40 20 0 5km 10 km

15 km

Distance Between Units (km) Fig. 12 Angle between zero sequence voltage and current on the cable radial close to the bus bar as single phase to earth faults are simulated on the cable radial 50 and 20 km from the bus bar and on the overhead radial 10 km from the bus bar [5]

[6]

The fault location dependence of the simulated angles is limited. It does however not decrease when the distance between the units increase. The angle measured by directional zero sequence over current relays does not seem to be a reason to keep the distance between the units short. VIII. DISCUSSION The active earth fault current drastically increases with the length of the cable radials in centrally compensated networks. As seen from simulations and analytical calculations presented in this paper the use of distributed compensation significantly decrease the active earth fault current. Local compensation of radials longer than 20 km can influence the earth fault current. Some major Swedish distribution network owners plan to have a distance of approximately 5 km between the Petersen coils in locally compensated networks. The simulated results demonstrate that investment costs and losses alone do not motivate a distance of 5 km between the compensation coils since a distance of 20 km between the coils would create approximately the same losses. In addition to investment cost and losses, maximum earth fault current on the MV side of the transformer and proper location with respect to disconnectors are factors that influence the rating of compensation coils. This paper describes a number of analytical calculations and PSCAD simulations. Actual test on the ZNdyn transformer and distribution network system has not been carried out as a part of this work. This shall be taken in to consideration when reading the paper. The references used in this paper are all published in Sweden. This is a consequence of the fact that Swedish distribution companies are the first to experience the extensive change in rural distribution network design. Up till now it has not been urgent to analyse this problem in other parts of Europe. IX. ACKNOWLEDGMENT The authors gratefully acknowledge the contributions of S. Lindahl. His advice and his previous work in this field of