CI R ED

20th International Conference on Electricity Distribution

Prague, 8-11 June 2009 Paper 0527

PRACTICAL EXPERIENCE WITH FAULT LOCATION IN MV CABLE NETWORKS

Frans PROVOOST Alliander - The Netherlands Frans .Provoost@Alliander.com Walter VAN BUIJTENEN Phase to Phase - The Netherlands W.M.vanBuijtenen@Phasetophase.nl magnitude and angle. In the second stage short circuit calculations are performed on a model of the network. The results ofthese calculations are compared with the results of the first stage, in order to fmd a location giving the same fault impedance. The developed fault location system calculates the fault reactance based on FFT over one period. This makes it possible, to give an indication of the fault impedance, even if a fault lasts only two or three cycles . However, the accuracy of such calculations depends on the effect of transient phenomena during these short events. Experiences from this pilot system resulted in the implementation ofa fault location system in 10 substations, that were equipped with the new developed substation automation system , called SASensor, described in [3]. The high sampling rate of this system and the diversity of earthing and operation of the geographically spread sub stations showed new phenomena in voltage and current. Improvement ofthe SASensor system and the fault location software made it possible to correctly interpret the various registrations and, in case of a real fault, to pin point the location close to the reallocation. This paper describes the experiences with this system.

ABSTRACT
As the major cause ofinterruption time is afault in the MV network, the Dutch grid operator Alliander puts effort in faster location offaults in these networks, which entirely consist ofburied cables. After severalpilots afault location system was implemented in 10 substations equipped with the SASensor protection and control system. The SASensor system uses a high sampling rate and collects event recording data as soon as a set pointfor the current is exceeded. Therefore also events like inrush currents and transients due to selfextinguishing single phase faults are recorded. The data processing routine in the substation is able to distinguish between a real fault and other events. The process is fully automated. All events are shown on a PC screen and, in case of a fault, printed on paper. The system performs according requirements on accuracy and processing time.

INTRODUCTION
The ideas of implementing fault location in the MV networks of Alliander started in the last years of the 20 th century. In order to be a success, several requirements were specified. The location of the fault should be known within five minutes after its occurrence. The accuracy should be 100 meters for two and three phase faults and 500 meters for a single phase to earth faults. This implies that for a two or three phase fault its location can be determined to either the faulted cable section or the neighbouring section. For single phase to earth faults the accuracy is less, because of the limited accuracy of the zero sequence data of the network. In order to be introduced on a large scale, the fault location system had to be cheap and should not require extra boxes and wiring in the substation. Since 1997 data from faults in the substation ofZaltbommel has been collected during a pilot for fault location [1]. The principle ofa fault location system could be proved, but the costs of implementation were too high at that time. The introduction of digital protection equipment with embedded functions for transient recording opened new possibilities. The fault location could be based on the information from these devices . In 2002 , the refurbishment of the old protection system in substation Zaltbommel was used to implement a pilot system for fault location. A detailed description of this system is given in [2]. After the occurrence of a fault , the relevant sampled data of voltage and current is sent to the control centre. Here the data is processed in two stages . In the first stage the type of fault and its impedance is calculated by a new developed tool. This tool uses Fast Fourier Transformation (FFT) to transform the sampled data into complex variables with

FAULTLOCATION SYSTEM WITH SASENSOR

In 2007 the first 10 substations were equipped with the SASensor protection and control system. In order to implement fault location in these substations several new questions arose : Where to process data Which data should be transmitted from the substation to the control centre In this case it was decided, that all relevant data should be sent from the substation to the control centre (Figure 1).

Control Centre
Substation 3

Substation N

Figure 1: Data collection and transmission

CIRED2009 Sess ion 3

Paper No 0527

CI R ED

20th International Conference on Electricity Distribution

Prague, 8-11 June 2009 Paper 0527

There the already developed tools for data processing and fault location could be used. In this way , the developer of the SASensor product only had to focus on triggering, data collection and the transmission of this data . The company developing the applications for the control centre could focus on the data processing and further developments in fault location. Directly after the SASensor system (using a high sampling rate) was implemented in the substations new types of events were recorded. Examples are: Extinguishing single phase faults Single phase faults in networks with isolated neutral Inrush currents Contributions of generators to faults Registrations of non disturbed feeders due to faults Registrations due to testing of equipment. The data processing routine was enhanced in order to deal with these various events. Only in case of a real fault , the data was sent to the fault location system. It turned out, that, as expected, the fault location system worked well for two and three phase faults and for single phase faults in earthed networks. The method for location of single phase faults in networks with isolated neutral had to be improved. However, this has a low priority, due to several reasons: These faults are not switched off. These faults either extinguish or develop into a multi phase fault, which can be located with enough accuracy. It was decided to generally transfer these networks into impedance earthed networks.

INTERESTING SITUATIONS
One ofthe problems with single phase to earth faults is, that it takes time before the fault has fully developed. Many recordings of single phase faults show events that are self extinguishing. Some registrations show transients, that make clear that the fault is not always stable. This is visible , not only in networks with isolated neutral, but also in networks that are earthed by means of an impedance.

Self extinguishing single phase faults

In case of self extinguishing single phase to earth faults, especially in repetitive sequences, the FFT will result in a smoothed registration, that looks similar to a single phase to earth fault of longer duration and smaller amplitude. As a consequence, the calculated values for current and voltage in these situations will result in a wrong impedance for the fault , and thus in a wrong location. It is therefore necessary to add a criterion regarding self extinguishing faults . Self extinguishing single phase faults are characterized by relatively large zero sequence currents that last less than a quarter of a cycle. Due to this current there is a change in the zero sequence voltage. After extinguishing of the fault, the magnitude of the zero sequence voltage slowly decreases. A test on self extinguishing faults, must be performed on the sampled data and is based on the following items : A threshold value for the absolute value of deviations in the momentary zero sequence current. A threshold value for the absolute value of the zero sequence voltage. A possible extinguishing fault starts when the deviation of the zero sequence current is above the threshold value In the next halfcycle less than halfofthe deviations of the current must be above the threshold value The zero sequence voltage at the end ofthis half cycle must be above the mentioned threshold value It must be noted, that these items also may occur for single phase faults in networks that are either directly or impedance earthed. However, in these situations it is unlikely that the fault is self extinguishing. Therefore the test on self extinguishing faults must be performed in the case that, based on the FFT evaluation, either no fault has been detected or a single phase to earth fault in a network with isolated neutral has been detected. Ifin these situations a self extinguishing fault has been detected, the system must change its conclusions towards the self extinguishing fault in the following cases: The FFT evaluation did not detect a fault The FFT evaluation detected a single phase fault , but the previous event of a self extinguishing fault was a few samples before the calculation moment of the single phase fault. In this way , a self extinguishing fault , resulting in a permanent fault during the same registration, will still be considered as a single phase fault.

PERFORMANCE
The problem with the first pilots of fault location in Zaltbommel was , that only a limited number of faults occurred. In average about 4 faults in a year. Due to the fact, that SASensor is implemented in 10 substations, more faults could be observed. Table 1 shows the result for the period between October 2007 and December 2008.
Location

Fault Typ e Recordings OK Adja cent >lRMU Three Phase 7 3 3 1 Two phase 10 S 1 4 Sing le phas e (ea rthed networks) 4 2 0 2 Sing le phas e (isolated neu tra l) 3 2 0 S

Table 1: Results of fault location

From this table it becomes clear, that most ofthe faults were located in the correct section or in the adjacent section. This is within the required accuracy. A few results were outside the required accuracy. Two possible reasons can be given: The fault was far away from the substation, so specific assumptions for calculating the fault impedance were no longer valid The network model was incorrect. The accuracy of the results can be improved by improving the network model of these substations.

CIRED2009 Session 3

Paper No 0527

CI R ED

20th International Conference on Electricity Distribution

Prague, 8-11 June 2009 Paper 0527

Non stable single phase faults

Measured voltages

uc - - Ie

IFault

1.S " - - - - - - - - - - - - - - - - 200 150 100 SO

1.5
-so
-100

g
;:
u

0.5

l S0

~
~
~

-i.s
0
-05
-1
460 46 5 470 47S 480
Time (ms)

200 48S 490 49S SOO

Figure 3: Detailed waveform of voltage and current during a non stable single phase fault in a network with isolated neutral. o
500 1000
Time (ms)

- 1 .5 +--~-~-~-~-~-~-~-~

1500

2000

Measured c urrents

160
140

120 100 80
60
_ ~ 40 20

~ -20 u -40

-60 -80

-100 -120
-140

-160 -l--~-,....--~-,....--~-,....--~----,

500

1000
Time (ms)

1500

2000

The FFT ofsuch a signal will not result in the proper values for voltage and current. This may result in a large error for the calculated corresponding impedance. Therefore this event is not suited for the fault location algorithm . A correction algorithm must find a time interval of at least one cycle, in which no extinction or re-strike of the of the fault occurs. From the figure the following characteristics for these events can be deduced : An extinction of the fault is characterised by a value close to zero for both the fault current (IFault) and its derivative . A re-strike of the fault is characterised by a sudden change in the fault current, resulting in a derivative which is much larger than the maximum slope of the expected current curve. The search algorithm must use the momentary samples instead of the FFT signals . Preliminary tests of this algorithm show promising results as is shown in Figure 4.
- - - - - - - - - - - - - --------,
- X lc Ca XOK X steady

Figure 2: Measured voltage (top) and current (bottom) during a single phase fault in a network with isolated neutral.
Figure 2 shows the registration ofvoltage and current in the faulted phase during a single phase to earth fault in an MV cable. At first glance it seems that the fault current is stable, but the voltage shows some variations. When zooming in on such a variation, it can be observed that during these variations, the fault extinguishes and almost immediately restrikes (Figure 3). The fault current (IFault) could be derived from the three measured phase currents. At t=470, the fault current crosses zero and the fault extinguishes. From that moment the voltage in the faulted phase (UC) starts to rise. When (at t=472) the voltage reaches a value near to 1.3 kV, the fault re-strikes . This can be seen in the rapid decreasing of the voltage and the sudden increase of both the fault current (IFault) and the current in the faulted feeder (IC). A same event happens at t=480 .

10
9

J
i

I 2x

1 ,

a .
200 400 600 800 1000
Time (ms )

1200

1400

1600

1800

2000

Figure 4: Calculation of the fault impedance as a result of a non stable single phase fault in a network with isolated neutral

CIRED2009 Session 3

Paper No 0527

CI RED

20th International Conference on Electricity Distribution

Prague, 8-11 June 2009 Paper 0527

Here, the calculated impedance (XCalc) is compared with the calculated impedance ofa registration at the moment the fault was stable (XSteady). The correction algorithm was able to identify two stable points for the correct calculation of the impedance to the fault location. These points are circled in Figure 4. The corresponding impedances are given as (XOK). The figure shows that the values for XOK correspond well with the values calculated during the registration of the stable fault (XsSteady).

experience is gained. This may obstruct the acceptance of use in daily practice. It must be prevented, that operators see the product as an extra workload, instead of a useful assistance in their daily work.

CONCLUSIONS
It is possible to implement robust fault location systems based on simple data processing techniques in the MV network. These systems can pin point two and three phase faults within 100 meters. Single phase faults in earthed networks can be located with an accuracy of about 500 meters. The time between the actual fault and the calculation ofthe possible fault is less than five minutes and is mainly dependent on the transmission time ofthe relevant data from substation to the control centre. The actual calculation time is less than 20 seconds. The implementation of the system in more substations was possible due to the simple concept of SASensor. Directly after the system was implemented in different substations with various network structures, new events could be recorded. The system is adapted for a correct interpretation of these events. Study of self extinguishing faults, unstable faults and other transient phenomena resulted in measures to minimize the number of false alarms. A major concern is the actuality and correctness of the network models and the acceptation of the network operators and their managers.

Other phenomena
In practice the devices that are used to record fault situations will start recording when the measured current is above a certain level. This means, that also other events, like switching large loads, inrush currents of transformers and contributions ofgenerators and motors to faults in other feeders will generate a recording. The system must conclude from the wave forms, that these phenomena are not caused by faults in the monitored feeder. From our experience with fault location, it can be concluded, that this can be achieved by utilising a proper set ofthreshold values for the different types of faults.

REMARKS
A key factor for success ofsoftware tools implementation is a good communication between operators, developers and manufacturers. The wishes of the operators have to be translated towards the developers. The modem software used for fault location has to be accepted for use in daily practice. It also may involve extra training. The operators need to see the product as a useful assistance in their daily complex work and not as an extra workload. Therefore it is required, that the software itselfis straightforward and that its interaction with the operators is simple and clear. The amount of input has to be minimized, the processes have to be automated and the results must be easy to interpret. It is also required that the number of false alarms should be minimized. For a good functioning ofthe programs, up to date network information is necessary. This implies special requirements for side processes like Geographical Information Systems and the people involved. The accuracy ofthe fault location method, strongly depends on the correct model of the network which is used for the short circuit calculations. These network models must include necessary information for fault location, such as protection devices, double bus bar systems and information of the HV network and the HV IMV transformers. As MV networks are subject to changes, these network models must be updated on regularly base. The update of the networks within Alliander is performed in a geographic information system (GIS). This system is able to generate the required network models. Although, the fault location system is implemented in 10 sub stations, the number ofrecorded faults is not large. This means, that there are large time intervals between the actions of the fault location system. In this way not much

REFERENCES
[1] M.M. Saha; F.Provoost, 1999, "Fault Location in Medium Voltage Networks", Proc. 1999 International

Conference on Electricity Distribution - CIRED, session 3, Paper no 2 [2] P.M. van Oirsouw; F.Provoost, 2003, "'Fault Localization in an MV Distribution Network", CIRED 17th International Conference on Electricity Distribution, session 3, Paper no 64 [3] F. Volberda; M.J.M. van Riet; A. Pikkert, 2007, "The Power of Simplicity", CIRED 19th International Conference on Electricity Distribution Vienna, session 3, Paper no0252

CIRED2009 Session 3

Paper No 0527