The present invention relates to the location of faults in cables, such as low voltage power cables. SUMMARY OF THE PRIOR ART

Many existing low voltage power cables are of the paper-insulated type. If water penetrates such cables, the insulation provided by the paper decreases, until breakdown occurs. At that breakdown, a high current flows to ground, by e.g. the sheath of the cable.

It is necessary to detect the location of that fault, in order to repair it. One known method of determining the location of the fault relative to a fixed point in the cable is to use a pulse-echo or Time Domain Reflectometry technique. This technique is disclosed in e.g. the book "Underground Cable Fault Location" by B Clegg pp 42-71. In such a technique, a known signal pulse is applied to the faulty cable, and that pulse is reflected at the fault. By determining the time between the transmission of the pulse, and the reception of its reflection, the distance to the fault may be estimated. When the cables are to carry multi-phase currents, it is necessary to prevent the pulse passing onto parallel cables, which could provide multiple reflections. It is therefore usually necessary to

interpose a substantial inductance between the power supply point and the device for applying the pulse to the cable, so that the pulse path is limited to the cable being tested. One problem which occurs in cables, particularly of the paper insulated type, is that the high current which flows when insulation breakdown occurs may be sufficient to dry-out the insulation at the vicinity of the fault. The fault then disappears and cannot be detected until the level of insulation has decreased again. This prevents the pulse echo technique described above being used. Even when such a technique can be used, there is a practical limit to the size of the device which can be connected to the cable, and thus limitations of the power of the pulse, and the power limitations limit the maximum distance between the device and the fault which can be determined sufficiently accurately.

This problem was addressed in GB-A-1539118. That document proposed that a plurality of voltage detectors were provided along the cable which repeatedly supplied the voltage on the cable. When a fault occurred, the voltage measurements at the instant of the fault at each detector were analysed. The analysis was on the basis of a voltage gradient technique, in which the change in voltage along the cable was used to determine a line corresponding to the voltage gradient on each side of the fault, with

the intersection of those lines then being the location of the fault.

SUMMARY OF THE PRESENT INVENTION

As mentioned above, GB-A-1539118 provided a plurality of detectors along the cable. For a buried cable, this is inconvenient since all the detectors have to be buried with the cable and be arranged to transmit their voltage measurement data to a control processing site. This increases the complexity of the detectors. Moreover, for the voltage gradient technique to be used, there need to be detectors on each side of the fault, and the accuracy of the measurement depends on the spacing of the detectors, hence a large number of detectors must be used. The present invention is based on the realization that the waveform of the signal on the cable at the time of the fault contains sufficient information to enable the distance between a single measurement point and the fault to be determined. Analysis of that waveform by autocorrelation techniques thus enables a single measurement point, and hence a single detector, to be used.

Thus, the present invention seeks to make use of the normal current that is present on the cable, in order to detect the position of the fault. When the fault occurs, the current waveform will have components which enable the fault to be located.

Therefore, according to the present invention,

the signal on the cable is continuously sampled, for successive sampling periods. Then, when a break in supply due to a fault is detected in any sampling period, the sample of a previous (usually the immediately previous ) period is investigated to determine the waveform components present during that previous period, which then enables the fault to be located. This investigation of the sample of a preceding period is triggered by e.g. disconnection of supply or tripping of fuses detected by voltage monitoring, when the signal on the cable has dropped to a predetermined level because of the existence of the fault. Normally, when there is a fault, the signal of the cable will decay with time-repeating components, and by looking at a preceding sampling period, those components can be investigated, for the time interval therebetween enabling the distance from the fault to be determined.

The signal needs to be sampled at a rate which is high enough accurately to represent the highest frequency of interest, which will determine the range resolution. If the sampling period is too short, insufficient information will be obtained. Any such insufficiencies generally limit the resolution of distance that can be obtained. Thus, to achieve a resolution of x (i.e. to determine the location fault to a particular length x of the cable), it is necessary to sample at a rate greater than K.c/x where

K is the velocity of propagation in the cable. The signal should be sampled long enough to record a number (e.g. 3) round trips from the most distant possible fault to the measuring site. The maximum length of the sampling period is determined by the size of the buffer in which the sample or samples are to be stored.

As described above, the present invention can be used to derive the distance between a single measurement point and the fault, thereby permitting a single detection to be used. However, the accuracy of the measurement of the fault position then decreases with the distance from the detector to the fault. To reduce this inaccuracy, it is possible to use two or more detectors with known separation( s ) , to measure the distance between the fault and each detector, and then to compare the distance measurements thus obtained, using the known detector separation(s) to reduce the error in the determination of the location of the fault. A plurality of detectors may be mounted in a portable unit which is brought into proximity of the cable when a measurement is to be made.

It can be seen that the spirals detected by each detector are used to measure the distance between that detector and the fault, it is not necessary to provide detectors all along the cable as in GB-A-1539118. Moreover, since the determination of the position of the fault does not use voltage gradient technique, it

is not necessary for there to be detectors on both sides of the fault.

In one arrangement, the current on the signal is investigated, e.g. by use of a suitable transformer coupled to the cable. Such a transformer may be of the split-core type, since this can be applied around the cable at a convenient location so that the distance to the fault can be determined whilst the cable remains operational. This is particularly useful where the fault dries-out the insulation, since the first occurrence of the fault may then trigger the placing of the transformer around the cable, to detect a subsequent insulation failure later.

It is also possible, however, to monitor the voltage on the cable. Normally, that voltage will be sinusoidal, but the fault will produce a variation which will differ in dependence on the distance to the fault.

As a third alternative, magnetic fields associated with the signal on the cable, and the current due to the fault, may be used. BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the present invention will now be described in detail, by way of example, with reference to the accompanying drawings, in which:

Fig. 1 shows a cable fault location arrangement, being a first embodiment of the present invention;

Fig. 2 shows a cable fault location arrangement

being a second embodiment of the present invention;

Figs. 3A and 3B are auto-correlation functions used in the present invention;

Fig. 4 is a cable fault location arrangement being a third embodiment of the present invention;

Fig. 5 is a block diagram showing an investigation device for use in the embodiments of Fig. 1, Fig. 2 or Fig. 3; and

Fig. 6 is a schematic view of the waveform that will be sampled using the present invention. DETAILED DESCRIPTION

Referring first to Fig. 1, a cable 10 is connected to a transformer 11, which transformer applies a signal to a cable 10. A split-core current transformer 12 is mounted around the cable, and connected to a current investigation device 13. When a fault 14 occurs at a point P on the cable, a high current flows which can be detected by a suitable detector to break the connection between the transformer 11 and the cable 10. This can be achieved by disconnection or tripping of a fuse or circuit breaker.

When that happens, the behaviour of the signal on the cable 10 is investigated by the investigation device 13 for the period immediately prior to the occurrence of the high current. In particular, the investigation device 13 continuously samples the current on the cable 10, and when the high current

occurs in any sampling period, the immediately preceding sampling period is investigated. Such investigation involves analysis of the signal on the cable 10, using autocorrelation techniques well known in investigation of signals. Such autocorrelation techniques are discussed in e.g. "Digital Signal Process - A Practical Approach" by I Leachor and Jervis, published by Addison-Wesley pp6-10, the book "Introduction to Digital Signal Processing" by Proalas and Manolalis published by Maxwell-Macmillian pplll- 115. The signals will have time-repeating elements which permit the distance between the split-core transformer 12 and the fault point P to be determined, making use of the signals and the reflection of the signals from the fault.

To carry out such sampling, the investigation device 13 may employ a circular buffer memory which stores data corresponding to the signal for the immediately preceding time period, at the same time that it is receiving the data corresponding to the data corresponding to the signal for the current time period. When this data for the whole of the current time period has been received, the data currently in the buffer memory (from the preceding time period) is replaced by the data from the current period, and the cycling starts again.

The embodiment of Fig. 1 is satisfactory where the is a single cable 10. However, as shown in Fig.

2, the cable 10 may branch into separate branches 20, 21 and 22. If a fault 14 occurs at a point FI, measurement using a split-core current transformer 12 and device 13, as in the embodiment of Fig. 1, means that the distance to the fault can be determined, but the branch 20, 21, 22 in which the fault is present cannot be determined. Thus, if the arrangement of Fig. 1 is used, the system can detect that there is a fault at points FI, F2, F3, being points on the branches 20, 21, 22 which are equidistant from the split-core current transformer 12, but cannot determine which corresponds to the fault.

Therefore, in the second embodiment of the present invention, voltage monitors 23, 24 and 25 are connected to the respective branches 20, 21 and 22 such as by way of a live power socket. Each voltage monitor 23, 24, 25 continuously and periodically samples the voltage waveform on the respective branch 20, 21, 22 of the cable and stores signals corresponding to that voltage waveform for preceding sampling periods. Normally, the voltage waveform will be sinusoidal. However, a fault in the cable will produce a waveform dependent on the location of the fault. Assuming that the fault is at point FI on branch 21, the voltage waveform detected by the voltage monitor 24 will decay virtually to zero. The same fault will cause changes in the signals to the voltage monitors 23, 25 but these will have a quasi-

sinusoidal decay pattern which will depend on the fault distance.

The auto-correlation functions which are then used are illustrated in Figs. 3A and 3B. Fig. 3 shows the auto-correlation function used when the fault is relatively close, whereas Fig. 3B shows auto¬ correlation function when the fault is relatively distant.

In the embodiment of Fig. 1, the accuracy of location of the point P at where the fault 14 occurs will depend on the distance between that point P and the split-core current transformer 12. If this distance along the cable 10 is too large, the location of the fault 14 may be insufficiently accurate. To avoid this problem, the embodiment of Fig. 4 may be used. In that embodiment, there is a portable receiver 30 which has a pair of aerials 31,32 which detect magnetic fields. The aerials 31, 32 are arranged so that they can be parallel to the cable 10, since in that state they will not detect any magnetic fields due to the signal on the cable 10 itself.

When the fault 14 occurs, however, the direction of the fault current causes magnetic fields to be generated which will be detected by the aerials 31, 32. If the outputs of the aerials 31, 32 are monitored continuously, and periodically, and samples stored in a buffer in a similar way to the other embodiments, then a section of high current caused by

the fault will enable preceding sampling periods for the receiver 30 to be investigated. For example, the investigation device 13 may generate a transmission signal 33 on detection of that high current, which triggers the investigation of signals stored by the receiver 30. By comparison of timing of signals at the two aerials, 30, 31 and since the distance between those two aerials will be known, the relative location of the receiver 30 and the fault 14, together with an estimate of the distance thereof can be determined. It can be seen that the embodiment of Fig. 4 can operate even if the split-core current transformer 12 is a long distant from the fault P, provided that the receiver 30 is close thereto. Fig. 5 shows the structure of the investigation device 13 used in the embodiment of Fig. 1, Fig. 2 or Fig. 4. The signal from the transformer 12 passes via an amplifier 40, a high-pass filter 41 and a further amplifier 42 to an analogue-to-digital converter 43. The digital signal thus generated is passed to a buffer in the form of a RAM 44 in which the digital data is stored. The storing of the data in the RAM 44 is controlled via a CPU 45 and an address generator 46. The CPU 45 and the address generator 46 ensure that, at all times, there is data corresponding to at least one sampling period in the RAM 44.

Data thus arriving at the RAM 44 is immediately stored therein at addresses determined by the address

generator 46, and data which was collected in an earlier sampling period, and is now no longer necessary, is deleted from the RAM 44. The CPU 45 determines, at any time, which data needs to be retained in the RAM 44 and which can be deleted.

When a fault occurs, the current on the line changes and such a change is then detected by voltage monitoring, to trip a fuse or circuit-breaker. The characteristics of the signal over the period between the occurrence of the fault and that tripping, will enable the distance to the fault to be determined. Thus, when that tripping occurs, a signal is received via a communication link 47 at the CPU 45, and this causes the data corresponding to one or more preceding sampling periods to be retrieved from the ram 44 and passed to the CPU 45 for investigation. That investigation permits the distance of the fault to be determined.

Fig. 6 shows a typical sample at the time of a fault. Initially, prior to the occurrence of the fault, the signal has a sinusoidal characteristic 50. When the fault occurs, at time T, however, there is a sudden change in the current and an oscillation occurs which then determines the distance of the fault. In considering Fig. 6, it should be appreciated that Fig. 6 represents a sample backwards in time from point S at which the tripping of a fuse or circuit- break occurs, and the CPU 45 is triggered to release

the sample from the RAM 44. The length of the sampling period, backwards from the point S, thus needs to be sufficiently long to ensure that a suitable sample is obtained. The sample should be long enough to record a suitable number (e.g. 3) round trips from the most distant possible fault to the point at which the transformer 12 is located. This determines the maximum length of the sample, but there is also a minimum length because the signal needs to be sampled at a rate which is high enough accurately to represent the highest frequency interest, since this will determine the range resolution, i.e. the accuracy to which the fault can be located. If range resolution of x is required, it is then necessary to sample at a rate greater than K.c/x where K is the velocity of propagation of signals on the cable.