Partial discharge diagnostics on very long and branched cable circuits

Transcription

1 11 Nordic Insulation Symposium Stockholm, June 11-13, 2001 Partial discharge diagnostics on very long and branched cable circuits Nico van Schaik, E. Fred Steennis, Wim Boone and Dick M. van Aartrijk KEMA T&D Power Arnhem, the Netherlands 1. Summary Partial discharge diagnostics is a proven tool to diagnose the condition of cable circuits. However the length to be measured is limited because of attenuation and dispersion of the partial discharge signal and because of its reflections, used as indicator for the location of weak spots. Result is an upper limit up to 4 km circuit length that can be diagnosed. A method to overcome this length limitation is to measure partial discharge signals at both ends simultaneously. This has Figure 1 Defect near the conductor (burned material by repeated partial discharge activity) found in a bitumised joint been realised by means of a patented method. The upper limit for the length is thus increased up to 10 km for point to point connections. To diagnose branched cable circuits, simular problems can be solved with success by application of multi-sensor testing. The method used for long lengths and for branched cable circuit will be discussed in detail. Results will be presented. Besides, attention will be paid to the partial discharge detectability of one specific type of defect, water treeing. 2. Introduction Partial discharge diagnostics on medium-voltage and high-voltage cables are a proven tool to reduce the number of unexpected outages. This paper will focus on partial discharge diagnostics for medium-voltage cables only. One of these methods is called in this paper reflection based pd testing and will be explained more extensively in one of the sections.

2 12 Due to the successful application of the reflection based pd testing, which is applicable to non branched cable circuits with lengths up to 4 km, there was a need to develop a test system that could cover cable circuit lengths up to 8 km or more cable circuits that are branched (mainly for the USA) This was solved, by making a test system where multiple sensors are used. Sensors are placed at each cable circuit end. In a long cable there are two sensors, in a branched cable circuit the number of sensors is three or more. For long cable circuit lengths, the advantage of this method is that a partial discharge does not need to travel up and down the cable before it is measured. For branched cable the multiple reflections on branch connection points do not disturb the partial discharge pulse anymore (like it is doing in the reflection based pd testing method). Each of the sensors measure the partial discharges, and localisation of a partial discharge source is done with help of the difference of arrival times of these partial discharges. The problem of accurate time synchronisation was solved by means of a GPS system added to each sensor, which method is patented. Accurate time synchronisation is needed in order to get a reliable localisation of the partial discharge sources. The test method is called in this paper multi-sensor based pd testing method and will be discussed in more detail in one of the following sections. Important to be mentioned is, that partial discharge testing cannot be used to find all types of heavy water treeing in medium-voltage extruded cables. Some more background information will be given in the following section. Also, new extruded cables normally do not show partial discharge activity if the cable was screened on partial discharges before delivery as is usually done during routine testing. Only extruded cables that have been loaded to extremely high temperatures (for instance at hot-spots) may have voids or loose screens and therefore may show partial discharge activity indeed. Bus this is not often seen. Therefore, the focus of partial discharge testing on medium-voltage cables is on accessories for extruded cable circuits cable and accessories for paper insulated lead covered cable circuits 3. Water treeing and partial discharges Because there is some confusion among experts and cable network owners world-wide about the effectiveness of partial discharge testing in the case of an extruded cable with water trees, our opinion on this matter will be given here.

3 13 A water tree is a well-known degradation process of extruded cables, especially in older cables that do not have any water (tree) protection. The background of water tree growth will not be discussed here, many references deal with this subject [1]. There is some agreement, that a water tree is a kind of chemical corrosion of the polymer. It is important to realise that even a large water tree is not a conductor, but still an insulating material (of inferior quality). Consequently, the electric stress enhancement at a tip of a water tree is Figure 2 One large and two small very limited. Therefore, water trees can vented water trees in an extruded cable, bridge the whole insulation of a cable no voids could be found that may without giving a breakdown immediately. generate partial discharges. There are no voids in the water tree that are available or large enough to give rise to partial discharges. A typical large water tree as found in the insulation of a steam-cured 10 kv cable is shown in Figure 2. For these reasons, a field test in many cases will not show up partial discharge activity even for cables with large water trees..but, such water trees are very dangerous. They suddenly can initiate an electrical tree, especially during an over-voltage. Such electrical trees do grow very fast and within minutes they bridge the whole insulation and cause cable breakdown. During this process of growth of the electrical tree, partial discharges are generated which indeed can be measured. However, such discharges are only found at the last stage of life of a cable and for a real Tip of a vented diagnostic test they do not represent the early Figure 3 tree with an electrical tree warning needed. An experiment in the laboratory revealed such a situation. Various 10 m service aged cable samples with large water trees were subjected to a test voltage, starting at the working voltage 1 Uo (which is defined as the system voltage between the conductor and neutral wire). During the increase of the test voltage (with a ramp of about 1 Uo per 5 minutes) the partial discharges were measured. In all cable samples, partial discharges appeared between 2 Uo and 4 Uo. In some cables, the test voltage level was reduced to a level where the partial discharges just did not disappear and it was found

4 14 that the cable samples broke down within minutes after the first appearance of the partial discharges. In some other cable samples, the first partial discharges were measured with high-frequency sensors on both cable ends and stored in a 500 MHz digitizer. Immediately, after that the test voltage was switched off. The discharging location could be traced with an accuracy of about 1 cm. At this spot, the cable sample was cut and the insulation was subjected to a visual inspection. All locations showed a large water tree and also an electrical tree that not had fully bridged the insulation. It is assumed here that this fast growing electrical tree has produced the partial discharges that was measured just before switching off. An example of a water tree with such an electrical tree is shown in Figure 3. Clearly, the open carbonised black channels from the branching electrical tree can be distinguished from the water treed area (the dark finger in Figure 3 represents one water tree branch). These findings explain what could happen in the field during normal service operation and during partial discharge testing. During normal service operation, over-voltages will create electrical trees at the tips of one or more large water trees. If the extinction voltage of the electrical tree is lower than the following normal operating voltage of the cable, soon the cable will breakdown. If it is higher, new over-voltages will further weaken the cable insulation by repeated growth of the electrical tree until breakdown follows. Anyhow, the cable has become very sensitive for over-voltages, also for those occurring during switching operations. A diagnostic test that has test voltages above the normal working voltage of the cable is therefore a threat for a cable with large water trees. It may create electrical trees or if the electrical trees are already there it will further weaken the cable (causing breakdown during testing or soon after returning to normal service operation). Therefore, such a partial discharge test cannot be seen as a real diagnostic test. On the other had, it can trace water trees that are in their final stage of life. But, many network owners would like to know the cable condition in a stage that is preceding the final stage of life, in order to take timely decisions about proper maintenance actions. Dielectric spectroscopy is a real diagnostic test alternative T2 T3 that has proven to be effective [2]. T1 x 4. Reflection based pd testing defect We come back now to real partial discharge testing, focussing on other types of defects than water trees. Briefly, the reflection based pd testing will be explained here (see also Figure 4). The cable is switched off and is subjected to a Figure 4 Principles of relection based pd testing.

5 15 test voltage. The test voltage source can be for instance 0.1 Hz, an oscillating wave or a power frequency based voltage. There is some dispute going on what the most realistic test voltage is, but far more important is the fact that they all generate partial discharges at defects. KEMA is using or recommending all these test voltage sources as long as it is proven that they are effective in finding specific defects. So far, KEMA s experience is based on a test voltage of 0.1 Hz and some tests with power frequencies. The partial discharge pulses are measured with a sensor, which is normally a capacitive divider. The actual partial discharge source location can be derived from the differences in arrival times (identified in Figure 4 with T1, T2 and T3, corresponding with the first, second and third pulse arrival). The sensors normally operate with a frequency bandwidth between 1 and 10 MHz, which is enough to have a localisation accuracy between 0.5 and 2 %, the actual localisation accuracy is depending strongly on noise conditions, cable type and cable length. For very short cables (10 to 100 m range), it is for instance required to have sensors and digitizers with a bandwidth of 10 to 100 or even 400 MHz in order to reach the same localisation accuracy. Three factors influence the success of a partial discharge test much more than the actual equipment applied. These factors are often neglected too much when discussing the right measuring equipment: the experience of testing technicians in reducing the noise, selecting the right test voltage, knowing how to calibrate and in the end knowing how to interprete the data the co-operation with the network owner, who must be able to inform the measuring technicians about the actual cable length, cable types and accessory types. Moreover, earlier experience is very important for the right diagnosis. For example, if in an oil joint many large partial discharges ( pc or more) are found then this will cover (at least) two different possible defects: a too low oil level (which is not very critical) or water ingress (which is very critical). If in this example in the past several joints in the same circuit did fail due to water ingress, there is now a good reason to assume that the remaining joints will suffer from the same problem, in which case immediate replacement of the joints is needed. a database with knowledge rules, that represents all past experience and that can be used to select possible defects (assuming the circuit data given by the network owner is correct). The database should not contain partial discharge magnitudes only, but also partial discharge intensity information, test voltage level, phase information and if possible also pattern recognition tools. 5. Multi-sensor based pd testing method With sensors on both sides of the cable circuit the cable circuit length that can be diagnosed is in principle doubled due to the fact that partial discharge pulses do not

6 16 have to travel up and down the cable anymore. That means that the cable length that can be diagnosed has increased from 4 to 8 km. But there is another additional advantage, making it possible to diagnose even longer cable circuits. This is the fact that a reflected pulse in the reflection based pd method is sometimes hidden or defect partly hidden by the tail of the 1 2 first pulse that will arrive. And this tail is full with oscillations due to the measuring circuit setup. With two sensors, it is not 1 2 needed to try to identify the time reflected pulse, simply the first Figure 5 system. Multi-sensor based pd testing pulse that will arrive at each sensor contains the information needed. Experience has shown that it is possible to diagnose cable circuits up to 10 km. It must be repeated also here that the actual cable circuit length that can be diagnosed is also depending on practical circumstances such as noise conditions and cable type. An important element in the multi-sensor based pd method is the time synchronisation, for which a GPS system at each sensor is applied. The accuracy is such that it will not harm the Figure 6 The signals from two different sensors, from accuracy that can which the partial discharge location can be found. be reached with the bandwidth of the sensors (1 to 10 MHz) and therefore, the total accuracy of the measuring system is still in the order of 0.5 % to 2 %. Also, the alternative solution, the application of atom-clocks was studied, but it was found that these atom-clocks are very sensitive to mechanical shocks, temperature variations and magnetic fields. Partly this could have been solved, but the sensitivity for the direction of the magnetic field could not easily be overcome. And even then, it was not certain whether the atom-clock would

7 17 be accurate enough. However, both solutions (GPS and atom-clocks) have been patented Figure 7 Discharge map, a paper insulated cable circuit of almost 8 km. Both sensors are connected to a digitizer and PC, which is called sensor-set. Each sensor-set has the ability to trigger independently, but under normal operation conditions they operate in a master slave configuration. On a trigger of the master sensor-set, the connected PC will investigate the received data. If it is recognised as a possible partial discharge, the slave defect 1 2 time Figure 8 Testing set-up for a branched cable network with the multi-sensor based pd method. 3 sensor-set will get a signal so that it will store the data in a certain time frame. In this way, the total amount of received data is minimised to reliable information. All this data is sent via one of the cable phases or with another communication device to the master sensor-set for further analysis. An example of the partial discharge pulses from one source, received at both sensors is given in Figure 6. The further analysis is similar to that of a reflection based pd method. An example of a discharge map for a cable circuit of almost 8 km is given in Figure 7. In this map it is seen that many joints have high partial discharge levels. This is an example of a cable circuits, where the oil filled joints have a too low oil level. After refilling the joints, the partial discharge intensity was clearly reduced. In the case of a branched circuit, the measuring principles are identical to that of a long cable. Figure 8 shows the principle set-up in the case of working with three sensors. In principle, any partial discharge will result in partial discharge voltage pulses at each

8 18 sensor. But each branch will attenuate the pulse due to branch reflections, therefore the total attenuation is strongly depending on the travelling path along the cable for a specific pulse. For each partial discharge, an optimal set of pulses can be selected and with that the discharging location and discharge magnitude can be found. In very complex circuits, with many branches it is better to install a few sensors (three or four) on the end of the longest branches. If in this case discharges are found from a T-splice, in which branch there was not a sensor, this indicates that the branch itself including the T-splice has discharges. But the discharging locations can only be identified if in a second round of testing a sensor is placed at the end of this specific branch too. This pre-selection can be very effective for such complex cable circuits. An example of a relatively simple branched cable circuit, including a summary of the test results, is shown in Figure References [1] R. Ross, Inception and propagation mechanisms of water treeing, IEEE Transactions on Dielectric and Electrical Insulation, Vol. 5, Nr. 5, pp [2] N. van Schaik, P. Wirelius. Dielectric Spectroscopy, the Diagnostic Method for Diagnosing Watertree Deteriorated MV XLPE Cables, ERA 2000 Conference, paper S2P4. Results are shown on the mapping diagram for test #1. All distances are shown from Location A to Location B. Results are shown on the mapping diagram for test #2. All distances are shown from Location A to Location B. 8 kv 4,300 pc 12 kv 7,700 pc Retest or replace the section in 1 year Substation Test #2 Location B Test Cap PD Zone 8 kv 2,100 pc 12 kv 3,300 pc Inspect the splice. Retest the section in 1 year kv 2,100 pc 12 kv 3,500 pc Inspect the splice 8 kv 2,100 pc 12 kv 3,300 pc Inspect the splice kv 3,300 pc 12 kv 6,000 pc Inspect the splice Test #1 Location B Test #2 Location B kv 2,500 pc 12 kv did not increase Retest the section in 1 year Test #1 Location B kv no PD 12 kv 3,000 pc Retest the section in 1 year 8 kv 5,200 pc 12 kv did not increase Inspect the splice Figure 9 Example of a test result of a branched cable circuit