EVALUATION OF ON-SITE DIELECTRIC RESPONSE METHODS FOR NON- DESTRUCTIVE TESTING OF WATER TREED MV XLPE CABLES

Transcription

1 EVALUATION OF ON-SITE DIELECTRIC RESPONSE METHODS FOR NON- DESTRUCTIVE TESTING OF WATER TREED MV XLPE CABLES Sverre Hvidsten*, Peter Werelius**, Jørgen Christensen*** * SINTEF Energy Research (SEfAS), Norway ** Royal Institute of Technology (KTH), now with Programma Electric AB, Sweden *** Research Institute for Danish Electric Utilities (DEFU), Denmark SUMMARY Condition assessment of power apparatus by diagnostic testing is becoming increasingly important. The commercial available diagnostic techniques for detection of water tree degradation are generally based upon measurements of the dielectric response, either by measurements in the time or frequency domain. In this study the results from a joint Nordic project for condition assessment of water tree aged 1 and 4 kv XLPE cables started in 1997 are presented. The purpose of this work has been to examine the influence of the degree of water tree ageing, different terminations and on-site conditions on the prediction of the insulation status. There have been four main objectives in the project; The first objective of the work has been to compare the diagnostic results with that obtained from measurement of AC breakdown strength and degree of water treeing on the same cables. Four commercially available diagnostic techniques, all based upon measurement of the dielectric response, have been used for nondestructive testing of different types of medium voltage XLPE cables. The second objective of this work has been to study the dielectric response from different types of terminations, which can become large compared to the response from the cable and may cause an incorrect interpretation of the cable condition. The third objective has been to evaluate the reproducibility of the on-site measurements and to examine different on-site conditions using two of the frequency domain methods. The on-site dielectric response measurements have been performed at different times during a year. Finally, the degree of water treeing can be strongly dependent upon cable design, cable manufacturer, and so on. Information about these factors can accordingly be very important for making a reliable estimation of the degree of water tree ageing based upon dielectric response measurements. Therefore a database and has been developed where both the diagnostic measurements and a detailed information about the cable system are included. The results show that the condition of old cables with a high density of vented water trees, are correctly assessed by all the test methods. To detect cables with water trees bridging the insulation, it is necessary to also measure the polarisation currents when using the time domain methods. The results indicate that in order to avoid dielectric breakdown during diagnostic testing the test voltage should be lower than U o. The good correlation between the estimated cable condition and the actual cable condition, indicate that the test voltage can be limited to U o. However, it can become necessary to increase the test voltage to above the service stress (U o ) in order to assess the condition of cables with very few but long vented water trees. Terminations can have a voltage and frequency dependent dielectric response similar to that observed for water tree aged cables. Thus, during diagnostic testing of short cables in service, the response from the terminations should therefore be known in order to avoid wrong prediction of the cable condition. However, water tree aged cables with high density of trees can in any case be correctly assessed as the response from the cable is likely to be much larger than from the termination. The on-site measurements have revealed that high air humidity can cause misinterpretation of the cable condition due to large leakage currents across the terminations. This is specially the case for measurement made at air humidity larger than 80%. However, a guard ring can reduce the contribution from such currents. Information of the service record of the cables should be used to verify the results from on-site measurements. The database for storing diagnostic measurements and information of the cable system can be used for making a more reliable condition assessment.

2 EVALUATION OF ON-SITE DIELECTRIC RESPONSE METHODS FOR NON- DESTRUCTIVE TESTING OF WATER TREED MV XLPE CABLES Sverre Hvidsten*, Peter Werelius**, Jørgen Christensen*** * SINTEF Energy Research (SEfAS), Norway ** Royal Institute of Technology (KTH), now with Programma Electric AB, Sweden *** Research Institute for Danish Electric Utilities (DEFU), Denmark ABSTRACT This paper presents results from a joint Nordic project for condition assessment of water tree degraded 1 and 4kV XLPE cables. The results show that that the condition of old cables with a high density of vented water trees, are correctly assessed by all the test methods. To detect bad cables with water trees bridging the insulation, it is necessary to also measure the polarisation currents when using the time domain methods. In case of cables with very few but long vented water trees, the evaluation criteria need to be further developed. During diagnostic testing of cables in service, the response from the installed accessories and the air humidity should be known in order to avoid wrong interpretation of the cable condition. INTRODUCTION In this study the results from a joint Nordic project for condition assessment of water tree aged 1 and 4 kv XLPE cables started in 1997 are presented. The purpose of this work has been to examine how the degree of ageing, different terminations and on-site conditions can influence the prediction of the insulation status. In addition, a database has been developed in order to more effectively use these factors during the evaluation of cable condition. The main objectives of the work have been; to compare the diagnostic results using four commercial available diagnostic techniques with that obtained from measurement of AC breakdown strength and degree of water treeing on the same cables. to study the dielectric response from different types of terminations. to examine different on-site conditions using the two frequency domain methods. to develop a database where both the diagnostic measurements and detailed information about the cable system can be included. EXPERIMENTAL Non-Destructive Diagnostic Test Methods The following four diagnostic test methods were used during the laboratory testing of the XLPE cables [1, ]: a) Depolarisation current measurements after application of 1kV DC for 0 minutes (IRC). After seconds of short-circuiting, the depolarisation currents are measured the following 0 minutes. An empirical ageing A-factor is calculated based on these currents to classify the ageing status of the cable insulation, indicated by residual breakdown voltage, k U o. b) Return voltage measurements after the application of DC charging voltages up to U o. The cable is subjected to a DC voltage for minutes, discharged for seconds, and subsequently followed by measurements of return voltages for 10 to 40 minutes. The return voltages typically reach a maximum after about 1 minute. A linearity factor (L) is calculated as the ratio between the maximum values at U o and U o, and values larger than indicate ageing. Condition is classified as Good up to Strongly water tree aged. It is also possible to measure the current during charging of the test object (polarisation currents), but this was initially not used for diagnostic characterisation. c) Tan δ measurements at U o and U o at 0,1Hz. The magnitude of tan δ at U o, as well as the degree of nonlinearity are used to classify the cable as Good, Aged, or Strongly damaged. d) Capacitance and tan δ measurements in the frequency range from 10 to 0,1Hz, and voltages up to U o. The frequency characteristics as well as the magnitude of tan δ at U o, and the difference between the values at U o and 0,U o are used as a diagnostic criterion given by the residual breakdown voltage (k U o ). The diagnostic testing was performed in the following order, based upon the highest applied test voltage; depolarisation current (1kV, DC), capacitance and tan δ (U o, AC), return voltage (U o, DC) and finally tan δ (U o, AC). The classification of ageing status was performed by the manufacturers of each method before destructive characterisation of the cables. The frequency domain methods were used during the on-site testing of the cables. During the laboratory measurements of the terminations only the capacitance and tan δ method was used. Description of Cables for Non-Destructive Testing in Laboratory A summary of the cable characteristics is given in Table 1. The examined cables had been used in the Scandinavian distribution network for more than thirteen years before they were shipped to the laboratory for diagnostic testing. The cable lengths ranged from 80 to 190 meter (each phase). Cable 1 and 4 showed good service

3 performance, while the other cables had suffered from one or two breakdowns during service. TABLE 1 - Description of the seven service aged 1 and 4 kv XLPE cables used to compare the four different non-destructive test methods in laboratory. All cables had aluminium conductors. Cable Type of cable Type of insulation screen Service voltage [kv] 1 1xx0 mm Paint and tape xx40 mm Strippable 11 79/80 1xx0 mm Strippable 1, x1x400 mm Strippable xx0 mm Paint and tape 1, xx0 mm Strippable xx0 mm Graphited 10, 1979 Electric Breakdown Testing of the Cables After the diagnostic testing, the cables were cut into 10 meter long sections followed by an AC step test until breakdown. This was done in order to examine the variation of the degree of ageing along the cable length. The step test was performed by applying U o for minutes, and subsequently every fifth minute, the voltage was increased by U o until breakdown occurred ( U o is defined as the system voltage divided by according to IEC 60 18). If breakdown occurred during the non-destructive diagnostic testing, the defects were cut away, and joints were installed. Water Tree Analysis The water tree analysis was performed by microscopy analysis of 0,mm thick microtomed slices, stained in methylene blue dye solution. In each slice the length of the longest tree, and the tree density were recorded. Description of Terminations The terminations were mounted on a new 6m long 1 kv XLPE cable. Year of installation TABLE - Description of the different types of terminations (M = manufacturer) Termination Type of termination M Mounting type 1 Stress relief A Slipover Stress relief A Elbow connector Stress relief B Elbow connector 4 Stress relief C Slipover Stress relief D Slipover 6 Resistive field grading A Tape 7 Resistive field grading A Slipover 8 Resistive field grading B Cold shrink 9 Resistive field grading E Heat shrink 10 Resistive field grading E Heat shrink 11 Resistive field grading F Cold shrink During the measurements a guard was used in order to avoid contribution from the cable insulation to the measured response []. On-Site Diagnostic Testing The on-site measurements were performed on more than twenty old cables. They were mainly equipped with insulation screens of graphite painting and semiconductive tapes and produced before Some of the cables had suffered from breakdown during service [4]. In order to measure the influence of air humidity, the reproducibility due to different load conditions and surrounding temperatures, the measurements were performed at different seasons during the year. Measurements were also performed in order to investigate the influence of the coupling network, applied voltage and the terminations on the measured response. EXPERIMENTAL RESULTS AND DISCUSSION Electric Breakdown Testing of the Cables All the examined cables had reduced breakdown voltages compared to unaged cables. The results presented in Table show that Cable and 7 had the lowest breakdown voltages (U o ). In case of Cable 7, breakdown occurred during the diagnostic testing of Phase 1 (U o at DC) and Phase (U o at 0.1Hz). Cable had the highest breakdown values (6-7U o ). The variation of the breakdown voltage along the cable length was found to be small, indicating a relatively uniform degree of ageing along the cables. Water Tree Analysis In most of the examined cables the longest observed vented water trees were usually growing from the insulation screen, with maximum lengths as shown in Table. Exceptions were however observed; in Cable 7 a long bow-tie tree was found to bridge the insulation. In Cable and 7 vented water trees growing from the conductor screen were found to bridge the insulation. Two of these trees are shown in Figure 1. Thus, it is demonstrated that both vented and bow-tie trees can bridge the insulation. The density of vented water trees was particularly high for the cables equipped with insulation screen consisting of graphite painting and semiconductive tapes (~0/cm ). The density was lower for the cable with graphited insulation screen (~/cm ). The cables with strippable insulation screens had even fewer trees (~ 0,/cm ). The measurements were performed using the capacitance and tan δ method at voltages up to U o.

4 Laboratory Non-Destructive Diagnostic Testing The results from the condition assessment using the different methods are summarized in Table. Cable and Cable Cable 7 Figure 1 Examples of long water trees bridging the,4mm thick insulation. 7 were diagnosed different by the methods. The frequency domain methods characterized these cables as very bad/ strongly damaged and the time domain methods characterized the cables as good/mid-life. These cables, having the lowest breakdown values and the longest water trees bridging the insulation, were therefore both incorrectly assessed as good by the time domain methods. In order to correctly assess such cables with water trees bridging the insulation using the time domain methods, it is necessary to also measure the polarisation currents [1]. In frequency domain such trees are detected by a strong increase of tan δ at low frequencies at a sufficiently high voltage. The results show that Cable 1 and had high density of long water trees and low breakdown voltages. These cables were correctly assessed as bad by all the diagnostic test methods. Cable, 4 and 6, having the highest breakdown voltages, were relatively correctly assessed by the frequency domain methods. However, the test method based upon return voltage diagnosed Cable incorrectly as strongly damaged, and the IRC-method predicted Cable 4 incorrectly to have breakdown voltages up to 1 U o. The other methods characterized this cable as good with breakdown voltages of about 7U o. Influence of Terminations The results from measurements on the terminations are summarized in Table 4 []. When taking into account the response from the terminations during on-site measurements, it is more convenient to represent the response using a complex capacitance model (the relation between the tan δ and the complex capacitance is tan δ = C''/C'). The measured responses can be classified into two different types: linear or non-linear (i.e. a voltage dependency in C and C ). The terminations equipped with a geometrical field grading are found to have a linear response, while the terminations with a resistive and refractive grading have a non-linear response. Example of non-linear responses of the capacitance and losses is shown in Figure. TABLE - Summary of the cable condition measured by the diagnostic test methods. Destructive methods Non-destructive methods Cable no. 1 Phase Lowest measured breakdown voltage Longest measured vented water tree Depolarisation current Time Domain Return voltage Tan (0,1 Hz) Frequency domain Capacitance and tan (10-0,1 Hz) 1 48 (4Uo) 68 Critical/ 4Uo Damaged Strongly damaged,- Uo 48 (4Uo) 6 Critical/ Uo Damaged Strongly damaged Bad,- Uo 48 (4Uo) 6 Critical/ 4Uo Damaged Strongly damaged Uo (Uo) 67 Mid life/ 1Uo Good Aged 4 Uo 1 (Uo) 100 Mid life/ 1Uo Good Strongly damaged Bad Uo 0 (Uo) Old/ 10Uo Good Aged 4 Uo 1 7 (6Uo) 14 Mid life/ 7Uo Strongly damaged Aged 4, Uo 84 (7Uo) 14 Mid life/ 7Uo Strongly damaged Aged Good 4, Uo 7 (6Uo) 10 Old/ Uo Damaged Aged 4 Uo 1 0 (Uo) 44 Old/ 11Uo Good Aged Uo 6 (6Uo) 47 Critical/ 7Uo Good Good Good 6 Uo 6 (6Uo) 0 Old/ 1Uo Good Good 7 Uo 1 6 (Uo) 77 Critical/ 4Uo Strongly damaged Strongly damaged, Uo 6 (Uo) 86 Critical/ 4Uo Damaged Strongly damaged Bad, Uo 6 (Uo) 6 Old/ Uo Strongly damaged Strongly damaged -, Uo 1 60 (Uo) 49 Old/ Uo Good Aged 4 Uo 7 (6Uo) 44 Old/ Uo Good Aged Good Uo 84 (7Uo) 46 Old/ Uo Good Aged Uo 1 1 (Uo) 100 Old/ Uo Strongly damaged x Strongly damaged 1-, Uo 7 18 (Uo) 87 Old/ 6Uo Good Strongly damaged Bad Uo 1 (Uo) 100 Old/ Uo Good Breakdown xx, Uo Comments: x Breakdown at Uo DC. xx Breakdown at Uo AC.

5 TABLE 4 - Summary of the results from measurements on the terminations. C is the measured capacitance at 1 Hz and 1 kv. C is measured at 0,1 Hz and 6 kv. Termination Type of response C [pf] C [pf] tan δ 1 Linear 17,6 0,47,7x10 - Linear ,8 1,1x10-1 Linear 66, 0,,x10-4 Linear 6,1 0,0 1,1x10 - Linear 17, 0,0 1,x10-6 Non-linear 19, 1,16 6,0x10-7 Non-linear 0, 1,04,4x10-8 Non-linear 6,1 1,,0x10-9 Non-linear,9 7,9 1,4x Non-linear,6,68 4,8x10-11 Non-linear 4,6 0,8 6,1x10 - Some non-linear terminations also have a hysteresis effect similar to that observed for strong water treed cables [4]. However, some differences are observed. In case of terminations the losses increase as the frequency is lowered, while the losses from water treed cable have a frequency independent characteristic. In addition, the response from the terminations is non-linear even at relatively low voltage levels, well below the service voltage. The response from water treed cables with long and few water trees becomes non-linear at a relatively high voltage level. Capacitance [pf] Dielectric losses, C'' Relative capacitance, C'-C 1kV 6 kv 1kV 6kV Applied voltage [kv] Figure The response from termination no. 9. In order to determine the non-linearity of the capacitance (C ), it is presented as a "relative" capacitance (C -C ). Terminations with a linear response usually cause a minor increase of the measured losses of the total cable system. However, in case of termination no., the total losses can be significantly increased due to relatively large values of the capacitance and losses. Therefore, knowledge of the type of response from the most commonly used terminations, makes it possible to estimate the contribution to the measurements, then avoiding wrong interpretation of the cable condition [] On-Site Non-Destructive Diagnostic Testing Examples of the effect of humidity on the measured tan δ are shown in Figure. It is demonstrated that the measured response is strongly increased and voltage dependent during measurements at 100%RH compared to measurements at 7%RH (capacitance and tan δ method). This voltage dependency due to high air humidity could be misinterpreted, as such strong voltage dependence is typical for water treed cables. Similar measurements at 100%RH using the tan δ method at 0,1 Hz (voltages up to U o ) show that the leakage currents across the terminations causing the voltage dependence can be removed by using a guard ring. Dielectric loss tangent, tan δ (0,1 Hz) %RH 7 %RH tan δ method tan δ and capacitance method 4 6 Applied voltage [kv] 100 %RH Figure Influence on the tan by the ambient humidity. It was unfortunately not possible to compare the result from the on-site diagnostic testing with breakdown testing. However, some of the cable owners had a complete service record of their cables. Such records can be used to evaluate the on-site diagnostic testing, as shown in table. 6-9 months after the last measurement two of the cables presented in table were reported to have several breakdowns caused by over-voltages in the network. Both cables were assessed as bad during the on-site test, indicating that the evaluation criteria can be used for such cables. Cables that were classified as good or aged showed generally good reproducibility of the magnitude of the tan δ. However, large changes of tan δ values from one test to the next measured on the same cables were typically detected on bad cables with strong frequency dependence at low frequencies (leakage currents). However, the type of response did not change [4]. This is a typical feature of the dielectric response measured on bad cables with water trees bridging the insulation. It is therefore important to measure the tan δ as a function of frequency in order to get this information. If switchgears can be connected to the cables during diagnostic testing, the time for the test can be decreased. However, it is found that the switchgears contribute 0.01

6 significantly to the total measured losses. It is therefore recommended to disconnect the switchgears before performing the diagnostic testing. TABLE - Results from on-site measurements compared with service experience. Meas U o Prod. year Results Previous faults Good Non Bad Several faults (water tree) Good Non Bad Non* Good Non Good 1 fault (lightning**) Bad previous faults (water tree) Bad previous faults (water tree) /87 Aged Non Comments: *This was the only cable that were diagnosed as bad without a service record of previous faults. ** One fault was recorded (overvoltages generated by lightning strikes). The Database and the Application Program A database structure was specified and programmed using commercial software. In addition, a simple application tool was also developed. The main inputs into the database are: cable history, cable design, type of terminations, type of joints, conditions during measurements (weather, temperature etc.) and results from diagnostic testing (partial discharge, time/frequency domain dielectric response, previous destructive laboratory examinations etc.). In order to verify the applicability of the database, selected measurements from more than 00 field measurements was imported. The experience so far show that the database is well suited for storing diagnostic results and cable information, but a simpler version should be developed for application during on site testing. CONCLUSIONS The results show that the condition of old cables with a high density of vented water trees, are correctly assessed by all the test methods. To detect cables with water trees bridging the insulation, it is necessary to also measure the polarisation currents during the time domain measurements. The results indicate that in order to avoid dielectric breakdown during diagnostic testing the test voltage should be lower than U o. The good correlation between the estimated cable condition and the actual cable condition, indicate that the test voltage can be limited to U o. However, it can become necessary to increase the test voltage to above the service stress (U o ) in order to assess the condition of cables with very few but long vented water trees. For such cables, the evaluation criteria need to be further developed. for water tree aged cables, i.e. voltage dependent response. During diagnostic testing of cables in service, the response from the installed accessories should therefore be known in order to avoid wrong interpretation of cable condition. The on-site measurements have shown that high air humidity can cause misinterpretation of the cable condition due to leakage currents across the terminations. The database for storing diagnostic measurements and information of the cable system can be used for making a more reliable condition assessment. ACKNOWLEDGEMENT The Norwegian part of the work has been performed as a R&D-project in the framework of Norwegian Federation of Utilities (Enfo). Both utilities and Norwegian cable industry have contributed to and sponsored the work. The Danish part of the project was sponsored by the associations of the utilities ELFOR and Sjællandssamarbejdet. The Swedish part has been performed by the Competence Centre in Electric Power Engineering at KTH, Stockhom, sponsored by Elforsk and Swedish industry. The authors would like to express gratitude for their support. REFERENCES 1. S. Hvidsten and J.T. Benjaminsen: Condition Assessment of Water Tree Aged XLPE Cables, Technical Report, TR A180, SINTEF Energy Research, ISBN , 74 pages, August S. Hvidsten, H.Faremo, J.T.Benjaminsen and E.Ildstad: Condition Assessment of Water Treed Service Aged XLPE Cables by Dielectric Response Measurements CIGRE Session 000, Paper 1-01, pp 1-8, Paris, France, August 7 -September 1, A. Avellan, P. Werelius, R. Eriksson: Frequency Domain Response of Medium Voltage XLPE Cable Terminations and its Influence on Cable Diagnostics, ISEI 000, pp , Anaheim CA, USA, April -, P. Werelius, B. Holmgren and U. Gäfvert Diagnosis of Medium Voltage XLPE Cables by High Voltage Dielectric Spectroscopy, ICSD 98, Västerås, Sweden, - June, DEFU Report 447: Diagnostic Testing of XLPE Cables, 71 pages, October, 000 (In Danish). Sverre.Hvidsten@energy.sintef.no Terminations can have a voltage and frequency dependent dielectric response similar to that observed