Working Group 5 - 25 met at 10:45 a.m. on Monday April 21 with 163 members and guests present. Seven presentations were made.
Bruce Bernstein (Electric Power Research Institute (EPRI)) gave an update on the activities on distribution cables within EPRI. Ten projects have recently been completed, nine of which have been published. The projects along with ongoing and some new projects to begin in 1998 are listed in Appendix 5-A-1 of the printed minutes. It was suggested that some of the completed projects could be summarized at future 5-25 meetings.
Glen Bertini (UtilX) described a project to rejuvenate cables by injection with a silicone fluid, undertaken between April and September 1995 for Entergy in New Orleans. Five failures had occurred at the Entergy project which raised concern that cable injection might have caused a short term negative impact on cable performance. This resulted in an comprehensive analysis of all treated cables.
Over 6 million feet (>1,850 km) of polymeric cables with poor service performance in North America, Europe and Asia had been treated and 99.64% were still in service without further failures. The 0.36% of the cables which had failures after treatment were subjected to further investigation and analysis to determine the causes of failure. Most post-treatment failures occurred shortly after the cables were injected before there was a significant increase in the AC breakdown voltage as shown in Figures 1 and 2 of Appendix 5-B-1 of the printed minutes. The failures decreased with time, consistent with the known gradual increase of AC breakdown strength after treatment.
Three possible reasons were presented for the five failures at the Entergy project. Firstly, the failures had occurred during the peak failure period of untreated cables before the treatment had time to improve the condition of the cables. Secondly, voltage transients, produced during the switching and grounding operations before and after treatment, may have initiated electrical trees thus leading to electrical failure. A high impedance discharge device is now used to reduce the amplitudes of the transients. Thirdly, due to the impregnating procedure in use at that time, uneven penetration of the fluid might have occurred near the dammed ends of the cable. The damming procedure now takes strand compression into account and minimizes the amount of damming compound used for any cable geometry. Mechanical damage caused by pressure used to inject the fluid was ruled out as a cause of damage.
The combination of pre-existing electrical trees and the two months required for the silicone fluid to have a significant impact on the insulation performance suggests that a small percentage of treated cables will fail. Thus treated cables will not have 100% reliability.
The Entergy system consisted of about 304,000 circuit feet (~94 km) of 750 KCM feeder cable installed in the New Orleans area. The failure rate on the complete system was 58 failures/100 miles/year (36 failures/100 km/year) and 97 failures/100 miles/year (60 failures/100 km/year) when only the cables installed prior to 1980 were considered. The national average is 8 failures/100 miles/year (5 failures/100 km/year). Applying these failure statistics to the 15,700 circuit feet (4.8 km) of pre-1980 installed cables that were injected, three cable sections were expected fail in one year; this was the number that failed with an additional failure in two of the sections. The distribution of failures for a section of New Orleans in 1995, shown in Figure 3 of Appendix 5-B-1 of the printed minutes, clearly indicates that the majority of the failures occur in the late summer and fall, the same period as when the treated cables failed in 1995. Thus to avoid failures in newly treated cables, treatment should be carried out between December and May to allow enough time for the treatment to take effect before the fall failure season.
Possible causes to failures just after treatment, shown in Figure 4 of Appendix 5-B-1 of the printed minutes, were investigated in detail. They were considered to be due to either the injection itself or due to surges introduced by the injection treatment. The different causes examined were, (1) due to the injection:- the irregular penetration into the bulk of the insulation of the impregnant, damming effects near the cable ends and the effects of pressure, and (2) due to voltage transients created by:- current and/or voltage interruptions, the discharge of residual space charges during grounding operations, circuit re-energization on no-load and with load, faults, thumping, and lightning. The voltage transients can initiate electrical trees which ultimately result in failure under normal AC voltage.
An experimental program was carried out to determine the short term impact of the injection process on cable performance. Examination of one length of a treated cable that had failed revealed electrical trees indicating that the cable had experienced surges. No unusual distribution of the fluid in the cable was noticed. To examine the effect of impregnation pressure on AC breakdown strength, 40 foot lengths (12m) of service-aged but untreated 25 kV cable from Entergy were subjected to silicone injection at different pressures and the AC breakdown strengths then measured. Some lengths were not treated and served as control specimens. The results are listed in the table in Figure 5 of Appendix 5-B-2 of the printed minutes. The fluid uptake was significantly greater at the higher pressure as can also be seen in Figure 6, Appendix 5-B-2 of the printed minutes. Less compacted conductors did not show this pressure effect. A week after treatment a cable length was characterized as shown in Figure 7, Appendix 5-B-2 of the printed minutes. The AC breakdown data for both phases were combined and plotted in Figures 8 and 9, Appendix 5-B-2 of the printed minutes for each condition of pressure along with the control cables. There was a significantly difference in the AC breakdown voltages between the treated and untreated (control) cables.
The effect of damming was examined by subjecting nine lengths of cable to the same damming procedure as took place in the field and then performing infrared spectroscopic analysis over a period of 19 weeks. The axial distortion with time at 0.5 mm (20 mils) from the conductor shield are plotted in figure 10, Appendix 5-B-3 of the printed minutes. Another length of aged cable from Entergy was cut into shorter 26 foot lengths (8 m) and 7 lengths treated with silicone fluid but not dammed, 7 lengths treated and dammed to put the theoretical fluid interface in the middle of each specimen, and 7 lengths were left untreated to act as control specimens. The treatments were arranged in such a way that AC breakdown tests were carried out exactly three weeks after injection. The results are shown in Figure 11, Appendix 5-B-3 of the printed minutes. There was no significant difference between the results which showed that any negative effect of inadequate filling due to damming is quite small. Infrared spectroscopy of the specimens revealed that there was no discernible correlation between concentration and performance after three weeks, Figure 12, Appendix 5-B-3 of the printed minutes, and a nebulous correlation between the AC breakdown strength and the absolute concentration difference of two measurements 180 degrees apart at 0.5 mm form the conductor shield.
Dean Smith (San Diego G & E) asked if there had been a change in the end devices for larger diameter cables. G. Bertini replied that damming was still being used. To a question by B. Bernstein (EPRI) if ACLT tests had been performed on impregnated cables G. Bertini answered that tests had not been done.
Mark Walton (BICC - MTC) presented an update on an EPRI project entitled ‘ACLT test results on XLPE cables with supersmooth shields’. The project involves the accelerated life tests under various conditions in the laboratory on three different XLPE cables. Runs 1 and 2 were made from the same nominal conductor shield compound at different times and Run 3 had a supersmooth shield, see Appendix 5-C-1 of the printed minutes. Lengths of the cables were installed in 4 utilities to experience field aging and also at an external site at the Marshall Technology center to simulate field aging. A comparison of laboratory, field and simulated field aging would be made and a generalized aging model developed. The tests on the Run 3 cables enabled the evaluation of cables with a supersmooth conductor shield under ACLT conditions and also verified if the aging model developed for the Run 1 and Run 2 cables was valid for the Run 3 cable.
The Run 1 and Run 2 cables have been aged under 9 different conditions listed in Appendix 5-C-1 of the printed minutes, with voltages ranging from 2 to 4 times operating voltage (U0) and cyclic temperatures with maximum values of 60, 75 and 90°C. The aging model developed from the experimental results predicted the characteristic life at 63% probability of a 30 foot length (9 m) would vary from 0.18 years at 4U0 and 90°C to 69.7 years at U0 and 45°C, (see the table in Appendix 5-C-1 of the printed minutes for intermediate values). The Run 3 cable tests were being performed under only 4 conditions: 4U0, 60 and 90°C, 3U0 at 75°C, and 2U0 at 90°C. The time-to-failure tests were complete at both voltages at 90°C but only one cable had failed in each of the two tests at lower temperatures after almost 3 years under test. (Appendix 5-C-2 of the printed minutes).
The conclusions from the tests were:
A. Kong (Pacific G & E) asked if the failure sites had been dissected to reveal the failure mechanisms. M. Walton replied that the cables had not been dissected. S. Chmelyk (AT Plastics) inquired if different XLPE’s had been aged with supersmooth conductor shield to which M. Walton answered that only one type had been examined in his laboratory. F. Kuchta (Pirelli) asked if there would be difference in the results if strand blocking would have been used with a supersmooth conductor shield. The reply from M. Walton was that there was a definite effect of water between the strands and that their model is for a worst case condition.
Steve Szaniszlo (Union Carbide) gave an update on accelerated cable aging of tree retardant XLPE (TRXLPE). Previous presentations have shown that:
The conditions of different accelerate aging tests on TRXLPE cables are listed in Appendix 5-D-1 of the printed minutes. The test conditions include 2 and 4U0 at 30, 45 and 90°C. The objectives of the tests on 15 kV cables at 2U0 with 120 kV transients, 30°C and water in the strands were to clarify the effect of temperature on accelerated aging and also the effect of voltage transients on XLPE, TRXLPE and EPR at temperatures typical of URD cables service conditions. The XLPE and EPR cables had conventional furnace carbon black conductor shield while the TRXLPE cables had a supersmooth conductor shield. Cables were removed after 18, 36 and 48 months of aging and subjected to AC breakdown tests. The results are shown in Appendix 5-D-1 of the printed minutes. Field-aged cables with up to 9 years in service exhibited similar trends for all three materials.
An aging test at 4U0 and 45°C on 15 kV TRXLPE cables with a conventional conductor shield material showed no failures after 1200 days. This test is continuing. In another series of tests on 15 kV TRXLPE cables with conventional and supersmooth conductor shield materials, the test conditions are 4U0 and 76°C (temperature of the conductor in the water, 90°C conductor temperature under the stress cone). After 390 days of aging, one failure has occurred in the TRXLPE cable with the conventional conductor shield at 320 days, and one failure at 369 days in the cable with the supersmooth conductor shield. This compares with a geometric time to failure of 186 days for a 1980’s vintage TRXLPE cable with a conventional conductor shield tested under similar conditions.
The conclusions were:
E. Comata (AT Plastics) asked if all the cable lengths tested had been preconditioned as the residual cross-linking by products would affect the results. S. Szaniszlo replied that the cables in the first series of tests performed at CTL had not been preconditioned but cables in all the other ACLT tests had been subjected to 90°C for 72 hours prior to testing.
Eric Marsden (Nova-Borealis) made a presentation entitled ‘Accelerated Wet Aging of XLPE Cables - Effect of Thermal Stress on Different Insulation Materials’. The presentation addressed the effect of what temperature should be used in accelerated aging tests on medium voltage cables. Two basic concepts are used, cyclic temperature loading up to 90°C is specified in AEIC CS5 (1994) while in Europe lower constant temperature tests are specified (30°C in UK and 50°C in Germany). The question arises of how high temperature will affect the ‘wet’ performance of materials with lower densities and lower melting points (some metallocine compounds have melting points of about 90°C). This paper, presented as Appendix 5-E of the printed minutes, addresses this question.
Three materials were examined, (1) a standard cable grade XLPE, density of 0.922 g/cm3, (2) a linear very low density polyethylene with peroxide and antioxidant added, density of 0.895 g/cm3 and melting point of 86°C, and (3) a linear material polymerized by the same technique as (2) but having the same density and melting point as ‘standard’ XLPE, i.e., 0.922g/cm3 and 116°C respectively. Peroxide and antioxidant were added to the base resin. The choice of materials excluded possible influences from the polymerization techniques on the ‘wet’ aging performance.
Lengths of 15 kV cable manufactured with material (2), density of 0.895 g/cm3 and melting point of 86°C, were subjected to the AEIC CS5 accelerated water treeing test (AWTT), i.e., a voltage of 3U0 and temperature cycling to 90°C for 360 days. Three cable lengths failed after 68, 115 and 285 days of aging, a performance inferior to that of standard XLPE. An examination of a cable length aged for 360 days revealed large bow-tie trees growing from voids up to 20 mm (~1 mil) in diameter; the longest tree was 1.5 mm (60 mils) while the few vented trees were less than 50 mm (2 mils). A comparison with an unaged specimen revealed that the voids had been generated during the aging and formed a halo about 2 mm (80 mils) from the conductor shield.
Tests were also performed on cup-shaped specimens with a uniform field geometry and which had semiconducting material electrodes. The insulation thickness was 0.7 mm (27.5 mils) and the aging conditions were 15 kV/mm (380 V/mil) and temperature cycled to 90°C (16 h on 8 h off) for 3 weeks a diagram of the test specimen along with a table of the number and sizes of the water trees are shown in Appendix 5-E-1 and 5-E-2 of the printed minutes. The largest voids were formed in the material with the lowest melting point. Specimens exposed to the thermal aging but without voltage also revealed voids up to 12 mm (0.5 mil) in diameter.
The results clearly show that when the melting point of a material is close to the maximum allowable continuous operating temperature, mechanical stresses generated by thermal cycling will lead to the formation of voids even in cross-linked materials. The voids are 20 to 30 mm (~1 mil) in diameter and are concentrated in a halo; they have an adverse effect on the ‘wet’ performance of the insulation.
It was concluded that, in order to evaluate the possible effects of thermomechanical stresses that could occur in service, the evaluation of new compounds should be undertaken at the maximum continuous operating temperature, i.e., 90°C for XLPE.
B. Bernstein (EPRI) stated that more data are needed for the new materials. There are results available which shows that the geometric mean time to failure falls more rapidly for cyclic temperature aging than for constant temperature aging at the same maximum temperature as the cyclic aging. E. Marsden commented in reply that voids can be created in insulation if it is not cooled slowly.
Neal Parker (Puget Sound P & L) presented failure statistics that have been accumulated at his utility on their HMWPE, XLPE or TRXLPE, jacketed or unjacketed cables installed on the system. The data are plotted in Appendix 5-F-1 to 5-F-3 of the printed minutes. As can be seen in Figure 1 the number of failures of 1/0 cables has increased steadily from around 700 failures/year in 1987 to about 1300 failures/year in 1995 and remained at that level in 1996. The amount of cable installed increased steadily during the sixties and seventies reaching a peak of about 325,000 feet (100 km) in 1979 before decreasing to about 150,000 feet (46 km) in 1982, as shown in Figure 2. A slight increase in 1983 was followed by a large increase to just under 4,000,000 feet (1230 km) in 1984 and a return to 1,500,000 feet (462 km) in 1985. From 1985 to 1991 there was a gradual increase to about 3,000,000 feet (924 km) and remained constant at that level through 1994 and then decreased to about 2,000,000 feet (616 km) in 1995 and 1996. Higher numbers of failures in some vintage cables has led to increased amounts which have been replaced. For example, more than 65% of the cables installed in 1971 have been replaced, 40% of the 1972 cables, and less than 20% of the cables installed between 1973 and 1978, Figure 3. More than 500, 000 feet (154 km) of the 1979 cable has been replaced representing about 20% of the cable installed in that year. XLPE replaced HMWPE cables in 1981 and jacketed cables were first installed in 1985.
Several factors are involved to decide which cables should be replaced. These include cable manufacturer, number of previous failures, location and importance of the circuit, etc. A points system is used in making the final decision.
When the failure rate data are plotted, the 1971 cables are significantly worse than other vintages of HMWPE cables, peaking at 120 failures/100 miles/year (75 failures/100 km/year) in 1989 compared with about 40 failures/100 miles/year (25 failures/100 km/year) in 1996, as shown in Figure 4. XLPE cables have a failure rate less than 8 failures/100 miles/year (5 failures/100 km/year), below the national average. Figure 5 shows the trend for the average days between failures for the 1971 vintage cables; the first failure occurring after more than 2 years, the next after a further 16 to 18 months, and failures eventually occurring in within 100 days of each other. The cables are not DC tested but thumping is sometimes used.
The failure rates of unjacketed HMWPE and XLPE cables show a seasonal trend increasing in May to reach a peak in August and then decreasing to a steady level by November . All vintage HMWPE cables exhibited a similar trend as shown in Figure 6. The trend for unjacketed XLPE cables was much less pronounced, Figure 7. while there is no seasonal trend for jacketed XLPE cables, Figure 8. The isokauronic level is low and the heaviest load is in winter.
Dean Smith (San Diego G & E) asked when were feeders first installed with jackets and what was the failure rate. N. Parker replied that feeders have been jacketed since 1964 and that the failure rate has been about 1 failure/100 miles/year (0.6 failures/100 km/year) since 1972 for extruded shields and 2 to 7 failures/100 miles/year (1.2 to 4.4 failures/100 km/year) for taped shields (1964 - 1966 installed cables). R. Smith (Kerite) asked what type of insulation was responsible for the failures and could an acceptable failure rate be defined. N. Parker answered that 99% of the failures occurred in the older HMWPE cables and the reliability of this type of cable was not acceptable. However it was difficult to define an acceptable failure rate as it was mainly a customer satisfaction issue. B. Bernstein (EPRI) complemented the speaker for his excellent data and asked when did the changeover to jacketed cables occur. The reply was that PVC jackets have been used since 1986 on 1/0 cables to prevent concentric neutral corrosion and that a switch to LLDPE jackets was made in 1996. C. Katz (CTL) also congratulated the speaker for his excellent data and asked if any differences between the ‘good’ and ‘bad’ performing cables installed in the 70’s had been observed, for example, were there any changes in design (strippable insulation shields were introduced in the mid 70’s) or cable manufacturing? N. Parker stated that there were no clear physical differences between ‘good’ and ‘bad’ cables. Some cables had loose insulation shields but the population of cables with this type of defect was not known. Cables with loose insulation shields usually had water trees growing from the outer shield. A. Kong (Pacific G & E) asked if the HMWPE cables were unjacketed and direct buried to which N. Parker answered that all HMWPE cables were unjacketed and most were direct buried with the rest installed in ducts.
J. Fitzgerald (Okonite) made a presentation entitled ‘Thermal Characteristics of Medium Voltage Insulations’ in which he presented data comparing the physical properties of EPR with XLPE over a range of temperatures. The data, presented in Appendix 5-G-1 to 5-G-4 of the printed minutes, included oxidation induction time as a function of temperature, melting point data by differential scanning calorimetry, tensile stress/strain, flexural modulus, heat distortion, hot modulus, ultimate tensile strength, and elongation creep vs. temperature. The data showed that EPR had greater mechanical strength at temperatures above 90°C. It was concluded that EPR was a suitable material for the operation of cables at temperatures up to 105°C.
S. Szaniszlo (Union Carbide) commented that the presented data were well known, that it was important to have mechanical toughness at lower temperatures, e.g., during installation, and that tests have shown that XLPE cables did not show any deformation at 130°C. J. Fitzgerald replied that when 138 kV XLPE cables were subjected to load-cycling tests at Waltz Mill the copper neutral tapes burst due to the deformation caused by load cycling.
Working Group 5 - 25 met at 3:35 p.m. on Monday afternoon, November 3 with 243 members and guests present. Six presentations were made.
R. Keefe (BICC ITC) made a presentation entitled ‘Performance of New Elastomers in Medium Voltage Extruded Dielectrics’. The paper described recent advances in polymer catalysts (metallocenes) and processing technology that allowed control of the molecular characteristics of polymers. The resulting polymers had specific properties suitable for the power cable industry. Examples were given based on the type and ratio of comonomers used to make the polymer. From a laboratory study to determine the physical, rheological and electrical properties of a range of metallocene polymers, an ethylene-octene copolymer was selected for full-scale cable testing. The paper, which includes the results of the tests, is given in Appendix 5-A-1 to 5-A-13.
Haridoss Sarma (AT Plastics) asked for clarification on the type of systems that were studied. R. Keefe replied that filled materials were used for the cable tests and unfilled materials for the water treeing tests. Unfilled materials allowed the intrinsic properties of the polymers to be investigated.
B. Bernstein (EPRI) gave an update on the present and future activities on distribution cables within EPRI. Six projects are presently underway, three are scheduled to begin in late 1997 or early 1998, and one has been withdrawn. The projects are listed in Appendix 5-B-1. To a question from P. McTigue (Con Edison) about the voltage ratings under consideration in the project on thin-walled cables, B. Bernstein replied that the project was limited to 15 to 35 kV class cables.
S. Boggs (UConn) presented a paper entitled ‘Mechanism for Impulse Conversion of Water Trees to Electrical Trees in XLPE’. A mechanism of how voltage transients can cause failure of water treed XLPE insulation was proposed. The power dissipated, temperature rise, space charge generated, and the increase in electrical field in and around a water tree under impulse-voltage conditions were calculated for different conductivities of the tree channels. It was shown that the high electrical field formed could cause significant power dissipation in the channels. The resulting local increase in temperature caused a large localized increase in pressure within the tree and also a reduction in the yield stress of the polymer. The calculations showed that the water could reach boiling point over a range of four orders of magnitude in the conductivity of the water tree channel. These conditions could generate a cavity within which partial discharges could initiate and an electrical tree formed. The complete paper is given in Appendix 5-C-1 to 5-C-7.
S. Verne (BICC) asked how the chemical nature of the water tree would affect the results. S. Boggs replied that the chemical nature would affect the conductivity of the tree channel. J. Jow stated that there was evidence of 10 nm tracks in water trees according to fluorescence measurements made by C. Mayoux in France. S. Boggs responded by saying that although he was aware of the work showing the presence of the narrow tracks, larger diameter channels had to be used for the calculations.
N. Hampton (BICC Cables) presented a paper, ‘Weibull Analysis – Dealing with Real Data’, a copy of which can be found in Appendix 5-D-1 to 5-D-8. Time to failure or voltage breakdown data are often analyzed using the Weibull statistical distribution. A presentation by the same author at the Fall 1996 meeting stressed (a) the need for a critical assessment of the goodness of fit of a statistical distribution to the data, (b) how the statistical parameters are related failure mechanisms, (c) the interpretation of data, and (d) the number of tests required to achieve a given resolution. In this paper, Dr. Hampton explained how ‘imperfect’ data, i.e., data with termination failures, flashovers or no failures up to the maximum test voltage or time, should be analyzed. The ‘imperfect’ data should not be ignored. More than one failure mechanism usually resulted in a change in slope in the Weibull plot. Care must be taken when analyzing such data. The maximum likelihood method is often used to estimate the statistical parameters. However, this technique is valid for sample sizes of >50. It was shown that, for the small sample sizes typically used in cable tests, linear regression gave a more accurate estimate than the maximum likelihood method.
R. Hartlein (Neetrac) complemented the author for his lucid presentation and asked if there was any limitation on which data should be included or excluded in the statistical analysis, e.g., if a hole was accidentally created in the test cable, should it be considered? N. Hampton replied that one should use one’s judgement about including samples for data analysis. If the failure was caused electrically it should be included.
W. Boone (KEMA-USA) made a presentation entitled ‘Accelerated Aging at 500 Hz, Usable or Fake?’ He briefly reviewed the purpose of accelerated aging, i.e., to intensify specific parameters to accelerate aging in order to reach the same degree of aging in a short time as would be obtained under service conditions. Any test parameter, ‘natural’ or ‘unnatural’, could be used provided the aging mechanism remains the same. In accelerated water treeing tests, the test should, in a reasonable time period, distinguish between good and bad cables. The important parameters involved in water treeing tests are electrical and mechanical stresses, temperature, temperature gradient, chemical condition of the water, and the applied frequency. Two test methods were compared, the Unipede test at 50 Hz for 2 years in tap water at 2.5U0 at 30(C and a 500 Hz test for 4 months at the same voltage and temperature. The test length was 60 m and the cable was preconditioned in water for 1 month at 80(C. After the aging the cables were cut into 5-m lengths and each length subjected to a step voltage test. All lengths must survive 14 kV/mm, >73% to withstand 18 kV/mm and >40% to withstand 22 kV/mm. The breakdown strengths of cables aged for 2 years at 50 Hz and 4 months at 500 Hz were 48.7 and 44 kV/mm respectively. A comparison of laboratory aging at 500 Hz for 3000 hours with service-aged cables is shown in Appendix 5-E-1 to 5-E-3. The advantages of high frequency testing were that the aging mechanism was the same as occurred in service and that the aging time was 6 times shorter. The main disadvantage was that a high frequency generator was required.
S. Szaniszlo (Union Carbide) presented ‘Accelerated Cable Aging Tests: A Perspective’. Two accelerated aging test protocols are presently used in North America, (1) the accelerated water treeing test (AWTT), a fixed time test, followed by breakdown strength measurements, and (2) the accelerated cable life test (ACLT), when the times to failure are measured. The AWTT is also used in Europe and Japan although the test conditions are different. Improvements have been observed in both the AWTT and ACLT test results. These have been attributed to smoother and cleaner semiconducting shields, cleaner material processing, true triple extrusion, and improved quality management. Thus, it is essential to compare cables of the same vintage, with the same semicons and tested at the same time. It was also pointed out that the AWTT and ACLT tests could rank cables differently. AWTT data for two vintage cables are shown in Appendix 5-F-1 to 5-F-2.
The meeting ended at 5:15 p.m.
