2004 International Conference on Solid Dielectrics, Toulouse, France. July 5-9, 2004 Comparison of Moisture Contents in PILC Cables Measured by Chemical Titration and Determined from Electrical and Dielectric Measurements V. Buchholz, M. Colwell ', S. Cberukupalli ', J.P. Crine "and R. Keefe ' Powertech Labs Inc., Surrey, BC, Canada 'BC Hydro, Vancouver, BC, Canada 'Consultant, Candiac, QC, Canada 4EPRI, Palo Alto, CA, USA * E-mail :
[email protected] Abstract: Moisture content titration, using the Karl-Fischer method, was performed on small samples of paper tapes taken from thirteen MI-length field-aged distribution Paper Insulated Lead Covered (PILC) cables. It has often k e n suggested in the literature that the moisture content of paper-oil systems could be estimated from some electrical and dielectric measurements, especially from return voltage and from dielectric spectroscopy measurements. We performed these measurements on field-aged PILC cables. We have also performed other non-destnrctive electrical tests: the isothermal relaxation current (IRC) and the LIpATEST leakage current test. In addition, AC breakdown measurements were also performed on the same samples. The electrical techniques ranked the cable's condition consistently and it appears that the water content of the insulation is not the only factor influencing the dielectric losses and depolarization currents. In fact, the predictions of moisture content made by the electrical tests did not agree at all with the moisture content measured by Karl-Fischer titration. The presence of acids due to thermal aging andor pd activity is another parameter affecting the electical life of the examined PILC cables, especially by generating polar compounds. These results strongly indicate that moisture contents in any paper-oil system determined from indirect electrical measurements can be misleading and give incorrect estimation of the real moisture content. It seems much more likely that e ledcal and dielectric measurements are sensitive to all polar products, not only water, especially in aged insulation. INTRODUCTION The condition assessment of PILC cables is a crucial factor for many utilities. One objective of this paper is to present the results obtained with four diagnostic techniques designed for use on in-service cables, but used only to a limited extent on PILC cables [l]. They are the: L l p A T E S F method, isothermal relaxation current (IRC), return voltage method (RVM), and dielectric spectroscopy. In addition, the residual AC breakdown strength ofthe cables was also measured. The tested cables were dissected and the moisture content of the paper tapes was measured by Karl-Fischer titration. Another objective of the paper is to discuss the 0-7803-8348-6/M/$20.00 QZOM IEEE. differences noted between these titration measurements and the predictions made by AC breakdown, RVM and dielectric spectroscopy regarding the moisture content in the cables. These last two techniques are well known and have been recently used to predict the water content in paper-insulated systems, especially in large transformers [2]. Some discrepancies between the results given by the two techniques have led to many, discussions and questions regarding data interpretation [3]. Our own results obtained with PILC cables may shed some light on the origins of those discrepancies and it is hoped that these may apply to other paper-insulated systems. Note also that the diagnostic techniques used here may also be used to assess the condition of XLPE cables. Some of the techniques tested in this work were also used in the field and more detail on their limits and on the experimental procedures used can be found in [I]. CABLE SAMPLES The various field-aged cable samples supplied by several US and Canadian utilities are described in detail in Ref. 1 and are identified by letters D to S. There was a wide variety of cable constructions (sectored and round conductor), voltage ratings (5 to 25 kV), conductor sizes, and ages (from 2 to 69 years in service). In most cases, unfortunately, the loading history of these cables was unavailable. The tested samples were typically 9 m long, except 2 cables identified as P (173 m) and Q (151 m). MOISTURE MEASUREMENTS The moisture content was determined according to ASTM D 3277-95 standard, which depends on solvent extraction of the moisture in cellulose samples at room temperature and Karl-Fischer titration. Extraction is accomplished by stirring the solvent with small pieces of insulation. The paper tape samples were taken in at least three different radial and axial locations in the cables. They were removed from the cables at the last minute to reduce moisture evaporation to the lowest possible level. The samples size was approximately 25mm x 25mm and measurements were performed on three to five adjacent samples (same location) for statistical purposes. Since variations between the contents measured in the adjacent samples were below 20% it was concluded that the mailto:
[email protected] reported values were reliable. The moisture content was rarely constant within the insulation (see Ref. I for details) and Figure 1 shows the maximum value measured in all cables. Cables I and 3 are wet (more than 5 %moisture) and cables D, E and S can be considered as lightly moist (2-3 %moisture). Cables with less than 1% (F, G, H and P) are definitely dry and cables M, N , 0, Q and R are moderately moist. There is no obvious relation between cable age and moisture content. The two older cables have, in fact, low moisture content but they are single phase cables with a round-conductor and very thick insulation. 7 , I 0 D E F G H I J M N O P Q R S cables Figure 1- Maximum moisture content measured in all tested cables. The principal problem with the above moisture content measurements is the fact that it is a destructive technique requiring samples to be taken kom a cable removed from service. This is of limited value for on-site assessment of cable condition. Some of the electrical diagnostic techniques described in the following sections are reported to yield results wrrelaied to the moisture content [2,3,5,6,7]. In order to validate those observations, we will compare the above moisture content with the measured electrical properties using these diagnostic techniques. ELECTRICAL MEASUREMENTS LIpATEST and IRC Results LIpATEST is a Powertech in-house technique based on the leakage current flowing in a dielectric under the application of a high dc voltage. The automated setup applies various dc voltages from 4 to 20 kV in steps of 4 kV for a duration of 1 minute. We have shown in Refs. 1,4 that the current measured during LIpATEST is directly related to the capacitance, i.e. the length, of the cable. We have used the measured currents and normalized this to the cable capacitance. In a similar manner, since the maximum electric field at a given voltage was different in each cable because of the various conductor sizes. we have used the calculated maximum field applied during voltage application as the other normalizing parameter. Detailed results can be seen in [l] and in all cases, the leakage current increases rapidly between 0 and 1 kV/mm and then the current saturates between 1 and 2 kV/mm. Phase C of cable J is a notable exception since the current abruptly increases above - 2 kV/mm, whereas the current in the other phases follow the same trend as the other cables. We have plotted in Fig. 2 the normalized currents at lkV/mm and at 3kV/mm. Comparing Figures 2 and 1, it is obvious that the LIpATEST currents do not correlate with the moisture content of the tested PILC cables. f 100 s e 80 4 60 p 40 5 4 0 U E 0 .- 2 20 D E F G H I J M N O P Q cables Figure 2- Normalid LIpATEST results. The IRC test method measures the relaxation current during 30 minutes following the application of a dc voltage of 1 kV (see Ref. 1 for details). Seba KMT, the instrument manufacturer suggests plotting results as Ix t vs. log t. When this is done, we observed the same trend as with the LIpAâIEST measurements under 1 kV/mm (Fig. 2), i.e. no correlation with measured moisture content in the insulation. Return Voltage Method (RVM) As for the other electrical test, the RVM gives results dependent on the sample length. Oetjen and Kamenka have shown [5] that the influence of the length of the cable dimmishes with its increase in length and becomes negligible for cable lengths over 18Ometers. We have normalized the results by dividing the measured maximum recovery voltage by the length of the sample (Fig. 3). Once again, there is no correlation with moisture in the PILC cables tested (Fig. I). L a, p 50 40 2 E30 = 20 10 E z E & L t Cable RVM(from ratio of slopes) 0 Dielectric Karl-Fischer Specboswpy Titration (%) (from min. tan 6) (measured) D E F G H I J M N O P Q cabks Figure 3- Maximum return voltage per unit length. The ratio QW of the absolute return voltages, when measured at 1 & 2 kV over the entire test time of 30 minutes may provide B good indication of any decrease in the insulation resistance [5 ] . If we expect tha$ resistance is essentially dependent on moisture content, the measure of Qla should reflect the moisture concentration in the insulation. However, no direct relationship with the measured moisture contents was observed (Fig. 4) with this parameter. 3 2.2 7, .? 1.8 n S 1.6 m 1.4 .- I 2 1.2 Figure 4- Ratio of RVM slopes at 1 and 2 kV as a function of the moisture content measured by Karl-Fischer titration. Dielectric Spectroscopy Results It was suggested that the moisture content in PILC cable insulation can be estimated from the tan 6 measurement at low frequency using the following expression [6]: mc(%) = 15.3 + 2.53'ln(tan8J (1) where tansmm is the minimum value of measured tans in the frequency range IO-' to IO' Hz (see Refs. 1 and 6). All tested cables show a minimum loss from which moisture was estimated according to Eq. ( I ) . The estimated values are shown in Table 1 where it appears that the suggested relationship between measured moisture content and minimum tan 6 does not hold at all. Thus, dielectric We have run Fourier Transform Infra Red (FTIR) spectra of droplets of some oil samples and we have discovered the presence of acid groups in the oil of cables D, G and J. One likely cause would he a large pd activity, which is common to many PILC cables. The presence of wax between the paper tapes gives further support to this contention but this remains to be demonstrated. In order to validate the detrimental influence of the aging by-products of oil on the cable properties, we have measured the dielectric losses of small oil droplets extracted from the insulation of some cables. It clearly appeared (see Ref. 1) that the losses in the tested cables were directly related to the losses in the oil. In other words, the main degradation process of the examined PILC cables does not seem to he associated to moisture in paper but rather to the aging of oil. Much work remains to be done to identify the processes and the generated products. AC Breakdown and Water Ingress Tests AC breakdown of PILC cables has been assumed, for a long time, to depend on the insulation's moisture content. We have raised an applied AC voltage in steps equal to the operating voltage until the cable failed. Figure 5 shows the withstand field for various cables. Again, comparing this data with Fig.] does not reveal an obvious influence of humidity on the breakdown strength. The only possible exception is cable J but the low strength could also be due to the large amount of wax found between the paper tapes [I]. D I r Y 1 x-0 P L) R 8 Figure 5- Minimum AC breakdown field passed by the various cables in the raise-and-hold test In order to c o n h the detrimental effect of water on the electrical properties of aged PILC cables, we energized one 15 kV cable (P) in the bottom of a tank filled by Im of water under the l i e voltage and under a normal load current (8 h on, 8h om. To accelerate the ingress of water, we made a large artificial slit (1 cm x 12 cm) through the lead sheath of the cable. At regular time intervals, the return voltage and the LIpATEST current were measured. The LIpATEST current did not change vay significantly and the return voltage change was even smaller. Breakdown finally occured after 74 days of direct exposition to water, which is much longer than we expected. The dissection of the failed cable revealed the presence of a large amount of wax between the paper tapes suggesting that one likely cause of failure might have been overheating. Thus, the influence of water on PILC aging and failure is not obvious although it is well known that moisture will reduce the strength and increase the conductivity and losses of new paper. This suggests that aging induces a polarization process much more significant than polarization due to moisture. DISCUSSION AND CONCLUSION The evaluation of the moisture content in the insulation from the eleclrical diagnostic techniques is definitely much more complicated than originally expected. The moisture content deduced 6um RVM and dielectric spectroscopy does not agree with Karl-Fischer titration of the same samples. In addition, the two electrical techniques produced conflicting results, which gives further support tu our contention that they are sensitive to other phenomena and not moisture alone. One may speculate that the above conclusions could equally apply to the present simplistic estimation ofmoisture content in transformer's paper tapes using the same diagnostic techniques. Note also that the relation between the AC breakdown results and the moisture content (Fig. 5 ) is not obvious at all. Bessei and von Olshausen [7] have concluded that PILC cables soaked in water would remain dry as long as they are kept under load (which implies a thermal gradient driving water out of the insulation). The tank test results are in excellent agreement with this statement. Therefore, the main polarization process in field-aged PILC cables is due to something other than water. This also implies that PILC cables kept under load will not be significantly affected by water ingress. Finally, all the cables were ranked consistently by the various electrical tests (see Ref. l), confirming that these techniques are all sensitive to the same effects and could be useful as diagnostic techniques. ACKNOWLEDGMENTS We thank all the utilities that supplied cable samples for this study. REFERENCES [I] "Assessment of PILC Cable Condition from Electrical, Chemical and Metallurgical Tests", V. Buchholr S. Cherukupalli, M. Colwell, J.P. Crine, A. Rao, EPRl Report No. 1001724, Palo Alto, 2003. [2] Tettex Instruments. "Polarisation Spec- Analysis fur Diagnosis of Insulation Systems". Information 29, TI29-d/e-04, 1992. [3] U. Gafvert et al., "Dielectric Spectroscopy in Time and Frequency Domain applied to Transformers", 6& ICF'ADM Conf., Xian, pp. 825-30, June 1999. [4] J.P. Crine, V. Buchholz, M. Colwell, S . Cherukupalli and B. Bernstein, "Influence of Aging on Polarization Currents and Return Voltages Measured in PILC Cables", 2001 CEIDP,pp. 188-191,0ct.2001. [5] H. Oetjen and D. Kamenkq "The Return Voltage: a Non-deshuctive Diagnostic Method for Oil Impregnated Paper Insulation Systems", EPRI Workshop on PILC Cables, Edison, NI, May 2001. [6] R. Neimanis, "On Estimation of Moisture Content in Mass Impregnated Cables", Ph.D Thesis, Technical University of Stockholm, lSSSNl100-1593, Feb. 2001. [7] H. Bessei and R. von Olshausen, "Water ingress and moisture effects on MIND cable with respect to the application of polymeric accessories", 2"d Int. Qnf. on Power Cables and Accessories, IEE Pub. No 270, 1986.