LABORATORY AND ON SITE TESTS FOR THE DETERMINATION OF THE AGEING CONDITION OF MV PILC CABLES Nikola Kuljaca; Sergio Meregalli; Giuseppe Rizzi CESI RICERCA Italy sergio.meregalli@cesiricerca.it ABSTRACT In the major cities in Europe the electrical distribution network was firstly established at the beginning of the past century and it is not unusual to find electrical components having a service life of sixty seventy years, still having a failure rate non dramatically higher than younger components. As a consequence the electrical Utilities are looking for methods/tools able to quantify the real ageing state of equipment and to propose a priority list for their replacement. INTRODUCTION AEM Elettricità supplies electricity to Milan city and the nearby town of Rozzano. AEM Elettricità is part of A2A group SpA. The present dimension of the company was established in 22 when AEM Distribuzione incorporated the remaining supply network and business branch of Enel Distribuzione in Milan. AEM Elettricità manages a MV (Medium Voltage) and LV (Low Voltage) network of 38 km and 576 km, respectively. The number of customers is about 8. with a maximum output power of 154 MW. The medium voltage network is organised in four different voltage levels, namely 23 kv (28 km), 15 kv (8 km), 9 kv (7 km) and 6.4 kv (14 km). The level of 23 kv was chosen for the progressive unification of the MV network replacing the cable of the remaining voltage levels. The 6.4 kv network is the oldest and has been almost completely eliminated while 15 kv network is of marginal consistency. The 9 kv network is also being removed gradually and its replacement will be completed in about years; then it is of particular interest to verify its state of ageing to confirm if the original plan can be maintained or if it is necessary to speed up the replacement to continue managing the network with acceptable failure rate (in line with the targets set by Italian Regulatory Authority for Electricity and Gas). Aim of the research presented in the paper is the determination of the ageing state of 9 kv cable based on onsite and laboratory tests. MAIN CHARACTERISTICS AND CONDITIONS OF 9 kv NETWORK Typically, for 9 kv level, cables with the following main characteristics are used: copper conductors, impregnated paper with a normal/stabilised compound, non-radial electric field, lead sheath and PVC outersheath. It is to be underlined that the insulation is oversized and suitable for Stefano Fratti, Salvatore Pugliese; Cesare Ravetta AEM Elettricità - Italy salvatore.pugliese@a2a.eu voltages up to 15 kv. The average failure rates (expressed as number of failures per km per year) are available for each MV level, and namely: 2.2 for the 23 kv voltage (ex AEM Distribuzione), 9.7 for the 9 kv voltage and 4.6 for the 23 kv voltage managed before 22 by ENEL Distribuzione. Figure 1 shows the failure distribution in 9 kv network in the period of time 25-28. It is shown that approximately 68% originate on the joints, 25% failures are due to cables and the remaining due to different causes. 9 kv Level: Component Failure Distribution 9 kv JOINT 9 kv CABLE 9 kv OTHER CAUSES Figure 1: Failure distribution for 9 kv network The failure causes of on the cables mainly depend on the corrosion of the lead sheath with consequent deterioration of the insulation. The main causes of joints failure are the unsuitable type of filling compound used and the difficulty to control the electric field in the area where the cable three phases open. As mentioned, joints represent the most critical component of the underground 9 kv network. For setting up an optimised maintenance and replacement plan it is therefore important to understand if also the cable is in a very aged state. Replacement of the entire link is generally advisable in case of very aged cable while a proper maintenance to identify the most critical points of the line ad take suitable actions is advisable if the cable condition is still good or at least adequate to withstand network stresses. DIAGNOSTIC TESTS An experimental diagnostic campaign in field and in laboratory was planned with the aim of verifying the condition of network components, identifying the portions of the oldest network that need priority for replacement and verifying the effectiveness of diagnostic techniques to be used on site to avoid interruptions of service to customers. Field tests were based on measurement and location of Partial Discharge (PD) activity and measurement of loss angle (tg). Laboratory tests were performed on samples of
cables removed from service; in addition to the measurements performed on site, tests were carried out using two methods made commercially available in the recent years. These methods are particularly suitable for the diagnostic of paper insulated equipment. The first method is the measurement of Polarisation and Depolarisation Current (PDC) [1]. The system using a model of the tested object evaluates the measurement and provides different diagnostic indicators that can be used for a deep analysis of the insulation and so it allows a more precise assessment of the cable condition. The main test parameters are the voltage to be applied to the test object (in the range 2 kv DC ) and the duration of current measurements (in the range 1s 2s). Current measurement duration shall be optimised: for example setting a measuring time of 2s for both polarisation and depolarisation current will make the test to last more than 11 hours, a duration unsuitable for on site diagnostic of MV cables. The second recently developed method is the Frequency Domain Spectroscopy (FDS) [2], that is the measurement of tg over a wide frequency range (from,1 mhz to 1 khz). The equipment provides a series of sinusoidal voltages at specified frequencies evaluating, at each step, the dielectric response of the test object. The test voltage can be set by the User in the range -14 V (rms values) as well as the frequency range. A sweep over the entire frequency range may take about hours. On site tests About 7 underground lines (for a total length of 6 km) were investigated and the same number will undergo a similar diagnostic campaign in the near future. PD tests were performed using a system supplying a damped oscillating voltage (DOV). The applied voltage was gradually increased from 1 kv up to 2U and PD activity measured at different voltage levels; in case of very high activity the test was stopped at lower voltages. A system delivering a sinusoidal voltage at.1 Hz (VLF) was used for tg measurement. The measurements were performed at three different voltage levels in the range 1U 2U. The results of the tests have shown two cables with very high values of tg and one point (in the cable) generating a very high level of PD, see examples of figures 2 and 3. One joint was immediately fixed, while the other lines identified as very aged have been left in service. After about two years from the diagnostic campaign, two joints and a cable identified as non critical failed. The failure analysis highlighted that water penetration was the cause of the failures. Water penetrated in the cable because of lead sheath corrosion and in the joints because of wrong type of filling compound used that could not guarantee the correct joint sealing. Figure 2: Examples of on site tg measurement Figure 3: Examples of on site PD measurement Laboratory tests The entire 9 kv underground network is based on belted PILC cables having a service age comprised in the range 2-7 years. The design of the cables is expected to be quite similar but basic material (paper and impregnating compound) have changed over a so long time span. Three periods of time were identified as moments of important development of the network: for each period some cable samples (without accessories) were removed from service to be analysed in laboratory. The average length of the samples is about 5 m. Table 1 shows main sample characteristics. The table highlights that the cable removed have a service time of approximately 2-25 years, 45- years and about 7 years respectively. Table 1: main characteristics of the cable samples Sample (n) Installation date/service age (yr) C1 C6 C 1985/22 1983/24 1986/21 C3 C4 C7 C8 1962/45 1964/43 1962/45 1959/48 C2 C5 C9 1938/69 1944/63 1939/68 Several measurements were performed on the removed samples at high and low voltages. PD activity (with both DOV and power frequency PF) and tg with PF and VLF were measured at high voltage (up to 1.5U ), PDC and FDS systems were used at low voltage. Some measurements were performed at ambient temperature as well as at higher temperatures to investigate the related influence. At the end of the diagnostic campaign, breakdown test, visual and chemical analysis on cable components were also performed to have a better insight of insulation real ageing state.
PD measurements resulted in a very limited activity with both voltage stresses; in few cases PD activity was detected but it was verified by an acoustic sensor that weak points were located in the area where the cable three phases open. Tg measurements were performed in two different configurations: the first stressing one phase at a time being the other two connected to the grounded lead sheath, the second stressing the three phases connected in parallel for a total of 4 measurements per sample. Generally no significant differences were evidenced among the different measurements indicating an homogeneous ageing condition of the samples. For sake of simplicity in presenting the results in the following pictures, only one measurement is reported for each cable sample. Figure 4 shows tg values measured at 1U and at 1.5U with PF; it can be noted that independently from service age the measured values are quite low and comparable to those typical of new cables. 12 14 V and at 1 U with PF and VLF As shown in figure, the values measured at PF are very similar. [ -3 ] 4 3 3 2 2 1 Cables C1 and C2: at,1 Hz, Uo and 2Uo 2 3 4 6 7 Temperature [C] C1 3F Uo C1 3F 2Uo C2 3F Uo C2 3F 2Uo Figure 6: VLF voltage, variation of tg with temperature For VLF stress the agreement is quite good in most cases, the higher differences could be justified by small differences on cable temperature during tests. In fact as shown later, the temperature of the sample is a very important parameter especially in the low frequency region for older cables. 6 [ -3 ] 8 6 (5kV) D (5 - kv) [ -3 ] 4 3 2 Hz-FDS Hz-SBr.1Hz-FDS.1Hz-VLF 4 2 1 2 3 4 5 6 7 8 9 Cable Number 1-V 2-B 3-R 4-R 5-V 6-R 7-M 8-G 9-M -B Cable Phase Figure 4: tg and tg values measured at ambient temperature with PF Measurement performed at higher temperatures did not show any important differences respect those at ambient temperature and in some cases tg values decreased as shown in the example of figure 5. tg d [ -3 ] 12 8 6 4 2 Cables C1 and C2: at Uo 2 3 4 6 7 Temperature [C] Limit Figure 5: Variation of tg with temperature at PF Tg measurements at VLF showed values in the range 5 4* -3 at ambient temperature; remarkable differences between recent and older cables were found at higher temperatures as reported in figure 6 for sample 1 (22 years old) and 2 (69 years old). With FDS system tg values were measured at 14 V in a frequency range Hz - 1 khz. Figure 7 reports the comparison of tg values measured at 1R 1B 1V 2R 2B 2V Figure 7: tg measured at 14 V and 1U at PF and VLF Figure 8 shows tg values at three different temperatures. It can be noted that increasing the temperature the curve moves toward the right side without changing its shape according to the theory. 1.1.1 Cable 3: FDS Curves at different temperatures.1.1 1 Figure 8: Variation of FDS curve with temperature Moreover, the figure shows that the minimum value of the curve is comprised in a frequency range from few Hz to about Hz. This feature, constant for all the measured samples, justifies the decrease of tg value at PF when increasing sample temperature as shown in figure 5. Even if the minimum tg values measured for all the samples with FDS system are very similar, the shape of the curves exhibit some differences. In particular, as shown in figure 9, younger sample have flatter curve than older ones. 2C 4C 6C
1 C1 and C9: curves 9 RVM: Max Return Voltage 8.1.1 C1 C9 Voltage [V] 7 6 4 3 2.1.1 1 Figure 9: FDS curves for young and old sample This feature explains the strong increase of tg values of older cables at VLF at high temperature. In fact with young samples (having a flatter curve) the increase is reduced with respect to older samples. For the measurement with PDC system, a voltage of Vdc was chosen after comparison of measurement results at different stress values; a polarisation and depolarisation time of 4 s was chosen, for a total test duration of about two hours. As already mentioned from the recorded currents the system has the possibility to investigate different possible diagnostic indicators and namely: tg values at different frequencies, polarisation index, polarisation spectrum, return voltage parameters, insulation resistance and evaluation of humidity in the insulation system. For all the samples the measurements were performed at ambient temperature only. Figure shows a comparison between tg values measured by FDS system and those calculated by PDC system. tan Ι 1.. Ι.. Fase Polarisation Polarizzazione Fase Depolarisation Depolarizzazione... 1... Frequency [ Hz ] Figure : tg values measured by FDS system and calculated by PDC system The figure shows that the agreement is excellent in the frequency range.1-1hz, while for higher frequencies an important difference can be noted, due to the particular test sequence used by PDC system. The values of the different diagnostic indicators were reported as a function of service age to verify the presence of a correlation. Figure 11 shows the maximum value of return voltage. It can be seen that the voltages relevant to the younger cables have lower values than all the others, while the values belonging to the other two groups are not clearly separated. FDS 2 3 4 6 7 Age [years] Figure 11: maximum value of the return voltage as a function of service age At the end of diagnostic measurements, samples were subjected to the determination of the dielectric residual strength with the aim of finding possible correlation between the values of the diagnostic indictors and the real ageing state of the cables. The residual strength was determined at PF increasing the applied voltage by 5 kv steps (of 15 min each). Five cables (C4, C7, C8, C9 and C) were kept to breakdown. In any case no puncture occurred in the cable itself but in the terminations that, for assembling reasons were the weakest points of the test objects. To avoid the destruction of the samples without obtaining results directly related with dielectric properties of the cable, the breakdown test was then converted in a withstand test, with a minimum requirement higher than typical network overvoltages (for example those caused by failure) previously evaluated with EMTP program. All the remaining cables passed PF (2 kv) as well as the subsequent lightning impulse (8 kvp) withstand test. The results of breakdown/withstand tests are summarised in figure 12 where the maximum power frequency overvoltage expected in 9 kv AEM Elettricità system is also displayed. Voltage [kvrms] 6 4 3 2 C1 C2 Breakdown / Withstand Test C3 C4 C5 C6 Cable Figure 12: PF breakdown (red) and withstand (blue) tests For obtaining further information on the real aging of the insulation system, cables taken to breakdown were disassembled for visual and chemical/physical analysis. In the oldest cable the external armour was heavy corroded but the impervious lead sheath guaranteed a perfect seal against water or other contaminant ingress and inner components were still in good conditions. Figure 13 shows two insulating paper tapes taken from a 21 yr cable (upper) and from a 68 yr old cable (bottom), respectively. C7 C8 C9 C Max P.F. overvoltage
Figure 13: Comparison between paper tapes removed from a 21 (upper) and 68 years old (lower) cables The figure clearly shows the change of colour of the paper tape of the older cable sample. Chemical analyses were performed on small quantities of compound taken from insulation of cables with the aim of identifying the presence of typical products of ageing. At the purpose FT-IR Spectroscopy was used. The aim was to detect the presence of CO bond (at 1693 cm -1 ) that is a product of the ageing process of the insulation [3]. Figure 14 shows the spectra of a young and an old sample. Figure 14: Comparison between young (left) and old (right) cable FT-IR spectroscopy Vertical arrow identifies the presence of peak at 1693 cm -1 on old sample and absence of same peak on young cable. The reference peaks, typical of impregnating compound, are located between 3 and 28 cm -1. Moreover, paper tapes, taken from the belt area of the cables, were used for other tests: determination of water content, degree of polymerisation and the tensile strength. The results obtained are reported in table 2. Table 2: Results of physical/chemical tests Sample Water content (%) Degree of polymerisation Tensile strength (MPa) C4 (43 Y),15,17 44 54 66 84 C7 (45Y),14,16 1324 1328 66 78 C9 (68 Y) 1,42 1,46 46 52 15 29 C (21 Y),,12 89-96 27 73 The table shows that C9 cable, beside the colour of paper tapes, has a higher water content and a lower tensile strength. The very large spread of the tensile strength test results of sample C can be explained by the fact that the papers of this cable were very wrinkled. For the interpretation of the results of the diagnostic measurements it is necessary to have a correlation between the measured values and the real ageing degree of the components (cable in this case). The data available in technical literature are not completely exhaustive neither for the newest diagnostic systems nor for the traditional ones. Taking into account the data found in bibliography and our experience, for each of the diagnostic indicator two levels were identified. Values lower than the minimum indicate a safe situation (green light), values above highest limit indicate a very aged component (red light) and the values between the two limits indicate an intermediate ageing degree (yellow light) requiring deeper analysis or the need of subsequent checks. Figure 15 shows for the tested samples a traffic light like chart reporting the ageing degree derived by the interpretation of each of the diagnostic indicator considered. TanD Hz TanD VLF PDC FDS Vbdo Tamb Hot Tamb Hot Relet Cvlf/cfi Ipol Vmax Trmax Sp Tg min C1 C6 C C3 C4 C7 C8 C2 C5 C9 Figure 15: Chart with the interpretation of the diagnostic measurements The chart is divided into three zones according to the service age of the cable. It can be noted that for each zone different degrees of ageing conditions are identified, but in any case for the youngest cables the green evaluation is prevalent, for the oldest cables several red and yellow evaluations are present and the middle aged cables have an intermediate colour distribution. Considering that no very critical cables were found, it was decided to induce artificial defects on some samples. The defects chosen aimed to simulate real defects and namely the penetration of humidity inside the insulation and leakage of insulating compound. The lead sheaths were cut on two cables, having a service age of about 2 and 7 years, and the samples immersed into water. The terminations were removed from a third sample having a service age of 7 years approximately. The cable was hanged with the removed termination up-side down, in such a way to allow the spilling of impregnating compound. A heating tape was wrapped around the cable to increase the temperature to ease paper draining. Diagnostic measurements were performed at regular interval. Figure 16 shows for sample C1 the variation of tg curve as a function of immersion in water time. The figure shows that after 46 days tg variation in the very low frequency zone is negligible, while it is more consistent for frequencies higher than.1 Hz. In particular, the minimum value is increased confirming that it is correlated to humidity content of the insulation. After 46 days of immersion in water, tg of the immersed cable was measured with VLF at the voltage levels of 6 and 9 kv, the values being 14* -3 and 36* -3 respectively.
Values measured on the same cable and at the same voltage levels before starting the ageing were 8.6* -3 and 9.7* -3, respectively. This seems to indicate the importance of carrying out measurements at different voltages. Increasing the voltage a failure was obtained at a voltage of 12 kv..1.1 Cable C1: variation with water penetration.1.1 1 day 28 days 46 days Figure 16: Tg curve versus time of immersion in water Samples of sound cable having a length of, and 1 m were connected to the immersed cable (about 5 m long) to evaluate the sensitivity of the measuring method to detect short portion of humid cable. Figure 17 shows the results adding m of new cable to C1 sample. The figure reports the FDS curves of the short humid sample (blue line), that of the new m cable (purple line), that of the series of the two cables (green line) and that calculated (red dotted) using the characteristics of the two cables. The picture underlines the influence of the cable length on the sensitivity of the method in pinpointing short aged parts... Tgd as a function of cable lenght.1.1 1 C1 (5 m) BM ( m) C1-BM calc C1-BM Meas Figure 17: Influence of cable length on tg curve Measurements of tg as a function of draining time were performed on cable C9; the results do not show any appreciable variation with respect to initial condition. On the same cable PD measurements were performed after 15 heating cycles with PF and DOV. Figure 18 shows the results obtained with PF stress at U. Figure 18: PD pattern of drained cable The figure shows a clear PD activity typical of internal defects. The maximum level of about 3 pc makes this measurement not very sensitive for on site application. PD measurements performed with DOV on the sample show a similar pattern with a PD level of several thousands of pc. CONCLUSIONS A diagnostic campaign was carried out with the aim of assessing the insulation condition of 9 kv cables of AEM Elettricità. Electrical tests were performed on site and in laboratory. Laboratory tests were carried out on cable samples removed from service and representative of installed population; moreover chemical/physical analyses were performed on insulation system components. The results of the tests have shown that cable samples have still sufficient residual dielectric characteristics to withstand the service stresses. Beside the experience gained has shown that a single diagnostic indicator is not generally sufficient to assess the ageing state of PILC cables and several methods are necessary. For laboratory tests all the methods described in the paper have shown to be useful to assess the ageing state of samples representative of cable network population. Under on site conditions the number of tests must be optimised in order to reduce testing time. On the base of our experience the minimum set of test methods should include the determination of FDS curve and HV tests, namely the measurement of tg values at two voltage levels (e.g. in the range of U ) and the measurement of PD activity. Further activities in laboratory and on site are necessary to confirm the results. ACKNOWLEDGMENTS This work has been financed by the Research Fund for the Italian Electrical System under the Contract Agreement between CESI RICERCA and the Ministry of Economic Development - General Directorate for Energy and Mining Resources stipulated on June 21, 27 in compliance with the Decree n.73 of June 18, 27. Special thanks to Sergio Belvedere and Massimo Garotta for conducting laboratory activities. REFERENCES [1] W. Zaengl, 21, Dielectric spectroscopy in time and frequency domain for HV power equipment (transformers, cables etc) 12 th ISH, Bangalore, India [2] R. Neimanis, January 24, Diagnosis of moisture in oil/paper distribution cables, IEEE Transactions on Power Delivery Vol. 19 N 1 [3] V. Buchholz, 21 Condition assessment of distribution PILC cables, IEEE Electrical Insulation Magazine, Vol 2, pages 1-12