21, rue d’Artois, F-75008 PARIS http : //www.cigre.org A2-115 CIGRE 2012 Ultra high frequency (UHF) partial discharge detection for power transformers: Sensitivity check on 800 MVA power transformers and first field experience D. GAUTSCHI*, T. WEIERS ALSTOM Grid AG Switzerland G. BUCHS, S. WYSS Alpiq EnerTrans AG Switzerland SUMMARY Partial discharge (PD) activity in power transformers is often measured during factory acceptance testing. The conventional measurement according to IEC 60270 is usually applied but the direction in the technology leads towards the UHF method. The latter records partial discharge activity in the frequency range between 100 MHz and 2 GHz. This frequency range offers the advantage that suppression of external noise can easily be achieved. There is, however, no direct correlation between the conventional measurement according to IEC 60270 and the UHF method as used at present. This may be one of the reasons why the UHF method is not yet very common, neither during factory acceptance tests nor for condition monitoring of power transformers. Another issue is that big power transformers are designed according to the customer’s specification, which limits the numbers of each type of manufactured transformer. Also, transformers act as waveguides in the UHF frequency range and reflections and standing waves inside the transformer tank highly depend on its geometry (windings, cores, field deflections etc). That is why every change in the geometry inside the transformer tank results in a change of impedance. The result is a unique transmission behavior for a given frequency range. It is thus necessary for UHF sensors to be recalibrated for every transformer design. The new 400/220 kV, 800 MVA phase shift power transformer poles of the substation Lavorgo have been fitted with state-of-the-art, high sensitivity UHF partial discharge sensors. The sensors put the owner in the position to carry out UHF partial discharge measurements and to benefit from future monitoring solutions. A part of the sensitivity check similar to the one which applies to gas-insulated substations has been used to verify the positioning and the characteristics of the UHF sensors after installation. In addition, partial discharge measurements based on the conventional method and based on the UHF method have been carried out in parallel during the factory acceptance tests. During these tests partial discharges could be observed with both systems. A correlation between the two methods could be established for the specific defect which was found. In addition it was possible to locate the PD source. The transformer poles were then subjected to 120% of nominal voltage for one hour during the commissioning tests on site. During these tests the PD activity was measured again with the conventional and the UHF method. During these tests no significant PD activity was found. Since September 2011 three transformer poles are in operation and connected to a dedicated PD online monitoring system. No significant partial discharge activity has been recorded since energization. *david.gautschi@alstom.com The paper shows that for power transformers a similar procedure for a sensitivity check like for GIS substations may be used. Future technical guides or brochures should address possible implementations of PD sources during routine tests since it is not as simple to install an artificial defect inside a power transformer tank as it is in a GIS. Power transformers are built to operate for several decades. A retrofit of new sensors during service is costly. In addition, new procedures might be used in the future to effectively monitor the PD activity of a transformer and to locate the position of defects with an accuracy of some centimeters. It is thus recommended to install UHF sensors on new power transformers even though a standardized correlation between conventional and UHF PD measurements has not been fully established yet. KEYWORDS UHF Sensors, Partial Discharge Detection, Partial Discharge Monitoring, Condition Monitoring, Power Transformers, Cigré Sensitivity Check 1. INTRODUCTION The measurement of partial discharge activity in power transformers is mainly used during factory acceptance testing. The conventional measurement according to IEC 60270 is usually applied. The UHF method is a state-of-the-art approach for measuring partial discharge activity [1-3]. Special UHF sensors for partial discharge detection typically work in the frequency range between 100 MHz and 2 GHz. This frequency range has the advantage that a suppression of external noise can be easily achieved. There is, however, no direct link between the conventional measurement according to IEC 60270 and the UHF method existing at the moment. Several papers address this issue but no general standard for power transformers has been established so far [4]. This may be one of the reasons why the UHF method is not very common, neither during factory acceptance tests nor for condition monitoring of power transformers. Another issue is that big power transformers are designed according to the customer’s specification, which limits the numbers of each type of manufactured transformer. Also, transformers act as waveguides in the UHF frequency range and reflections and standing waves inside the transformer tank highly depend on its geometry (windings, cores, field deflections etc). That is why every change in the geometry inside the transformer tank results in a change of impedance. The result is a unique transmission behavior for a given frequency range. It is thus necessary for UHF sensors to be recalibrated for every transformer design. The new 400/220 kV and 800 MVA phase shift power transformer poles of the substation Lavorgo in Switzerland have been fitted with modern, high sensitivity UHF partial discharge sensors. The sensors put the owner in a position to perform UHF partial discharge measurements and to benefit from future monitoring solutions. 2. SENSITIVITY CHECK IN THE TRANSFORMER FACTORY 2.1 LOCATION OF THE UHF SENSORS Each transformer pole was equipped with 8 built-in UHF sensors which were installed around the circumference and in different height over ground according to the Figures 1 and 2. The dimension of one transformer pole is: length 11.9 m, width 8 m, height 7.7 m. 1 5 4 1 6 8 3 7 2 Figure 1: Top view of a transformer pole and locations of the UHF sensors 2 3 8 7 6 Figure 2: Side view of a transformer pole and locations of the UHF sensors (5 visible) The sensors themselves are designed as high sensitivity wideband antennas which protrude into the transformer tank. Their sensitivities were measured by using different calibration systems (GTEM and cone calibration system [5-6]) in order to know exactly their phase and amplitude behavior over the frequency range up to 3 GHz. The location of the sensors was chosen in a way that all relevant parts inside the tank like the windings or tap changers can be monitored with at least three sensors in parallel. The sensors are connected with the local control cubicle by high frequency cables. The attenuation and phase shift of the installed measurement cables is known and can be taken into account for the evaluation of PD activity. The sensors have sufficient sensitivity, so available triangulation methods can be applied [1, 7-8]. 2.2 SENSITIVITY CHECK PROCEDURE A standardized sensitivity check procedure for UHF sensors installed inside power transformers is not available at the moment. For gas-insulated switchgear a Cigré recommendation TF 15/33.03.05 [9] exists. That document deals with partial discharge measurements of gas-insulated switchgear and links conventional partial discharge measurements according to IEC 60270 in pC to UHF measurements. According to [9] the procedure consists of two steps: Step 1 – laboratory test as type test: A defect is installed in a GIS near a UHF sensor. Then a power frequency voltage is applied and raised up to a level where a specific partial discharge level is recorded with the conventional method according to IEC 60270. For GIS application this level is 5 pC because this is a widely used maximum allowed level of partial discharge1. In addition to the conventional measurement the spectra of the partial discharge is recorded by a UHF sensor and a dedicated UHF measuring device for example a spectrum analyzer with a low noise amplifier. Then the high voltage is turned off and an artificial pulse with known amplitude and very fast rise time (< 1ns) is applied to a UHF sensor which has to be as close as possible to the position in the GIS layout, where the defect was installed. The amplitude of the injected pulse is then raised until a similar spectrum will be measured at the second sensor like in the previous test where the high voltage was turned on. The amplitude of the pulse is noted. Step 2 – On-site test as a verification test on the completed installation: To check the layout of the sensors - especially the distance to each other - an artificial pulse with the amplitude determined in step 1 and by using the same pulse generator is applied to one sensor of the substation and then recorded on the neighbor sensors. The sensitivity check is passed if a difference in the spectra can be seen on each neighbor sensor when the pulse generator is turned on or off. The step 2 is repeated for each sensor in the substation to check every link. 1 In several cases for example bushings and voltage transformers the limit values differ from the mentioned 5 pC. 2 With the above mentioned test it is possible to determine that the sensors work well and that a partial discharge source of 5 pC can be measured by the UHF sensors. In the case of GIS substations it is easy to perform the above sensitivity check for each type of switchgear due to the fact that normally the number of different switchgears is limited. In the case of power transformer applications this is not true as almost every transformer is a unique design. The installation of a PD source inside the large transformer tank is as well not as easy as in gas-insulated switchgear where the recuperation of gas can be sometimes done in several minutes. Different ideas in this regards have been published in [10] but there should be a standardized procedure where the artificial defect has to be installed as well as which type of defect should be used (for example a protrusion, a moving particle or a spark plug). It would be interesting to place the artificial PD source in an oil valve in order to make its installation and removal as easy as possible. Due to the facts mentioned above it was not possible to perform the step 1 of the sensitivity check in the factory of the transformer manufacturer. To overcome this limit the step 2 of the sensitivity check was performed by the use of different impulse amplitudes between 2 V and 50 V. From the GIS application it is known that the amplitude is in the range of 10 V to 20 V for a detection level of 5 pC. 2.3 RESULTS OF THE SENSITIVITY CHECK According to [9] differences in the amplitude spectra shall be visible at neighboring sensors depending on whether the impulse generator is turned on or off. In order to prove this behavior the coupling between each sensor and the others was measured (in total 28 measurements). The largest direct distance between two sensors was about 12 m. The largest distance between two neighbor sensors was around 6 m. In the Diagrams 1 the amplitude spectra with different injection amplitudes are shown for two widely separated sensors (direct distance 12m). According to the measurements an injection amplitude of 2 V is the lower limit which yields a measurable difference. The spectra of two neighbor sensors with a distance of about 3 m are much higher and presented in Diagrams 2. Compared with the Diagrams 1 we see that the sensitivity of the sensors to the neighbor sensors is very high. Therefore the step 2 of the sensitivity check as per [9] is passed with injection amplitude of 2 V. a) b) c) Diagrams 1: Amplitude spectra between the diagonal sensors 2 and 4 with different artificial impulse amplitudes: a) 2V, b) 10V, c) 50V; blue: background noise, black: spectra of the artificial pulse 3 a) b) c) Diagrams 2: Amplitude spectra between the neighbor sensors 3 and 8 with different artificial impulse amplitudes: a) 2V, b) 10V, c) 50V; blue: background noise, black: spectra of the artificial pulse For the measurements the pulse generator was directly connected to the sensor. The measurements at the neighbor sensors were taken at the connectors inside the control cubicle. A correction of the signals with the attenuation and phase of the cables was not made. HF relay with integrated preamplifyer Spectrum analyzer Figure 3: Connections to the sensors in the control cubicle Figure 4: Measurement equipment used for the sensitivity check of the sensors A similar sensor which is installed inside the transformer poles is as well available for GIS applications. For the GIS sensor an injection level of 10 V is sufficient to be sensitive enough to detect a partial discharge of 5 pC. In power transformer applications the PD limit values according to IEC and according to customer specifications are normally much higher than those for GIS (for example 50 pC instead of 5 pC). We arrived to pass step 2 of the sensitivity check with injection amplitude of 2 V. Therefore the sensitivity of the sensors is high enough to detect partial discharge even at a very low level. During the measurements it was identified that the sensitivity of one sensor was weak because the high frequency cable between the sensor and the control cubicle was too long (more than 20 m). In this case the attenuation of the signal inside the cable was so high that the low noise preamplifier was not able to detect it any more even not at an injection level of 50 V. After a rerouting and shortening of the measuring cable the sensitivity check could be performed at an injection level of 2 V with a signal to noise ratio of 7 dB and at 50 V with a ratio of 30 dB. 4 3. FACTORY ACCEPTANCE TESTS To gain experience with the UHF sensors installed in the transformer poles and if possible to calculate a correlation between the UHF spectra and the partial discharge measurements according to IEC 60270 both measurement methods have been used in parallel during the factory acceptance test at the supplier’s factory (see Figure 5). Conventional PD measurement Transformer pole Spectrum analyzer HF relay with integrated preamplifier Figure 5: Parallel measurement of the PD activity according to the conventional method and by the use of a spectrum analyzer During the measurements eight 20 m long high frequency cables were used to connect the measuring equipment with the connectors in the local control cubicle (see Figure 3). The cables had an overall attenuation of about 4 dB at 650 MHz and were connected to a high frequency relay and a low noise preamplifier. It was therefore possible to switch to different sensors during the test. A measurement was normally taken during 60 seconds at each sensor and this for several times until the end of the test period. The voltage of 1.3·Um/3 was applied to each of the transformer poles over a total time of 60 minutes. During the tests partial discharge was detected. The Diagram 3a) shows the measured UHF spectrum at a voltage level of 1.7·Um/3 = 412 kV and 237.6 kV respectively which resulted in a PD activity of 180 pC and 60 pC respectively. The Diagram 3b) shows the PD spectrum at the same sensor but at the voltage level 1.3·Um/3 = 315 kV respectively 182.2 kV. The measured PD activity according to the conventional method was in this case about 120 pC for the 400 kV side and 40 pC for the 220 kV side. The phase resolved PD pattern measured according to the conventional method is shown in Diagram 3c). 5 a) b) c) Diagrams 3: UHF amplitude spectra during the routine tests measured at sensor 3: a) at a voltage level of 1.7·Um/3 with a PD activity of 180 pC respectively 60 pC and b) at a level of 1.3·Um/3 with 120 pC at the 400 kV side and 40 pC at the 220 kV side. blue: background noise, black: spectra of the partial discharge activity. c) shows the phase resolved PD pattern measured according to IEC 60270 on the 400kV side and at the voltage level of 1.3·Um/3. If we assume that the relation between the UHF amplitude and the apparent charge in pC is linear as described in [4] we can compare the amplitude spectra according to the Diagrams 3a) and b) and the measured PD activity with the conventional method. Therefore 6 dB difference in amplitude in the spectra is correlating with about 60 pC at the 400 kV side respectively 20 pC at the 220 kV side. The signal to noise ratio in Diagram 3a) is about 16 dB at 650 MHz. If we add 4 dB attenuation of the measuring cables we end up with 20 dB. It is therefore assumed that we can reach a sensitivity of about 18 pC at the 400kV side (180 pC divided by the factor of 10) and 6 pC at the 220 kV side when we measure directly at the control cubicle connectors. The sensor 3 where the Diagrams 3a) and b) were taken had one of the longest connecting cables from the sensor to the control cubicle. If we can measure directly at the sensor without an interconnecting cable we can reach a sensitivity of about 10 pC respectively 3 pC for the same PD activity measured during the factory acceptance test and with the same measuring instruments. With dedicated measuring systems or devices and of course depending of the fault location much higher sensitivities are conceivable. The measurements on the eight sensors in parallel gave as well the opportunity to locate the PD activity during the high voltage test. By the comparison of the different spectra by their amplitude and their frequency bandwidth figures the origin of the discharges could be assumed to be near the 400 kV high voltage winding. With dedicated online-monitoring systems a better localization of the defect would be possible even today. 4. COMMISSIONING TESTS After the erection of the transformers at the substation Lavorgo, the sensors have been connected with a dedicated online monitoring device. On site commissioning tests were performed in Mai 2011 by an independent company. The three transformer poles and the reserve pole were tested with 1.2 times the nominal voltage during 60 minutes via the 33 kV tertiary system. A diesel generator which produced 60 Hz was used as a power source. During these tests only small PD activity below 17 pC was recorded and the UHF measuring system which was used as well showed only little activity. 6 5. FIELD EXPERIENCE The normal operation of the three transformer poles began in September 2011. The online monitoring system has been running since then. Until now no critical partial discharge activity has been recorded. 6. CONCLUSIONS The use of the UHF method to detect partial discharge in power transformer applications offers many advantages. The biggest advantage of the application of the UHF method is the possibility to locate sources of PD activity. Different techniques have been proposed and these will be further developed in the future. In addition, the sensitivity of the devices is higher compared to the conventional method according to IEC 60270 and the frequency spectrum and bandwidth can be chosen in a way that the background noise can be suppressed effectively. A correlation between the conventional PD measuring method and the UHF signals for power transformers has not yet been fully established. The Cigré recommendation [9] which applies to GIS might be used in the same way to get a better understanding. Even if different approaches have been made for the installation of an artificial defect [10] there is still a lack of technical documentation that details their implementation during tests. Since power transformers operate for decades and since a retrofit with new sensors during service is costly, it seems adequate today to install UHF PD sensors in new equipment. BIBLIOGRAPHY [1] J. Gui, W. Gao, K. Tan, and S. Gao, “Locating Partial Discharge in Power Transformer by Electrical Method” (Proceedings of the 7th International Conference on Properties and Applications of Dielectric Materials, Nagoya, June 2003). [2] R. Jongen, P. Morshuis, S. Meijer, and J. Smit, “Identification of Partial Discharge Defects in Transformer Oil” (Annual Report Conference on Electrical Insulation and Dielectric Phenomena, October 2005). [3] F. Massingue, S. Meijer, and J. Lopez-Roldan, “Partial Discharge Pattern Analysis of Modeled Insulation Defects in Transformer Insulation” (Conference Record of the 2006 IEEE International Symposium on Electrical Insulation, June 2006). [4] S. Coenen, S. Tenbohlen, S.M. Markalous, T. Strehl, “Sensitivity of UHF PD Measurements in Power Transformers” (IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 15, No. 6, December 2008). [5] D. Gautschi, and P. Bertholet “Design and testing of a novel calibration system for UHF sensors for GIS“ (17th International Symposium on High Voltage Engineering, Hannover, August 2011). [6] D. Gautschi, and P. Bertholet “Calibration of UHF sensors for GIS: Comparison of different methods and testing of a calibration system based on a conical antenna“ (International Conference on High Voltage Engineering and Application, New Orleans, October 2010). [7] S.M. Markalous, S. Tenbohlen, K. Feser, “Detection and Location of Partial Discharges in Power Transformers using Acoustic and Electromagnetic Signals” (IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 15, No. 6, December 2008). [8] J.Y. Kim1, I.J.Seo, B.W.Lee, J.Y.Koo, J.T.Kim, “A chip type UHF sensor applicable to find the PD location in gas insulated transformer” (XVII International Symposium on High Voltage Engineering, Hannover, Germany, August 2011). [9] CIGRE Task Force 15/33.03.05, “Partial discharge detection system for GIS: sensitivity verification for the UHF method and the acoustic method” (ÉLECTRA No. 183, pp. 74-87). [10] S. Meijer, M.D. Judd, S. Tenbohlen, “Sensitivity Check for Radio Frequency Partial Discharge Detection for Power Transformers” (2008 International Conference on Condition Monitoring and Diagnosis, Bejing, China, April 2008). 7