Comparing IEC60270 and RF Partial Discharge ...

5 downloads 0 Views 273KB Size Report
1 University of Strathclyde, 204 George St., Glasgow, G1 1XW, UK. 2 Glasgow Caledonian University, 70 Cowcaddens Rd., Glasgow, G4 0BA, UK. *E-mail ...
1

Comparing IEC60270 and RF Partial Discharge Patterns Alistair Reid1*, Martin Judd1, Brian Stewart2 and Richard Fouracre1 2

1 University of Strathclyde, 204 George St., Glasgow, G1 1XW, UK Glasgow Caledonian University, 70 Cowcaddens Rd., Glasgow, G4 0BA, UK *E-mail : [email protected]

Abstract-- Defect classification using partial discharge (PD) measurement has historically been achieved by representing the data in phase-resolved form. This paper presents phase-resolved PD patterns obtained using radio frequency (RF) and conventional IEC60270 techniques for a number of different defects including free particles, protrusions, floating electrodes and voids. Oil, SF6 and resin were used as insulating media. Each PD geometry was placed inside an aluminium test chamber fitted with an internal UHF antenna. Around 2000 RF and IEC PD pulses have been recorded simultaneously along with their phase position for each PD source configuration using a 3 GHz digital oscilloscope. Results indicate that due to the different responses of each system to the original PD current pulse, variations exist in the spread of discharge magnitudes, implying that more accurate defect classification should be achieved if a system is trained exclusively using either IEC or RF data. In other words, if a classification algorithm has been trained using measurements from a conventional IEC system, the accuracy of diagnosis will be reduced if more modern RF sensors are then fitted without retraining the algorithm. Index Terms--Epoxy resin insulation, fault diagnosis, insulation, partial discharges, power system faults, power system monitoring, SF6, UHF couplers.

I. INTRODUCTION

M

EASUREMENT of partial discharge is an established technique, used to assess the deterioration of high voltage insulation systems and is applied in testing a wide variety of components such as bushings, capacitors, transformers and switchgear. Two effective PD measurement methods are conventional measurements as defined in IEC Standard 60270 [1] and radio frequency techniques. Defect classification has historically been achieved by representing the data in phase-resolved form [2]-[5]. The phase-resolved partial discharge (PRPD) pattern consists of PD data recorded over a period of time plotted on a 3-dimensional φ-q-n or φ-Un graph, where φ is defined as the phase angle, q is the apparent charge, U is the energy of the RF pulse and n is the number of discharges. Today, the analysis of such PRPD patterns is a commonly used diagnostic technique for assessing the condition of an insulation system and is employed by most commercially available PD measurement systems. Attempts have been made to automatically classify This research was funded by the EPSRC through grants GR/S86747 (University of Strathclyde) and GR/S86760 (Glasgow Caledonian University).

PD sources based on statistical techniques. Such classification is often based on

Fig. 1. Experimental set up. Test objects are placed within a conducting chamber of side 1 m. RF and IEC PD signals are captured simultaneously along with the phase reference using a 3 GHz bandwidth oscilloscope.

conventional PD measurements. However, given the increasing popularity of RF PD measurements for fault diagnosis on large electrical plant items such as power transformers [6]-[10], questions arise as to whether the same classification techniques can be applied, or whether a system trained using IEC data can classify a PD source using RF data. This is because RF measurements respond to the dynamics of charge motion rather than the integral of the PD current pulse. II. EXPERIMENTAL PROCEDURE RF, IEC and phase data have been captured simultaneously using a 3GHz bandwidth oscilloscope, as illustrated in Fig. 1. Figs. 2 and 3 show two examples of partial discharge test cells. Fig. 3 shows a free metallic particle in SF6. Voltages of up 20 kV are applied to the upper electrodes. The particle is located on a concave aluminium dish which is at ground potential. Other PD source topologies tested include a free particle in SF6, a point-plane configuration in SF6 and a void in resin [11]-[14]. Since each pulse-pair on the respective PRPD plots can be distinguished individually and related back to the original PD current pulse, the results presented represent the first true simultaneous phase-resolved measurement. The phasereference circuit (PRC) produces a ramp output voltage that is proportional to the phase position of the reference wave. This overcomes any phase ambiguity introduced by the short time

2

base on which it is necessary to capture the signals and by the presence of harmonics or voltage spikes in the a.c. supply.

corresponding IEC pulses, the peak values of which are proportional to apparent charge. The last trace shows the output voltage from the PRC circuit which is linearly proportional to phase angle, with 2.56 V corresponding to 360o. III. RESULTS Figs. 5 - 9 compare PRPD patterns produced using PD data measured simultaneously using both RF and IEC techniques. As expected, the floating electrode, Fig. 5, produces high amplitude discharges both in terms of RF energy and apparent charge. Clear differences exist in the two sets of patterns. A large spread exists in the magnitude of the measured RF pulses, whereas the apparent charge values seem to be fairly consistent. In both cases, discharges occur in the first and third quadrants. There is no significant variation in the amplitude or phase position with applied voltage.

Fig. 2. Point-plane test cell containing pressurised SF6.

Fig. 3. Free particle test cell containing pressurised SF6.

Fig. 6 shows the results for a free particle in oil. For both IEC and RF measurements, discharges occur around the zerocrossing points, with the highest pulse density occurring on the zero-crossings. In this case, φ-q-n and φ-U-n patterns are virtually indistinguishable. φ-q-n and φ-U-n patterns for a free particle in SF6 are shown in Fig. 7. As would be expected, discharges are enveloped by the a.c. reference wave. Similar patterns are produced using both measurement techniques. It appears that RF measurements are more sensitive to this particular defect topology since some IEC pulses occurring on the positive half cycle appear to have an apparent charge of 0 pC, whereas the energy of the corresponding RF pulse is finite. This is due to all data being triggered on the RF channel. Results for the point-plane test cell are shown in Figure 8. Due to the electrode configuration, corona discharges appear on the positive half-cycle at inception. Although phase positions are similar for both φ-q-n and φ-U-n patterns, with discharges occurring around the peak of the positive half cycle, the discharge cluster is skewed toward the upper end of the apparent charge scale for conventional measurements, whereas the maximum number of discharges seems to correspond to the mean RF energy value in the φ-U-n graph. Fig. 9 shows φ-q-n and φ-U-n patterns for a resin sample containing an unknown number of voids. As expected, discharges occur mainly in the first and third quadrants. For the φ-U-n pattern, a larger pulse density occurs at the lower end of the RF energy scale.

Fig. 4. Screenshot from the oscilloscope. A sequence of 20 signals is shown on each channel. CH1 – RF pulses, CH2 – IEC pulses, CH3 – phase reference.

A screenshot from the oscilloscope, capturing all 3 data sets, is shown in Fig. 4. The upper trace shows 20 RF pulses captured sequentially. The second trace shows the

Simultaneous measurement allows the option of triggering on either the RF or IEC pulses. A comparison between the two triggering options is made in Figs. 10 - 12. For each of the three examples shown, a lower cut-off point in exists in the measured apparent charge that is directly related to the IEC trigger level. For the point-plane configuration in Fig. 7, although the discharges occur around 90o for both triggering

3

Number of discharges

Number of discharges

arrangements, a higher discharge density can be seen to occur at the lower end of the RF scale and at the upper end of the IEC scale. Similarly, for the void configuration, discharges occur at identical phase positions; however a wider spread in amplitude exists for IEC pulses.

Phase angle (deg)

Phase angle (deg)

Apparent charge (pC)

RF energy (nJ)

Number of discharges

Number of discharges

Fig. 5. (a) IEC and (b) RF partial discharge patterns for a floating electrode in SF6. 20 kV, 4 bar.

Phase angle (deg)

Phase angle (deg)

Apparent charge (pC)

RF energy (nJ)

Number of discharges

Number of discharges

Fig. 6. (a) IEC and (b) RF partial discharge patterns for a 2 mm diameter free particle in oil (L10B), 14 kV.

Phase angle (deg)

Phase angle (deg)

Apparent charge (pC)

RF energy (nJ)

Number of discharges

Number of discharges

Fig. 7. (a) IEC and (b) RF partial discharge patterns for a 2 mm diameter free particle in SF6. 2 bar, 15 kV.

Phase angle (deg)

RF energy (nJ)

Number of discharges

Number of discharges

Fig. 8. (a) IEC and (b) RF partial discharge patterns for a point-plane in SF6. 14 mm tungsten needle, 25 mircon tip radius, 1 bar, 17 kV.

Phase angle (deg)

Apparent charge (pC)

IV. DISCUSSION With the exception of free particles in oil and in SF6 it is clear that the spread in discharge magnitudes varies noticeably between RF and IEC techniques. Given the difference between phase-resolved patterns using the two techniques, classification algorithms which rely on phase-resolved information to identify defects may prove more accurate if trained exclusively on either IEC of RF data. For example, a diagnostic system containing a library of known faults, characterised by PRPD patterns, may produce an inaccurate diagnosis if the library data has been obtained using IEC measurements while the on-line data has been measured using RF techniques. On examination of the results, there is apparently no significant variation in phase position between φ-q-n and φ-Un patterns. This may be due in part to the manner in which the PD signals are triggered. As would be expected on any oscilloscope, signals from both IEC and RF channels are captured when a pulse is triggered on one of the channels. In order to achieve a more rigorous comparison, it may be necessary to devise a measurement setup that has the facility to trigger on either the RF OR the IEC channel. This will be explored further in future. In figs. 10 – 12, differences between RF and IEC channel triggering are most pronounced in the case of the point-plane in SF6. With IEC triggering applied, pulses below the trigger threshold are lost. Although it may be argued that lower magnitude discharges are irrelevant since they are unlikely to damage the insulation system, the nature of the PRPD pattern is changed significantly when triggering on the IEC channel. For the present measurement setup, this change may be more pronounced since IEC signals are taken from the hardwareintegrated output of the IEC measurement system and are therefore monopolar. The trigger level will therefore have a more direct correlation with the absolute magnitude of apparent charge, creating a measurement threshold on the IEC channel. V. CONCLUSIONS

Phase angle (deg)

Apparent charge (pC)

Fig.9. (a) IEC and (b) RF partial discharge patterns for a void in resin. 15 kV.

Phase angle (deg)

RF energy (nJ)

Phase-resolved patterns have been obtained for PD measured simultaneously using RF and IEC techniques for a number of different defects sources including free particles, protrusions, floating electrodes and voids. Simultaneous measurement allows points on the respective patterns to be related back to the same original PD current pulse. Due to the responses of each system, clear differences exist in the phaseresolved data, implying that more accurate defect classification should be achieved if a measurement system is trained exclusively using either IEC or RF data. It has been noted that similarity in phase positions between the respective PRPD plots may be partly due to the necessity to trigger all acquisitions on a single channel. It may be possible to

4

Number of discharges

Number of discharges

overcome this issue by using a measurement system that is capable of triggering all traces on either of the respective channels.

Phase angle (deg)

Phase angle (deg)

Apparent charge (pC)

RF energy (nJ)

Fig. 10. Comparing (a) IEC and (b) RF triggering for a 2 mm free particle in SF6. 2 bar, 15 kV.

Phase angle (deg)

Apparent charge (pC)

RF energy (nJ)

Number of discharges

Number of discharges

Fig. 11. Comparing (a) IEC and (b) RF triggering for a point-plane configuration in SF6. 1 bar, 12 kV.

Phase angle (deg)

Apparent charge (pC)

M. D. Judd, L. Yang, and I. Hunter. Partial discharge monitoring for power transformers using UHF sensors part 2: Field experience. IEEE Electrical Insulation Magazine, 21(3):5–13, May/June 2005.

[8]

L. Yang and M. D. Judd. Propagation characteristics of UHF signals in transformers for locating partial discharge sources. In Proc. 13th Int. Symp. on High Voltage Engineering, Delft, August 2003.

[9]

M. Judd, L. Yang, C. Bennoch, I. Hunter, and T. Breckenridge. Condition monitoring of power transformers using UHF partial discharge sensors: Operating principals and site testing. In Proceedings of Euro Tech Conference, Manchester, November 2003.

[10] M. D. Judd, G. P. Cleary, C. J. Bennoch, J. S. Pearson, and T. Breckenridge. Power transformer monitoring using UHF sensors: site trials. In IEEE International Symposium on Electrical Insulation, pages 145–149, Boston, MA, USA, April 2002. [11] A. J. Reid, L. Yang, M. D. Judd, B. G. Stewart, and R. A. Fouracre. An integrated measurement strategy for simultaneous fault identification: Combining the UHF and IEC60270 techniques. In 14th International Symposium on High Voltage Engineering, page 394, Beijing, China, August 2005.

Number of discharges

Number of discharges Phase angle (deg)

[7]

[12] A. J. Reid, M. D. Judd, B. G. Stewart, and R. A. Fouracre. Frequency distribution of RF energy from PD sources and its application in combined RF and IEC60270 measurements. In IEEE Conference on Electrical Insulation and Dielectric Phenomena, 2006. [13] A. J. Reid, M. D. Judd, B. G. Stewart, R. A. Fouracre, and S. Venkatesan. Identification of multiple defects in solid insulation using combined RF and IEC60270 PD measurement. In 9th IEEE International Conference on Solid Dielectrics, 2007. [14] A. J. Reid, M. D. Judd, B. G. Stewart, R. A. Fouracre, and S. Venkatesan. Correlation between RF energy and IEC60270 apparent charge for selected partial discharge source geometries. In 15th International Symposium on High Voltage Engineering, 2007.

Phase angle (deg)

RF energy (nJ)

Fig. 12. Comparing (a) IEC and (b) RF triggering for a 1.25 mm void in resin.

VI. REFERENCES [1]

IEC International Standard 60270. High Voltage Test Techniques – Partial Discharge Measurements. International Electrotechnical Commission (IEC), Geneva, Switzerland, 3rd Edition, 2000.

[2]

T. Hong, M. T. C. Fang, and D. Hilder. PD classification by a modular neural network based on task decomposition. IEEE Transactions on Dielectrics and Electrical Insulation, 3(2):207–212, April 1996.

[3]

L. Satish and W. S. Zaengl. Can fractal patterns be used for recognising 3-D partial discharge patterns? IEEE Transactions on Dielectrics and Electrical Insulation, 2(3):352–359, June 1995.

[4]

C Heitz. A generalised model for partial discharge processes based on a stochastic process approach. Journal of Applied Physics, 32:1012–1023, 1999.

[5]

R. Altenburger, C Heitz, and J Timmer. Analysis of phase-resolved partial discharge patterns of voids based on a stochastic process approach. Journal of Applied Physics, 35:1149–1163, May 2002.

[6]

M. D. Judd, L. Yang, and I. Hunter. Partial discharge monitoring for power transformers using UHF sensors part 1: Sensors and signal interpretation. IEEE Electrical Insulation Magazine, 21(2):5–14, March/April 2005.