ANALYSIS, SIMULATION AND TESTING OF TRANSFORMER INSULATION FAILURES RELATED TO SWITCHING TRANSIENTS OVERVOLTAGES.

Transcription

1 1, rue d'artois, F Paris Session 00 CIGRÉ ANALYSIS, SIMULATION AND TESTING OF TRANSFORMER INSULATION FAILURES RELATED TO SWITCHING TRANSIENTS OVERVOLTAGES. J.LOPEZ-ROLDAN, H.DE HERDT J. DECLERCQ T.SELS, D.VAN DOMMELEN. M.POPOV, L.VAN DER SLUIS. Pauwels Trafo K.U.Leuven Delft University of Technology BELGIUM BELGIUM THE NETHERLANDS 1 INTRODUCTION Transformers in electrical networks can be submitted to overvoltages with a broad spectrum of frequencies. The relevant standards can only cover a small part of them. Worldwide many transformer insulation failures have been reported caused by switching operations, while those transformers had previously passed all the standard tests and complied to all quality requirements. These phenomena can occur in both distribution and transmission networks. For distribution transformers, an investigation is usually only started in case of repetitive faults [1]. The problem is generally associated to the highfrequency overvoltages produced by the re-strikes and pre-strikes during the opening or closing of a switching device. These phenomena are a inherent to all circuit breakers. Especially the vacuum circuit breaker (VCB), which shows a high ability of interrupting HF currents of several hundreds of khz, may cause switching overvoltages. However the behaviour of the circuit breaker depends on the network itself. Furthermore, the response of each transformer to a wide range of frequencies will be different. Then it becomes an EMC problem: circuit breaker and transformer may be incompatible in a specific network arrangement. In order to study this problem thoroughly, it is required to build a proper analytical model of the principal elements, being the transformer and the circuit breaker. It is also necessary to have knowledge of the material behaviour, i.e. the behaviour of the transformer insulation under the specific HF overvoltages produced. In this paper a study of the interaction between transformers and vacuum circuit breakers is presented. TRANSFORMER MODELS.1 High-Frequency Transformer Models To get an accurate view on the behaviour of a distribution transformer, a detailed high-frequency transformer model is required. Most transformer manufacturers already have these models available for impulse voltage calculations. To simulate very high frequency phenomena, some of these models even work on a turn-to-turn level, thus enabling the prediction of the voltage in every turn of the winding.. Transient Transformer Model To simulate the high-frequency behaviour of a transformer in its electrical environment, a reduced model is required. The general high-frequency models are usually too large to be incorporated in a general system model, but with the appropriate reduction techniques, these models can be reduced to a more convenient size. For efficiency reasons, a second option was selected in this project: a new model was built, aiming for a compromise between calculation time, flexibility and accuracy in the prediction of the first resonance frequencies, as they are typically excited by transient network disturbances. Different modelling techniques can be used. The five main streams in this field of modelling are based on self and mutual inductances, on leakage inductances, on the principle of duality, on transfer function measurements and on electromagnetic field calculations. The duality based model, introduced by Cherry [], represents the leakage fluxes by an inductive polygon. Antwerpsesteenweg 167, B-800 Mechelen, BELGIUM

2 The elements of this polygon can be derived from the corresponding short-circuit inductances, that can be obtained both numerically as well as experimentally. This approach was chosen by Van Craenenbroeck et al. [3], since it is flexible, reasonably simple and has already proven to represent the essential transformer resonances (Adielson et al. [4]). In this model, the high voltage layer winding is broken up into smaller segments. Depending on the required accuracy, this can be done on a per turn basis, on a per layer basis, or with any segmentation in between. Fig.1 shows the principle of the model structure. For sake of clarity the high voltage winding in this figure is only broken up in 3 segments, which is of course inadequate for practical calculations. runs or frequency scans. Extra options are available to investigate the influence of connected cables and surge protection devices..3 Results In fig. - 3 some results are shown for a specific distribution transformer design. [%] LV 0 HV1 0 voltage distribution in the 10 HV layers HV HV3 Fig. 1: Transient model with 4 winding segments The model is built around a leakage inductance polygon, describing the leakage field of the transformer. The actual winding segments are connected to this polygon by means of ideal transformers, allowing the inductive elements to be on a common voltage level, preferable the voltage level of the (arbitrarily chosen) reference winding segment. The leakage inductances are calculated in a four-step process. The first step consists in the calculation of the short-circuit reactances X ij. These are calculated using a modified Rabins' procedure [5]. In a second step the bus impedance matrix Z BUS is determined. The corresponding admittance matrix Y BUS is then obtained by inversion (step three), and finally the leakage inductances are calculated (step four). The non-linear magnetizing inductance is connected at the terminals of the inner low voltage winding, by analogy with the traditional equivalent circuits of a transformer. A parallel resistance is included here as well, in order to represent the frequency-dependent core losses. The model is finally completed with a series resistance, to account for the copper losses in each winding segment, with series capacitances for each winding segment, and with shunt capacitances between adjacent winding segments. The whole modelling process has been integrated in an EMTP [6] preprocessing program. The inputs are the geometric and material properties of the transformer; the output is an EMTP-inputfile to be used in transient Fig. : Calculated voltage distribution (%) in the HV layer winding, at 50 Hz and 34 khz In fig. the calculated voltages to ground in the HV layer winding are plotted for an excitation frequency of 34 khz. Voltages are expressed in percentage of the excitation voltage. The winding contains 10 layers, modelled with 9 nodes each. In most nodes the voltages to ground are lower than in the 50 Hz distribution, but a resonance pattern can clearly be distinguished, with reflections at the end of each layer. [%] khz 50 Hz 34 khz Fig. 3: Calculated voltage differences across the layer insulation of the HV layer winding, at 34 khz The effect of the resonance pattern on the voltage differences between the layers is shown in fig. 3. For each pair of layers the calculated voltage differences along the winding height are plotted. The graph suggests that a systematic failure observed between layers 3 and 4 is caused by this specific resonance pattern at 34 khz.

3 3 VCB MODELLING When dealing with overvoltage estimation and small current switching, the model of the VCB has to include HF reignition components, depending on the properties of the VCB and the surrounding network. The VCB is modelled by means of the: cold withstand voltage characteristic of the VCB, HF quenching capability, chopping current. The cold withstand voltage characteristic of the VCB is a function of the contact distance. One of the parameters that is of influence, is the speed of contact separation and many researchers have investigated the withstand capability by experiments [7]. It is known that the data vary with a statistical distribution. Smeets [8] represented the withstand voltage characteristic with an exponential expression, while Glinkowski et al. [9] showed that the reignition can take place at short gaps (<1 mm), so it is sufficient to use a straight line. This approach is shown in figure 4. The HF quenching capability is defined by the slope of the HF reignited current at HF current zero. Earlier, many authors assumed the slope to be constant, but later it has come clear that the slope depends also on the reignited voltage and that it shows also a time dependent behaviour. The chopping current depends mainly on the contact material, but also the surge impedance of the load side is of influence. In our calculations however, we consider the chopping current constant at 3 A. The characteristics describing whether or not reignition occurs are [9]: ( ) U b = AA t t open + BB (1) ( () ) di / dt = CC t t open + DD where topen is the moment of contact opening. The quantities Ub and di/dt represent the dielectric and arc quenching capability of the VCB respectively. The value of the constants chosen in (1) and () are: AA=1.7E7 V/s, BB=3400 V, CC=-3.40E10 A/s, DD=55E6 A/s. Those are values proposed by Glinkowski et al [9]. However those parameters can be adjusted later using experimental data of the CB, as will be explained further. This simple model is sufficient enough for most of the purposes : it will give a relative estimation of the overvoltages expected when switching a transformer under certain conditions Step-up transformer 3. Shunt capacitor 5. Cable 7. Test transformer 9. Inductive load Step-up Generator 15 kva 0.44/ AC VT 4. VCB 6. HF current probe 8. HF voltage probe 10. HF earthing. Vacuum Circuit Breaker 50nF Test 15 kva 6.6/ nF Fig. 5: Experimental set-up. 4 SWITCHING TESTS In order to have a better understanding of the phenomena and the chance to check the computer models, a series of transformer switching tests were performed. A laboratory set-up was designed to reproduce similar switching re-strikes and fast transients overvoltages as it could be experienced when disconnecting a transformer from the network. The disconnection of the transformer with an inductive load on the LV side was found to cause higher overvoltages and a more severe re-strike chain, than switching the unloaded transformer only. The transformer with inductive load was therefore chosen as the test circuit to be benchmarked, in order to improve the computer model and to study the effects of possible transformer protective devices, such as surge arresters and RC snubbers. A single phase circuit was found to be easier to implement, and sufficient enough to benchmark the models, but it does not allow to reproduce virtual-current chopping phenomena [10]. 4.1 Test set-up: V AA BB t Fig. 4: Straight line model of Ub (t) for a VCB. Figure 5 outlines the circuit arrangement. A step-up transformer (15 kva, 0.44/10.75 kv) is used to rise the low voltage of the mains supply to a more suitable high voltage between 3 and 5 kv simulating the network source voltage. A capacitor of 50 nf is added to keep the source voltage stable during the switching.

4 A commercial vacuum circuit breaker rated 17.5 kv, 150 A and 5 ka is chosen to switch the MV/LV test transformer (15 kva, 6600/69 V). This test transformer was specially designed to be able to measure the voltage at some internal points in the HV winding which otherwise could not be reached in standard oil-filled transformer designs. Extra bushings were internally connected to the beginning and end of the first and second HV layer of the layer winding. A Point-onwave switching relay was designed to command the opening of the circuit breaker so that the opening angle could be chosen. The test transformer is connected by 1 meter single phase cable. A 1nF capacitor is added at the end of the cable in order to represent the capacitance of a longer cable. The LV load of the transformer is a 1.4 mh inductor since inductive loads provide the most favourable conditions for breaker re-strikes to occur. Standard high voltage and current transformers measure the AC voltages and currents. 3 HV Tektronix probes measure the high frequency voltages at transformer HV terminals and Pearson HF current probes measure the current crossing at each side of the breaker. The probes outputs are collected by a 4 channel High Frequency Digital Oscilloscope (Nicolet Pro, 10MHz) PC controlled. 4. Experimental and Analytical results: The first experimental results allowed to fit more closely the specific values for the constants A and B of the Ub/t curve shown in figure 4. A straight line is fitted through the maximum voltage points the re-strike chain as shown in the figure 7. This fitting was repeated on several different measurement results and the difference in slope values was found to be small. However, it can be seen how after about 1 ms the curve starts diverging from a straight-line. Thus, the longer the re-striking time, the more dispersion in the results will be obtained. This method to characterise the re-striking behaviour of a vacuum circuit breaker is found to be more practical than the method described by [7]. There a large number of dielectric tests has to be performed in the HV lab, for every contacts position, in order to find each of the curve points, while the statistical nature of the process will always limit the accuracy of the simulation anyway. a more simple ATP model is used to calculate the voltage at the transformer terminals only: for this purpose the Saturable Component provided by ATP was used [6]. To extend the bandwidth of the model, the capacitances between HV and LV windings and earth were added to the standard model. Ub 35 kv Fig. 7: Measured voltage at transformer terminal. Figures 6 and 7 show the measured voltage at the HV terminal of the test transformer during the opening of the VCB, approximately at the instant of maximum current. The source voltage was 5 kv rms and the maximum overvoltage measured at the transformer terminal was of 35 kv peak (5 pu). Numerous re-strikes can be observed during a period of 1. ms. The frequency of the voltage oscillation after the re-strikes is 900 Hz and is function of the L and C at the load side of the breaker. Due to the limitation of the test set-up, the maximum primary current was of 1 A. This is below the expected current chopping level of the breaker (3 A). Then the current is chopped as soon as the contacts separated. This is why the maximum overvoltage appears when the breaker opens at the peak of the current. For the same opening time of about ±0.5ms around current peak, 8 different openings produced a maximum overvoltage of 5.7 pu, a minimum of 5 pu and an average of 5. pu. The variations are due to the intrinsic stochastical nature of the disconnecting operation. It is not possible to get, for every switching operation, the same values of current chopping, opening time, contact speed and circuit breaker parameters, such as breakdown voltage and HF current quenching capability. Restrikes (V) 50 [kv] 35 0 Ub 40 kv 5-10 Current pulses -5 Fig. 6: Re-striking process during CB opening. Together with the transformer model described above, developed to calculate the internal voltage distribution, [ms].0 Fig. 8: Computed voltages as in figure 7.

5 Figure 8 shows the results of the computer simulation at the same conditions as in figure 7. Here the voltage peak is about 40 kv (5.7 pu). Considering the statistical nature of these phenomena, the calculation results match the experimental results quite well in maximum value, frequency and re-striking time. Deviations are due to the random nature of the phenomena described above, as well as to inaccuracies in the estimation of the circuit parameters, the load current at the moment of the opening, and the approximations made in the circuit breaker model (e.g. the cold withstand voltage curve). 5 TRANSFORMER SURGE PROTECTION The Fast Transients problem is well recognized in motors and generators and the same solutions used to protect motors and generators apply to transformers. Several protecting devices are available: - Surge arresters: Typically ZnO surge suppressors are used. They do not affect the rate of rise of the voltage transient and they do not have any effect on internal resonances. They only work if the surges pass a clamping limit. Transient voltages below this limit are not affected, in magnitude nor in rate of rise. Figure 9 shows the effect of a surge arrester in the switching test described above. - Surge capacitors and Combination RC snubbers: A capacitor will reduce both the surge impedance and the frequency of the oscillation. Some manufacturers add a resistor as well, as a matching impedance to avoid surge reflections at the transformer terminals. Figure 10 shows how re-strikes as in figure 6 are eliminating by incorporating a RC (30Ω, 0.17 μf) to our test set-up. The RC is connected by a short link to the HV bushing of the test transformer. Fig. 9: Voltage clamping effect of a surge arrester. V I CB Opens 33 kv SA clamping (3 kv) Fig. 10: VCB Re-striking is avoided by using a RC -ZORC: It includes a RC snubber and a ZnO arrester in parallel. So it combines together both protective features. V DC Fig. 11: Test circuit with Tesla Transformer 6 HF TESTING OF THE INSULATION It is possible to test in the HV laboratory the strength of insulation structures against repetitive HF pulses over imposed to the AC working voltage. This non-standard application can easily be produced with a circuit containing an extended Tesla transformer. The theory around Tesla transformers is well known and the main basics are described in [11]. Some test circuits containing a Tesla transformer and recently used for material testing are shown in [1] and [13]. The main element of the circuit, shown in Fig. 11, is a Tesla transformer (TT) built with air-coupled coils L 1 - L. Contrary to the traditional use of a TT, producing a high voltage with both a constant high frequency and amplitude, this circuit designed with the same TT principle generates decaying high frequency pulse trains superimposed on the 50 Hz mains voltage. The TT principle is based on energy transmission between two coupled (through air) resonating circuits C 1 - L 1 and C - L. The second capacitor C is the overall capacitive value formed by the capacitors from the measurement circuit C d, C p and C q, the coupling capacitor C 3 and the capacitor formed by the test object (TO). The two circuits are theoretically tuned when (3) is satisfied. The resonance frequency f of the secondary C - L circuit is then (4). L f 1 C1 = L C RDC C1 (3) 1 1 R = (4) π L C 4L R1 TH L 1 L R3 The primary capacitor C 1 is charged by a DC supply to a certain value. By firing a thyristor TH at the appropriate time, the capacitor will be discharged across the primary coil L 1. This initiates an RLC-resonance phenomenon from which only the first half wave cycle is obtained to produce an oscillating damped sine wave in the secondary circuit (L - C ). Only with perfectly tuned circuits there will be an RLC-resonance in the secondary circuit. R C 3 ~ DRIVER ERA-III SCOOP V AC A R d Cd PD Cq C p TO COMPUTER DATA LOG.

6 The 50 Hz main voltage, V AC, is applied to the secondary circuit through a coupling capacitor C 3, which acts like a short circuit for the pulses and like an open circuit for the 50 Hz main voltage. The final wave (combination of pulse and 50 Hz waveform) that will be applied to the TO is also shown in Fig. 11 at the connection point of R d. Measuring the ageing of the test object can be done by looking to the partial discharge (PD) level of the insulation after the application of a determined number of HF pulses to the test object. 7 SUMMARY Numerical and experimental tools have been developed to analyse transformer insulation failures due to switching transients. The two main elements to model are the circuit breaker and the transformer. As an example a simple EMTP model to simulate the re-striking phenomena in a vacuum circuit breaker has been proposed. To estimate the voltage at the transformer bushing a simple -terminal model of the transformer is enough. To calculate the effect of the specific surge inside the HV windings a more detailed model has been described. A series of switching tests on a inductively loaded transformer has been performed. Computer simulation results demonstrate good correlation with the overall statistical results in frequency and amplitude of the surge. Several transformer surge protections have been tested. A surge capacitor at the transformer terminals eliminates the VCB re-strikes. A combination of RC with a surge arrester (ZORC) provides the most complete surge protection. Once the HF voltage distribution inside the insulation is known, it is necessary to forecast the long-term behaviour of the material under repetitive application of those non-standard voltages. This special test can be performed using an extended Tesla transformer circuit. Acknowledgement The authors are grateful to the IWT of the Flemish government for providing financial support for this research. 8 REFERENCES [1] D.J.Clare. Failures of encapsulated transformers for converter winders at Oryx Mine. Elektron magazine, March 1991, pp.4-7 [] Cherry E C, 1949, "The Duality between Interlinked Electric and Magnetic Circuits and the Formation of Transformer Equivalent Circuits", Proc. of the Phys. Soc., Vol. (B) 6, pp [3] Van Craenenbroeck T, De Ceuster J, Marly J P, De Herdt H, Brouwers B, Van Dommelen D, 000, "Experimental and Numerical Analysis of Fast Transient Phenomena in Distribution Transformers", Proc. IEEE/PES Winter Meeting, Singapore, CD-ROM (6P) [4] Adielson T, Carlson A, Margolis H B, Halladay J H, Resonant Overvoltages in EHV Transformers - Modelling and Application, 1981, IEEE Trans. on Power App. and Syst., Vol. PAS-100 No. 7, pp [5] Rabins L, 1956, "Transformer Reactance Calculations with Digital Computers", AIEE Trans. 75, pp [6] Scott-Meyer W. ATP rule book BPA, 1994 [7] Roguski T. A.: "Experimental Investigation of the Dielectric Recovery Strength Between the Separating Contacts of Vacuum Circuit Breaker", IEEE Trans. on PWD, Vol. 4, No., pp , April [8] Smeets R.P.P., et al: "Types of Reignition Following High-Frequency Current Zero in Vacuum Interrupters with Two Types of Contact Material", IEEE Trans. on PS, Vol. 1, No. 5, pp , April [9] Glinkowski M., et al.: "Voltage Escalation and Reignition Behaviour of Vacuum Generator Circuit Breakers During Load Shedding", IEEE PES Summer Meeting, July 8-August , 96 SM 40-8 PWRD. [10] Damstra G.C.: "Virtual Chopping Phenomena Switching Three-Phase Inductive Circuits", Colloquium of CIGRE SC 13, Helsinki, September [11] Heise W., 1964, "Tesla-Transformatoren", Elektrotechnische Zeitschrift, 85 Jahrgang, 1-8 [1] Hardt N., Koenig D., 1998, "Testing of Insulating Materials at High Frequencies and High Voltage Based on the Tesla Transformer Principle", Conf. Record of the 1998 IEEE Int. Symp. on Electrical Insulation, [13] Sels T., Van Craenenbroeck T., Brouwers B., Van Dommelen D., De Ceuster J., "Simulation of Transformer Behaviour subject to Fast Transients using a Tesla Transformer", Eighth Int. Conf. on Dielectric Materials, Measurements and Applications, IEE, Edinburgh, Sept. 000 [14] Popov M., van der Sluis L., Paap G.C.: "Investigation of the Circuit Breaker Reignition Overvoltages Caused by Noload Transformer Switching Surges", Eur. Trans. on Electrical Power, ETEP Vol. 11, No. 6, November/December 001, pp