DIELECTRIC HEATING IN INSULATING MATERIALS AT HIGH DC AND AC VOLTAGES SUPERIMPOSED BY HIGH FREQUENCY HIGH VOLTAGES

DIELECTRIC HEATING IN INSULATING MATERIALS AT HIGH DC AND AC VOLTAGES SUPERIMPOSED BY HIGH FREQUENCY HIGH VOLTAGES Matthias Birle * and Carsten Leu Ilmenau University of technology, Centre for electrical power engineering, Research group of high voltage technology, Germany *Email: matthias.birle@tu-ilmenau.de Abstract: In this paper the measuring of the dielectric heating of insulating materials stressed by high direct voltages and high power-frequency voltages superimposed by high-frequency high-voltages (so called mixed-voltages) is described. First results are presented. High-frequency high-voltage stress results in an intense heating caused by higher dielectric losses commensurate to the frequency, the permittivity and the dielectric dissipation factor of the material. The aim of the paper is to make a statement regarding to the influence of superimposed high-frequency voltages on the dielectric heating of nonpolar and polar polymers in steady state operation. Insulating materials in HVDC systems and other power electronic high-voltage systems are stressed by mixedvoltages, high direct voltages, high power-frequency voltages and high-frequency alternating voltages in different forms. To ensure a reliable and sustainable electrical power supply, all components of the systems have to withstand the new demands concerning the dielectric stress. Superimposed alternating voltages are repetitive pulses or a ripple with different frequencies depending on the system configuration and operation mode. For this stress, no design criteria for insulation materials or standards for testing exist. Furthermore commercial test equipment for superimposed voltage tests, especially high-frequency test voltages are not available. The test voltages up to several 10 kv are realized by a test system consisting of an innovative high-frequency highvoltage generator combined with conventional high-voltage AC- and DC- test equipment. A constant sinusoidal high-frequency voltage up to 50 kv in a frequency range from 1 khz to 50 khz is superimposed on a high 50 Hz alternating voltage or a high direct voltage in different ratios. The correct measurement of the mixed-voltages is realized by special high-voltage dividers. The test assembly is made by a temperature decoupled electric field strength control of the samples where the heat transport mechanism is quantifiable. 1 INTRODUCTION The conversion of electrical energy into an optimal form for transmission, storage and consumers usage is done by power electronic components increasingly. Due to the switching operations of the valves, high direct voltages and power-frequency voltages are superimposed by ripple or repetitive transients. These voltage forms are called mixedvoltages. The form and amplitude of mixed-voltage depends on the principle of the inverter, the used filter configuration and the load characteristic [2]. High-frequency high-voltage stress and the associated high power losses can lead to life time reduction respective insulation systems [4]. The worst case is the failure of electrical equipment like the example of the Eagle Pass installation [6]. Experimental investigations of the breakdown of polymers at mixed-voltage stress show that the breakdown voltage is caused by dielectric heating due to the high-frequency voltage components [1]. To describe the power losses at voltages with a plurality of frequencies mathematically, a theoretical approach is presented in [5]. In this paper the temperature measurement of dielectrically heated samples during mixed-voltage stress is presented. Two important facts are in the foreground of the investigations: 1. Generation of mixed-voltages as test voltages with adjustable high-frequency components by the use of a novel high-frequency high-voltage generator. 2. Innovative test assembly for mixed-voltages, but especially suitable for high-frequency voltages, where the temperature of a sample can be measured during voltage stress and heat transfer mechanisms are quantifiable. Due to the use of sinusoidal voltage forms and the special test set up, a mathematical treatment of the measured temperature curves and complex calculation for the dielectric losses are possible. The measured temperature curves at mixedvoltage stress show the influence of the subordinated voltage on the dielectric heating for different polymers. 1025

2 THEORETICAL BACKGROUND The dielectric losses P DL, which are caused by an alternating electrical field E are derived from the electrical equivalent circuit for a lossy dielectric (RC - circuit). For sinusoidal alternating voltages P DL is given in Equation 1 where f = frequency, ε r = loss factor and V = volume. Due to the increasing heating of the sample and the associated temperature gradient, natural convection occurs and the temperature curve approaches to a maximum temperature max (see Figure 1). The steady state is reached, if the generated heat Q is equal to the dissipation heat Q c by convection. (1) The quantity of the heat Q of a sample with the mass m and the heat capacity c at a temperature gradient of is: The heat dissipation by convection is (2) (3) where = heat transfer coefficient, A C = effective surface for convection and t = time. In a test set-up with a sample stressed by an electrical field, where heat radiation and heat conduction dissipation mechanisms can be neglected compared to convection, the following equation is valid: The solution of the differential equation is: With: the temperature function is: ( ( ( ) ) ( ) ) (4) (5) (6) (7) The qualitative temperature function in correlation to Equation 7 is shown in Figure 1. At the time t = 0 the temperature gradient Δ is zero. The sample is at ambient temperature. This case can be approached adiabatically. The temperature curve in this case results in a rising straight and given by Equation 8 (see Figure 1) where the mass density is : (8) Figure 1: Qualitative temperature curve of the heating of a sample, only internal heating and convection 3 GENERATION OF MIXED-VOLTAGES AND VOLTAGE MEASUREMENT The generation of the mixed-voltages to use them as a test voltage is realized by the interconnection of conventional high-voltage test systems with a novel high-frequency high-voltage generator (herein after referred as HFHV-generator) as described in [3]. Figure 2 shows the high-voltage circuit for the generation of a mixed-voltage consisting of a high direct voltage and a highfrequency voltage (herein after referred as DC + HF - voltage) schematically. An example of the generated DC + HF - voltage is shown in Figure 3. Figure 4 shows the high-voltage circuit for the generation of the mixed-voltage consisting of a power-frequency voltage and a high-frequency voltage (herein after referred as AC + HF - voltage). In Figure 5 an example of the generated AC + HF - voltage is shown. To protect the elements of the test systems and to ensure a useful voltage distribution in the circuits, decoupling elements L d, R d and C d are used. The buildup and the function of the HFHVgenerator are presented in [3] detailed. It is based on the principle of resonance and generates a constant sinusoidal high-frequency high-voltage in the range of 1 khz to 50 khz with amplitude up to 70 kv, depending on the load. All capacitances and inductances in the circuit affect the resonant frequency of the HFHV-generator. This results in different frequencies at different high-voltage circuits (Figure 2 and Figure 4). 1026

Figure 2: High-voltage circuit for the generation of a mixed-voltage (DC + HF - voltage), EUT: equipment under test, FM: rotational voltmeter Figure 4: High-voltage circuit for the generation of a mixed-voltage (AC + HF - voltage), EUT: equipment under test Figure 3: Example of generated DC + HF - voltage (2,5 khz) The capacity of the test object has an influence on the resonant frequency, also. Here, the frequency variation due to the different sample materials is only a few 100 Hz. Furthermore, the HFHVgenerator is used to stress the samples with a constant sinusoidal high-frequency high-voltage only. Thus higher frequencies are possible. The used voltage forms and their resulting frequencies are compared in Table 1. Table 1: Voltage forms and resulting frequencies: Voltage form DC + HF - voltage AC + HF - voltage Sinusoidal HF voltage Frequency 8,7 khz 50 Hz + 17 khz 22 khz The precise measurement of the generated mixedvoltages is important for the validity of the investigations. The dielectric heating increases with the square of the voltage (see Equation 1, Chapter 2). An accurate measurement of the voltage has to be guaranteed. For the measurement of high-frequency highvoltages solid-insulated high-voltage dividers are only suitable for transient high frequency signals and not for steady state high frequency signals. Dielectric heating occurs in solid- and liquidinsulated high-voltage dividers also. Thus they can vary their divider ratio in steady state operation. Gas-insulated dividers (compressed gas capacitors) maintain a constant divider ratio in steady state operation at high frequencies, due to the insignificant loss factor of the used gaseous dielectric. Figure 5: Example of generated AC + HF - voltage (50 Hz and 2,7 khz) Another method to measure high-voltages with high frequencies or a plurality of frequencies is to use a divider according to Zaengl. The used selfconstructed Zaengl-divider is characterized by linear transfer behaviour over a wide frequency range so that an accurate measurement of mixedvoltages with several frequencies is possible. Therefore the Zaengl-divider is constructed of lowloss and temperature-stable capacitors and resistors. The current rating of the divider must be designed for the high alternating current (depending on capacity and frequency up to several hundred ma). The measurement of the DC + HF - voltage is performed by a measurement system where the high-frequency high-voltage is measured with an capacitive divider (C meas, Figure 2) and the direct voltage is measured by a rotational voltmeter, which has its upper limit frequency at 20 Hz. Both measurement signals are merged subsequent. 4 TEST SETUP 4.1 Test Vessel The aim of the investigations is to measure the dielectric heating of an insulating material sample during high-voltage stress. The dielectric heating of the sample is take place under two important conditions. First, the dielectric heating is caused by a constant sinusoidal voltage and a homogeneous electrical field. Second, the heat transfer to the ambient gas is the main heat dissipation mechanism and heat conduction in solid materials and radiation can be neglected. Then the equations of Chapter 2 are valid for the measured temperature curves and the temperature curve 1027

shown in Figure 1, Chapter 2 can be expected. No additional heat input into the sample other than the dielectric heating of the sample is allowed. This relates to other insulating materials which can be warmed up by dielectric heating. Partial discharges have to be completely avoided, also. At highfrequency high-voltages they result in an intense local temperature increasing and strong erosion of the sample. Figure 6 shows a sectional view of the test vessel with a sample (schematically) that fulfils these conditions. 1 10 2 6 7 8 Figure 6: Test vessel (schematic, sectional view), 1-sample, 2- stressed region of the sample, 3-shield ring assembly, 4-thin copper electrodes, 5-temperature sensor with earthed screen, 6-feed, 7-bushing (ground), 8- bushing (high-voltage), 9-SF 6 gas, 10-pressure vessel The thickness of the sample over the diameter is not constant. Two thin copper electrodes (4) contact the sample surface at the thicker middle volume (2). These electrodes are necessary for the displacement current of a few 10 to 100 ma at higher frequencies. Thus, the volume (2) in the sample is stressed by a homogeneous electric field. The outer volume of the sample body is thinner. Here, the electrical field stress in the sample body is much lower, since electrical field displacement is occurring in the area between the sample and the shield ring (3). This is caused by the difference between the permittivity s of SF 6 and the solid sample. Thus, there is no heat input from the outer volume of the sample into the stressed volume (2). The triple point (SF 6 -Gas (9), copper electrodes (4) and sample (1)) is optimized by the shield rings (3) so that no partial discharges occur. The shield rings (3) do not contact the sample, since they represent a high heat capacity and heat conduction in the shield rings is not desired. The creepage distance between the two copper electrodes (4) has to be long enough, since the surface resistance is fundamental for the flashover 5 4 9 3 at the DC + HF - voltage. The lower copper electrode (4) is connected with the high-voltage bushing by a thin conductive bar (6). The heat conduction resistance at the contact of the lower copper electrode (4) and conductive bar (6) is very large. The upper copper electrode (4) is connected with the ground side bushing by the shield of the temperature sensor (5). The temperature is measured directly on the surface of the upper thin copper electrode (4). As ambient medium (9) SF 6 gas at a pressure of 3 bar absolute is used. It prevents partial discharges completely and dielectric heating in the gas is not significant. Dielectric solids and liquids as embedding material would cause an additional heat input into the system by dielectric heating and result in a difficult mathematical treatment. The shield ring assembly (3), the shield of the temperature sensor (5) and the bushings (7) and (8) are made of high-grade steel. The thermal conductivity of high-grade steel is considerable smaller compared to copper and aluminium. The heat transport mechanism by thermal conductivity through the solid materials is suppressed. The pressure vessel (10) is made of an insulating material, which has low thermal conductivity and heat capacity. 4.2 Sample materials In this paper, the dielectric heating of polymers is studied. Polymers in power energy technologies are used for cables (cross-linked Polyethylene), capacitors (Polypropylene) and construction elements (Polyamide, Polyoxymethylen, Polytetrafluorethylen, and Polyvinylchloride). They are classified into polar and non-polar polymers. In contrast to non-polar polymers, polar polymers feature orientation polarization. This results in a higher permittivity ε r, a higher dissipation factor tan and thus a higher loss factor ε r '' (ε r '' = ε r tan ). Hence the dielectric losses of polar materials are higher than the dielectric losses of non-polar materials. Table 2 compares the measured materials, and shows the loss factor ε r '' for different frequencies at 25 C. Table 2: Comparison of the measured materials, their polarity and loss factor ε r '' (ε r '' = ε r tan ): Material Polarity Loss factor at 25 C 50 Hz 8,7 khz 17 khz PVC polar 0,186 0,084 0,085 PP non-polar 0,0011 0,001 0,001 PE non-polar 0,0004 0,0004 0,0004 The sample thickness of 3 mm at the homogeneous electrically stressed volume (see Figure 6 (2)) is equal for all materials. The used mixed-voltage configurations with subordinated 1028

voltages in the range of 10 kv to 50 kv lead to electrical field strengths that are comparable to those in power engineering equipment. 4.3 Test procedure First, the dielectric heating at constant sinusoidal voltage without a subordinated direct or powerfrequency voltage for the three different frequencies shown in Table 1 has been measured. Thereafter, the measurements were carried out for mixed-voltage. At mixed-voltages the subordinated voltage was set (power-frequency voltage or direct voltage) and then immediately the high-frequency high-voltage was adjusted to its value within one second. The temperature measurement with a sample rate of 1 Hz starts as soon as the highfrequency high-voltage is increased. Depending on the heating of the samples the thermal steady state is achieved after about 30 minutes. Before a new measurement is performed, the entire system is cooled down to ambient temperature. For statistical confidence the temperature measurements at different configurations where carried out several times. The measured temperature curves show a linear increase which rises exponentially to a final value like the mathematical characterization in Chapter 2. The marks in the diagram do not represent the total number of measurement points per curve. The voltage amplitude at PE compared to PVC and PP is set at 10 kv for all measurements. For smaller voltage amplitudes at PE no precise temperature increasing is measurable. Higher voltage amplitudes of the superimposed highfrequency voltage at PVC and PP have been avoided because of the danger of damaging the sample. The maximum temperatures of PVC are different to the temperatures of PE. This is caused by the fact that PVC is a polar material, orientation polarization occurs and the loss factor is significant higher (see Table 2) 5 RESULTS In Figure 7, the temperature curves of PE and PVC for different frequencies are presented. Figure 8: Measured temperature increase of PE, PP and PVC at DC + HF voltage Figure 7: Measured temperature increase of PE and PVC at constant sinusoidal voltage with different frequencies and amplitudes of 10 kv (PE) and 6 kv (PVC) Figure 9: Measured temperature increase of PE, PP and PVC at AC + HF - voltage The temperature curves at DC + HF - voltage for PVC, PP and PE are presented in Figure 8. The 1029

corresponding temperature curves at sinusoidal voltage with same frequency and same amplitude are shown in the diagram also. It can be seen that the heating of the sample at DC + HF - voltage is the same as for sinusoidal high-frequency voltage stress with same frequency and amplitude. The temperature curves at AC + HF - voltage are shown in Figure 9. It can be seen that in the case of PE and PP for both voltage forms the temperature curves are equal. In contrast, for PVC and mixed voltage configuration (20 kv 50 Hz and 6 kv 17 khz) the maximum temperature is higher compared to sinusoidal voltage with amplitude 6 kv and frequency 17 khz. This is caused by the fact that for PVC a temperature increase at power-frequency is measurable also, due to the high loss factor of PVC at powerfrequency. The power losses of PVC at 6 kv and 17 khz only about fourteen times greater than at 20 kv and 50 Hz (see Table 2 and calculate Equation 1). The maximum temperature of a PVC sample at alternating voltage with an amplitude of 20 kv and a frequency of 50 Hz is 1,2 K (see Figure 9). 6 CONCLUSION In this paper, the dielectric heating of polymers at mixed-voltage stress is presented. The used mixed-voltages are high power-frequency voltages and high direct voltages, superimposed by constant sinusoidal high-frequency high-voltages in the range of several khz. An innovative test assembly was build up where material samples (Polyethylene, Polypropylene and Polyvinylchloride) are stressed by a homogeneous electrical field. A direct temperature measurement of the sample during voltage stress is realized. The peculiarity is that the heat transfer mechanisms of the test assembly are quantifiable, because of the sample geometry, the sample connections, the shield assemblies and the used materials. A thermally decoupled electrical field strength control restricts the heat transfer by conduction in the solid elements. Thus a correct measurement of the actual sample temperature under voltage stress is realized. The reproducible measured temperature curves show the expected and calculable characteristic temperature increase. For mixed-voltage forms (high direct voltage superimposed by high-frequency voltage) with a total electrical field strength of 15 kv/mm the dielectric heating is equal to the dielectric heating of a sinusoidal high-frequency voltage with same frequency and same amplitude like the superimposed voltage. For these electric field strengths, the subordinated direct voltage has no impact on the dielectric heating due to a superimposed high-frequency voltage. The frequency components, the amplitude of the superimposed high-frequency voltage and the loss factor of the materials are decisive for the dielectric heating. At mixed-voltage forms (power-frequency voltage superimposed by high-frequency voltage) and for the polar material PVC, the measured temperatures show, that the configuration of the mixed voltage has an influence on the dielectric heating. The power losses depend on the involved voltage amplitudes and their frequencies respectively. That means, the dielectric heating and the maximum temperature respectively increase with the voltage and frequency of the subordinated and superimposed voltage form. At non-polar materials (PE and PP) this effect was not measureable, because of the small loss factor at power-frequency. For non-polar materials and at mixed-voltages with high-frequency components up to 10 % (application field strengths) a thermal breakdown can be excluded. In this case the temperaturedependent life time reduction is extremely low. 7 REFERENCES [1] Birle, M.; Leu, C.; "Breakdown of polymer dielectrics at high direct and alternating voltages superimposed by high frequency high voltages", 11th IEEE Conference on Solid Dielectrics, July 2013. [2] Cigré Broshure 447 Components Testing of VSC System for HVDC Applications, Working Group B4.48, February 2011. [3] Birle, M; Leu, C.; Bauer, S.; Design and application of a High-Frequency High-Voltage Generator, XVII International Symposium on High Voltage Engineering ISH, August 2011. [4] Koltunowicz, T.L.; Kochetov, R.; Bajracharya, G.; Djairam, D.; Smit, J.J., "Repetitive transient aging, the influence of repetition frequency", Electrical Insulation Conference (EIC), 2011, vol., no., pp.444,448, 5-8 June 2011. [5] Sonerud, B.; Bengtsson, T.; Blennow, J.; Gubanski, S.M.; "Dielectric heating in insulating materials subjected to voltage waveforms with high harmonic content", Dielectrics and Electrical Insulation, IEEE Transactions on, vol.16, no.4, pp.926-933, August 2009. [6] Paulsson L., Ekehov B., Halén S., Larsson T., Palmqvist L., Edris A., Kidd D., Keri A., and Mehraban B., High frequency impacts in a converter-based, back-to-back tie; The Eagle Pass installation, IEEE Trans. Power, Del., vol. 18, no. 4, pp. 1410 1415, Oct. 2003. 1030