Statistical analysis of partial discharges depending on different pressure and voltage rates in high voltage technique

High voltage has a very important role in power systems; not only for energy supply and transmission, but also it is necessary to minimize huge power losses. Installation of high voltage systems needs very excessesive attention and professionality due to high level costs of equipments and potential security risks. We, as electrical engineers, have decided to make researches about high voltage systems because of these advantages and risks as well. Because these features make them extremely important. Faults in high voltage systems are needed to be widely examined and one of them is partial discharges (PD). Partial discharges have very distinctive features. Partial discharges occur in insulation systems of electrical systems and can not be come out exactly. Faults, existing with the partial discharges in complete or partly of insulations systems, affect whole system and may prevent uninterruptable working. In this master thesis, we analyzed partial discharges measurements which were recorded in Stuttgart Institute of Power Transmission and High Voltage Technology. Insulation system of experiment was made of N2 gas in % 99 percentages and SF6 gas in %1 percentage and pressure levels are 1 bar, 2 bars and 3 bars. Experimental set up was made up one plane electrode and one rod electrode. The rod electrode was 100 mm long with the thickness of 0.5 mm. The plane electrode was 50 mm long and the thickness of it was 20 mm. The distance between the electrodes was 20 mm and electrodes were curvated. Experimental set up was cheked out by a partial discharge calibrator (Haefely 451) and a computer via two coupling capasitors. Optical signals were recorded with Omicron MPD 5400 and converted into electrical signals with Omicron MCU 502 fiber optical bus converter. There are five voltage levels measured in 1 bar pressure level. These are 8.2 kV, 10.2 kV, 12 kV, 15.6 kV and 18.3 kV. PD signals distribution according to these voltage levels are shown in Figures 4.9, 4.10, 4.11, 4.12 and 4.13. Similarly, there are six voltage levels measured in 2 bars pressure level. These are 10.93 kV, 12.39 kV, 13.62 kV, 15.38 kV, 18.47 kV and 20.24 kV. PD signals distribution according to these voltage levels are shown in Figures 4.14, 4.15, 4.16, 4.17, 4.18 and 4.19. There are five voltage levels measured in 3 bars pressure level. These are 13.54 kV, 15.5 kV, 16.75 kV, 18.94 kV and 20.54 kV. PD signals distribution according to these voltage levels are shown in Figures 4.20, 4.21, 4.22, 4.23 and 4.24. In order to compare PD distributions in common voltage levels (15 kV and 18 kV) of all three pressure levels, we compose two separate figures (Figures 4.25 and 4.26). 12 kV voltage level is chosen for analysis start point, from which PD`s are respectable. PDs are analyzed both for negative and positive polarities. Partial discharge rising time (Δt) is evaluated in µs and PD amplitude is evaluated in pC. Δt generally exists between 3 and 7 µs and rarely between 30 and 40 µs. Besides, PD values are shown on graphics in every voltage levels. These graphics change 3 to 4 periods depending on experiment duration. According to partial discharge distributions, basic amplitudes, approximately 0.6- 0.8 pC, exist at any time. So, some measurements are evaluated only for PDs which has an amplitude bigger than 2 pC. Besides, in some measurements, basic noise level signals and bigger ones are evaluated in two different categories. It can be said that, partial discharge levels are higher as voltage levels getting higher. As pressure levels getting higher, partial discharges are lately coming out because of the against force. Statistical values (mean, standart deviation, skewness, kurtosis and correlation coefficients) are calculated in order to evaluate both voltage and pressure effects on partial discharge characteristics. Calculated statistical values are expounded and discussed in the last two sections of this study. According to calculated statistical values, it is shown that partial discharges are naturally nonlinear systems. Therefore, statistical distributions of PDs are not symmetrical, but asymmetrical. Asymmetrical distributions have two remarkable parameters. First one is skewness and second one is kurtosis. Skewness values are shown on Figures 5.13, 5.14, 5.15 and 5.16. Kurtosis values are shown on Figures 5.9, 5.10, 5.11 and 5.12 as well. Kurtosis values are higher on 1 bar pressure level than 2 bars and 3 bars pressure levels. This is because, partial discharge levels and occurance times are very different from each other on 1 bar pressure level. Similiarly, skewness values are hihger on 1 bar and 2 bars than 3 bars. The reason is that, measurements on 3 bars are very near to each other and so, standard deviations are smaller on this pressure level. However it is very hard to comment statistical values of high values of Δt times and PD signals from graphics, it can be said that, these datas are comparatively few and very coincidentally. So that statistical graphics are not properly distributed and it is not very easy to make results with them. Mean values of Δt times, are getting smaller as voltage levels are getting higher. Because partial discharge signals are rising up properly as voltage levels get higher. This result has a supportive feature of our experiment. Correlation coefficients are not getting higher in high voltage levels but getting higher depending on pressure levels. In this circumstance, we can say that, partial discharge signals are to be similar at high pressure levels. It is hard to conclude high correlation between 1 and 3 bars on 15 kV voltage level directly. There are several variables both for voltage and pressure levels and partial discharge rise up stochastically as well. So that, these physical events are nonlinear and it is hard to make very certain judgements. A general notion can be put forth only if experiments are widened with more data and parameters. These experiments will be sequences of this master thesis. Partial discharges spectral analysis can be done and this frequency studies may result with skewness parameters of partial discharge signals.