Partial Discharges at Repetitive Rapidly Changing Voltages

Doctoral thesis, 2009

Electrical insulation can today be exposed to voltages having considerably higher frequency components than the conventional power frequency AC 50 Hz or 60 Hz, for example due to the increased use of power electronics in power systems. This has in a number of cases caused unexpected behaviour, including failures. One factor influencing the insulation is partial discharges (PDs), and the occurrence of these can also be an indicator of the condition of the insulation. There are therefore several reasons why PD measurements at these kinds of voltages are desirable. Measuring PDs at voltages with short rise times is however difficult as there is no longer a large difference in frequency contents between the PDs and the applied voltage waveform. This thesis reports on the development of a PD measuring system intended for use at repetitive, rapidly changing voltages, i.e. square-like voltages, also during the voltage flanks. Numerous experiments as well as theoretical derivations performed showed that the PDs are measurable, using a coupling capacitor and a coupling device, without necessarily requiring high order filters. The measuring system proposed still relies on the difference in frequency contents between the PDs and the applied voltage to certain extent, but also utilizes the stochastic properties of the PDs. Measurements at voltage levels up to 18 kV peak-to-peak with rise times down to 2 μs are reported. Studies on a number of vastly different kinds of test objects are reported, with the purposes of both ensuring a general applicability of the measuring system and increasing the knowledge on PD behaviour. For example, the appearance of corona is studied through analysis of statistical distributions, as function of steepness of the applied voltage. When analysing numerous periods of the applied voltage, it is found that the first PD after the switching event appears with some variation in time. This is an important observation since it implies that if a blind region is present during the flanks, some PDs will still be detected, as long as the blind region is shorter than the time interval where PDs are distributed. However, PDs are found to appear during the voltage flanks in many cases, why it is emphasized that measurements in these critical regions are crucial in order to get the full picture of the PD behaviour. Further, investigations of the PD extinction voltage in different types of test objects reveal that this parameter may be influenced by the rise time of the square-like voltage in some insulation systems whereas not in others. This observation implies that it is important to test insulation for the specific stresses it will be exposed to and such knowledge could ultimately be utilized when designing appropriate insulation systems for specific applications.