**6. Discussion**

The results obtained by the analysis of the partial discharge impulse current amplitude, the time difference between subsequent discharge impulses (respectively sequences), and the pulseless partial discharge current allowed a classification of the PD types occurring in SF6-insulated systems under DC voltage stress.

One main factor for the determination of the discharge types was the share between pulseless PD currents and current impulses. As expected from the literature, the pulseless currents (caused by the slowly moving ions) were always lower than the current impulses (caused by the fast moving electrons) (cf. Figure 2). The dependency on the applied voltage was more pronounced for the pulseless currents (Figure 13). At a positive protrusion, the pulseless current incepted at higher voltages than the streamer impulses.

**Figure 13.** Comparison of pulseless PD current and the peak value of current impulses (values from Figures 9 and 12).

The pulseless current was already described by various authors for gas-insulated DC systems [10,11,21], but so far, no detailed description of the occurrence in dependence of the electric field strength, gas pressure, and voltage polarity in addition to the description of the impulse currents can be found. These pulseless discharges were caused by a permanent occurrence of avalanche discharges in front of the protrusion, which formed a stable space charge region. The parallel elapsing avalanche processes could not be measured separately, only the constant ion drift could be evaluated. If this space charge region became instable or an avalanche discharge turned over to a streamer discharge, additional current impulses could be measured. Therefore, these current impulses were superimposed on the pulseless current. Since more discharge avalanches could build up at higher voltages due to the increased critical volume, this space charge region and thus also the pulseless current increased.

According to the evaluation of the data, four different discharge types could be differentiated (Table 3).

The first type occurred at the inception voltage of the positive protrusion independently of the gas pressure; strong PD impulses could be measured, but no pulseless current was present. Hence, a streamer discharge generation took place at the protrusion. Due to the long time differences between subsequent impulses, no stable space charge region could be generated, and therefore, no pulseless discharge was superimposed on the streamer discharges.

The second one, PD current impulses superimposed on a pulseless PD current, could occur at various voltage-pressure-variations. This meant that the charges generated by the streamer impulses led to a stable space charge region in front of the protrusion, which affected the distribution of the electric field strength and therefore the size of the critical volume. In this region, continuous avalanche processes took place, and an equilibrium of charge generation and the drift of ions was reached. If this region became unstable, a streamer discharge could occur, superimposed on the glow discharge.

The third one was special for positive protrusions. A strong first discharge impulse changed the space charge region and therefore the electric field distribution significantly. Subsequent impulses with a minor amplitude took place in the first discharge channel as described in Section 2.

A PD behavior without any current impulses, so-called glow discharge, could only be determined under a gas pressure of 0.1 MPa at a negative protrusion in the investigated pressure-voltage range. This space charge region was very stable, and therefore, the generation and drift of ions were always in equilibrium. It could not be excluded that a pure glow discharge, as for low gas pressures, occurred close to the breakdown voltage even at higher gas pressures since the maximum applicable DC voltage was limited in this investigations.

Combining the described data, a classification of the determined PD types was possible, as well as the description of the transition between different types (Figure 14).

At a positive protrusion, a continuous transition between the PD types took places. This could be a transition between pure impulse currents to impulses superimposed on a pulseless current at lower gas pressures (Figure 14a) or a transition to impulses with subsequent impulses superimposed on a pulseless current for higher gas pressures (Figure 14b). At a negative protrusion in a low pressure environment, the transition to a pulseless PD current was abrupt. No change in the PD behavior could be observed at a negative protrusion at high gas pressures.

Probably a more precise distinction would be possible considering more data points, especially at gas pressures of 0.3 MPa. A defined statement for 0.7 MPa was not possible due to the fact that the breakdown occurred close to the PD inception, and therefore, no measurements were possible at higher voltages than inception voltage in order to protect the measurement devices. However, the results obtained so far allowed us to assume that it was similar to the results obtained at 0.5 MPa.

Even though the presented effective ionization coefficient SF6 rose with enhanced electric field strength (Figure 1), the impulse amplitudes did not always rise in the same manner. Hence, the build-up of space charges influenced the formation of the streamer impulses due to a changing critical volume.

Looking at the gas pressures used in technical applications (*p*SF6 ≈ 0.5 MPa), it was evident that challenges arose during PD measurements under DC voltage stress. In principle, it seemed possible to identify a protrusion in gas-insulated DC systems with the well-known measurement principles based on IEC 60270 or by UHF measurements. it was difficult to distinguish between PD

impulses and noise, e.g., due to the long time differences between subsequent impulses at a positive protrusion or low impulse amplitudes at a negative protrusion. Hence, a low noise environment must be established during PD measurements under DC, to achieve high sensitivity. Further challenges may arise in the precise evaluation of subsequent current impulses with low time differences [9]. It may be beneficial for a meaningful PD measurement to measure and evaluate the pulseless currents as well.

**Table 3.** Classification of the determined PD types and their occurrence in the investigated experimental parameter range depending on the polarity of the protrusion, the applied voltage, and the gas pressure.


\* It could not be excluded that pulseless glow discharges occurred at higher pressures as well, because the measurement range (applied voltage) was limited in this investigation.

**Figure 14.** Transition between partial discharge types in dependence of the gas pressure, voltage polarity, and the applied voltage.
