| dc.description.abstract | The power electronic equipment needed in HVDC grids produces an AC ripple voltage with amplitude of around one-hundredth of the main DC voltage. The high voltage can result in partial discharges (PDs), which are a sign of defects and degradation in the insulation material system of the HVDC grid components. The present work aims to simulate and measure how a sinusoidal AC voltage ripple influences the PD process in dielectric bounded cavities.
This thesis describes a new stochastic Monte Carlo model proposed for the PD sequence at DC voltage and at combined DC and AC voltage, and provides statistical tools for a systematic extraction of physical discharge parameters from the PD data. The stochastic variables are the time lag and the phase-of-occurrence. The physical parameters in the model are the Townsend coefficient, the mean statistical waiting time, the capacitance in series with the cavity, and the rate of increase of the voltage over the cavity. The outputs for the Monte Carlo model are the individual discharge magnitude, the separation time, and the phase-of-occurrence. Individual discharges can be filtered out by setting a discharge magnitude detection threshold, emulating the conditions during the measurement of PD. A new statistical estimator is proposed, which performs the extraction of the physical parameters from PD data by the method of moments.
From model prediction and the measured PD data it was found that the main effect of adding a sinusoidal ripple to the DC voltage is a higher discharge magnitude and longer discharge separation time, compared to ideal DC voltage conditions. This was attributed to the increase in the time lag when an AC ripple is superimposed on the DC voltage. The increase in the time lag could be described mathematically using the concept of duty cycle, defined as the fraction of one AC voltage time period in which the cavity voltage is above the critical voltage.
The physical parameters could be estimated using the measured PD data, without resorting to simulations and the trial and error approach. It was found that the presence of a parasitic AC ripple voltage leads to a significant bias in the estimated parameters extracted by DC voltage estimation methods. | en_US |
