On-line signal analysis of partial discharges in medium-voltage power cables

Partial discharges are symptomatic of many degradation phenomena in power cables and may cause further deterioration of the insulation in many cases. Electrical im- pulses, generated by partial discharges, travel towards the cable ends, and can there be detected using appropriate sensors. Continuous monitoring of the insulation con- dition can be achieved by on-line detection and location of partial discharge (PD) signals. An important aspect of such a diagnostic is the analysis of on-line measure- ments. The research reported in this thesis is aimed at analysis of PD signals from on-line measurements and location of discharge sites. Signal analysis depends on knowledge of both signals and disturbances that are to be expected. To that end, characteristics of PD signals in medium voltage cables are studied in this thesis. The result of this study is a signal model of the propagation path between the discharge site and the sensors. The model accounts for cable sections with di®erent properties, and incorporates the propagation channel load impedances, i.e. the equipment to which a cable is terminated in an on-line situation. The exact propagation properties and load impedances depend on the speci¯c cable connection under test, and are unknown a priori. For this reason, research is conducted on meth- ods that enable experimental characterization of the parameters, by evaluating the response of the cable to applied transients. The presented methods rely on the ex- traction of pulses that are re°ected on impedance transitions within the cable system under test. On-line ¯eld measurements are corrupted by noise and interference, which impede PD signal detection and location. Generally, narrowband interferences resulting from radio broadcasts dominate the measurements, thus prohibiting data-acquisition trig- gered by PD signals. Broadband background noise is present within the entire PD signal bandwidth, and therefore poses a fundamental limit on PD signal analysis. Generally, existing extraction techniques for PD signals only partially exploit a priori knowledge of both signals and interference. In this thesis, matched ¯lters are ap- plied that are derived from the signal model, and are optimally adapted to the signals that can be expected. Besides signal extraction, matched ¯lters provide a means to estimate the PD magnitude and the signal arrival time. Likewise, discharge location methods based on the signal model are proposed, resulting in optimal location esti- mators. Computer simulations illustrate the e®ectiveness of the proposed algorithms and show that the attainable accuracy can be speci¯ed by theoretical bounds. Accurate PD location relies on estimation of the di®erence in arrival times of signals originating from the same discharge. In case of on-line detection, the cable is connected to the grid, and signals are not necessarily re°ected at the cable ends. Therefore signal detection at both sides is generally required for the purpose of dis- charge location. Synchronization of the measurement equipment is achieved using pulses that are injected into the cable connection. Finite-energy disturbances, such as PD signals that originate outside the cable connection under test, frequently occur in on-line situations. Since measurements are synchronously conducted at both cable ends, pulses originating within and outside the cable can be distinguished by examining the di®erence in time of arrival. Moreover, in many situations, the signal direction of arrival can be determined by detecting pulses in two di®erent current paths at a cable termination. This method is applied as an additional technique to discriminate PD signals and disturbances. Based on the results of research, a measurement system is proposed, which enables automated on-line PD detection and location in medium voltage cable connections. The conceptual design is validated by experiments, and the results demonstrate that the practical application is promising.