J. Veen

Effect of cable load impedance on coupling schemes for MV power line communication

Coupling of carrier wave frequencies up to 95 kHz (within the European CENELEC A-band) for online diagnostic data transfer in medium voltage cables is studied. Inductive and capacitive signal coupling is considered not only on basis of technical performance, but also on basis of practical aspects. The effectiveness of coupling schemes depends on the impedances of substation equipment at the cable terminals. The frequency response of a 10-kV, 400-kVA three-phase cast resin distribution transformer is investigated. In the frequency range of interest, the behavior is well described by a capacitance of typically 1 nF. The signal transfer over a 4-km paper cable, terminated by various load impedances to mimic real equipment is studied. From the results it is concluded that for inductive coupling performance within the CENELEC A-band may be sufficient, except for substations at the end of a grid. Transferring signals containing frequencies up to several megahertz, which is already required for synchronization of partial discharge detection and location equipment, is feasible under all conditions. Measurements on life substations indicate that up to these frequencies substation components can still be accurately modeled as lumped circuit impedances.

Sensors for on-line PD detection in MV power cables and their locations in substations

For developing a system capable of on-line detecting and localizing PDs in MV power cables, the choice of the sensor type is crucial. Sensors for detecting PDs can be divided into two main groups: capacitive and inductive sensors. In this paper, the advantages and disadvantages of these two types of sensors are discussed for on-line application. In multi-conductor cables, e.g. a three-phase belted cable, it is essential to distinguish the different propagation channels. When the cable is under normal operating conditions, all three the phases are energized simultaneously, implying that the PD pulses are propagating through two distinctive propagation channels: the Phase-to-Phase (PP) channel and the Shield-to-Phase (SP) channel. Measuring sensors can therefore also be subdivided with respect to the detected channel. If sensors can detect signals in all three conductors separately, both SP and PP channels are obtained. In this paper different positions in a substation for placing sensors are evaluated, with respect to the measured propagation channels, signal-and interference sensitivity, safety and practical applicability.

Synchronization of on-line PD detection and localization setups using pulse injection

The development of an on-line PD detection system has its main challenges in the area of sensors and signal processing techniques. If localization is included by means of simultaneous signal detection at both cable ends, the problem arises of synchronizing the measuring systems at both sites accurately. One method, already successfully applied in off-line PD measurements, is the use of GPS. Besides its high costs, this system has one main disadvantage: it requires an external antenna in the line of sight of several satellites. This requirement may be a problem for on-line measurements because of the long measuring time. Another option to perform the synchronization is the injection of high-frequency pulses into one cable end and detecting them at the other end. Taking the propagation time along the cable length into account, this results in accurate synchronization of the measuring sides, without being dependent on any other system. This paper discusses several aspects of the pulse injection technique. The obtained resolution realized during field experiments with the method of pulse injection is within the resolution obtainable by means of GPS (100 ns).

Time-base alignment of PD signals measured at multiple cable ends

PD measurement is one of the proven methods to determine the insulation quality of e.g., MV power cables. Location capability gives the results much extra value. If measurements are done on long power cables (>4 km), branched cable systems or if they are performed on-line, multiple measuring setups are required to enable PD location. Recorded signals should be synchronized accurately to allow extraction of relevant location information. In this paper several methods are discussed. It is concluded that the most promising method is the injection of pulses through the cable and aligning recorded data records to these pulses. Inaccuracy introduced by time-wandering of sampling clocks can also be compensated by this technique. Measurements show the applicability of the method up to at least 8 km, producing location accuracies of approximately 0.2%.

Propagation characteristics of three-phase belted paper cable for on-line PD detection

Partial discharge detection and interpretation benefits from accurate knowledge of the channel through which a discharge pulse propagates. In this paper the propagation of signals in a three-phase belted paper cable is studied for the purpose of online partial discharge detection. In this case, a general multi-conductor transmission line model characterizes signal propagation. However, if a specific method of measurement is applied, a simplified two-conductor transmission line model is sufficient.. Experiments support this observation. The simplified model is used to model the impulse response of the channel and therefore the measured waveform can be estimated. Model parameters are obtained through experiments and the model is verified by experiments.

Cancellation of continuous periodic interference for PD detection

Practical partial discharge detection is generally impeded by noise and interference. Especially in case of on-line detection in the field, the signal-to-noise ratio of the partial discharge pulses can be low. Primary disturbance sources include continuous periodic interference, such as amplitude modulated radio broadcasts. Typically, this type of interference is eliminated by digital notch filtering. Filters may introduce PD pulse shape distortion, which is generally not desired. As an alternative, adaptive interference cancellation is presented. Moreover, a mixed-signal implementation is proposed that enables interference rejection prior to digitization.

Determination of substation model for correct interpretation of on-line measured PD signals from MV cable systems

PD measurement is a proven diagnostic for MV cable system insulation. The PD charge is a relevant parameter for interpretation of test results. In case of off-line PD diagnostics, the measuring system is relatively easy to calibrate. For on-line measurements, the magnitude of signals from the cable under test is also dependent on the impedances in the connected substation. In order to obtain correct PD magnitudes for on-line measured signals, this impedance must be known. In this paper, a substation model is presented, as well a measuring method to obtain model parameter values. Experiments show good agreement with the model. The measuring method provides a solution to determine the substation impedance(s) on-line, so without any switching, disconnecting or making galvanic contact. This can be an important benefit to on-line PD measuring systems.

PD location in power cables using parametric models

In order to locate partial discharge in power cables, multiple pulses are detected using sensor(s) either installed at a single cable end or at both ends. The difference in arrival times of pulses originating from the same defect provides a measure for the discharge location. Multiple sources contribute to errors in the determined location, such as attenuation and dispersion introduced by the cable. Dispersion implies that the propagation velocity or cable speed is frequency dependent. As a consequence errors are introduced in the mapping of discharge sources along the cable length. Moreover, pulses spread and waveform reference points cannot be well defined, thus impeding accurate time of arrival estimation. Model-based signal processing provides a theoretical framework to deal with such inaccuracies. An efficient model-based method is proposed that enables compensation for dispersion related errors.