Partial discharge signal propagation, modelling and estimation in high voltage transformer windings

The thesis concerns partial discharge (PD) propagation within a high voltage transformer winding. Location of PDs and magnitude estimation are important tools for both diagnosis and prognosis of the health of large transformers. In reality there is limited access and knowledge of a winding and consequently any practical method requires the use of estimation techniques. The approach taken in this thesis is by considering lumped circuit parameter models. Firstly, a lumped capacitive parameter model was considered and secondly a transmission line lumped parameter approach developed. A technique of split winding analysis is introduced for both types of model. The derivation of the capacitive network considers the source location of a PD by defining the PD signal propagation in two directions. At the source, the currents are equal in magnitude and are attenuated as they flow in each direction. This provides information for a fixed distribution model equation. Under transmission line lumped parameter models, split winding analysis explains the development of accumulated harmonic waveforms of the PD propagation signal towards the neutral and bushing tap-point. At the source, a D’Alembert solution is employed to estimate the oscillation level and found to be in very good agreement with measured data using rectangular wave signal injection. PD signal behaviour is then considered using a time varying boundary conditions model with a principle of superposition equation of source signal. Duhamel’s principle is employed to find a solution for any waveform applied to some point on the transformer winding. Under the influence of losses and distortion, an accumulated harmonic amplitude analysis from the Duhamel’s principle estimates the PD propagation level. For different injection points along the transformer winding, the measured PD level at the neutral and bushing tap point caused by the accumulated harmonic amplitude reveals different patterns. As the PD injection point is altered from the bushing tap point to the neutral, the measured signals significantly change. This in turn contains information of the level of discharge signal at the source. From this analysis a technique based on minimum mean error (MME) calculation using the measurements at the bushing tap and neutral points can be used to identify the source location of PDs based on the analysis of accumulated harmonic amplitudes. With a known location, the information can then be used to estimate PD levels. As the actual charge transferred at the location of a partial discharge cannot be measured directly, by using the D’Alembert solution, the PD source level is found to have approximately twice the apparent magnitude. By using the predominantly capacitive model derived based on split current propagation, PD estimation at higher frequencies is also possible. As a result, an estimation of PD level can be estimated for measurement signals having bandwidth of up to 150MHz.