The University of Canterbury High Voltage Laboratory frequently uses partial core resonant transformers to conduct high voltage testing within the New Zealand power industry. These transformers utilise a non-standard core design in which flux is not constrained around the entirety of its path. This results in characteristics which are beneficial for high voltage testing, but also have negative consequences. The most significant of these consequences is a noticeable increase in losses, resulting in higher primary currents and a reduction in resonant effects. Present models are unable to estimate this loss in any accurate sense, instead relying on the experience of designers.

In order to better understand these losses, this thesis describes a study into the quantification of core losses within partial core transformers.

Previous studies are reviewed relating to the design and modelling of partial core transformers, and learnings and criticisms of these studies are offered. Theories relating to magnetic material loss derivation are collected, these theories relate to the traditional components of hysteresis and eddy current loss, and also to the more contemporary theories regarding excess loss.

A method of specification, design and construction for the cores of partial core transformers is developed. This method is used to create an inventory of partial core transformer cores, as well as providing numerical representation of the dimensions of these cores.

Testing is completed and described alongside relevant theory in order to determine what modes of core loss are present within partial core transformers, and to what amount these separate modes contribute to core loss. From this work it is determined that hysteresis is the dominant contributor to core losses, with lesser amounts of eddy current loss contributions. Excess losses are determined to contribute no meaningful losses.

Finally, an FEM model is developed taking into account the anisotropic and nonlinear nature of the core. This model is used to investigate flux distribution within the core, and this distribution is then combined with relevant theory to produce a model able to accurately estimate losses within the core. This model is shown to agree with the results of testing, provide a more accurate estimate than present models, and also to correctly determine where heating effects will occur within the core.