Prediction and measurement of high-frequency power losses in magnetic components under power electronics excitation

Abstract

The purpose of this work is to propose accurate approaches for measuring and predicting the core and winding losses of inductors and high-frequency transformers in power electronics applications. The large-signal procedures and practical efforts for precisely characterizing the magnetic component losses with the rectangular wave excitation are elaborated.
Among the new ways presented are offline and in-situ methods for reducing phase discrepancy error when measuring core and winding losses of magnetic components.
In the proposed offline approach, the phase discrepancy is compensated by shifting the measured current horizontally to align with the reference phase angle found from an offline impedance frequency sweep using an impedance analyzer in the post-processing of measured waveforms in the two-winding method.
Novel experimental methods to precisely measure the in-situ inductor and high-frequency transformer winding losses are presented, using the reactive voltage cancellation concept to separate the winding loss from the core loss and reduce the phase discrepancy error. The proposed in-situ measurement can capture the complete winding loss, including impacts from non-ideal field distributions and interactions between the core and windings with the presence of the magnetic core and the load on the secondary side of the transformer.
This study discovers that different load conditions can affect high-frequency transformers' core and winding losses. The physical causes of these loss variations are discussed and investigated by experimental measurements and 3-D finite element analysis results.
The magnetic component loss maps have also been improved to incorporate reliable data besides core loss, such as winding loss and total loss. These new loss maps aim to pave the way for standardizing magnetic components and provide a practical and accurate power loss estimation method embedded in the datasheets of the standardized magnetic components. Machine learning techniques are also utilized to develop applications based on magnetics' component loss maps. The proposed neural network aided loss maps not only show superior accuracy throughout the whole dataset, but also enable the loss maps to expand the dimensions to account for more factors (e.g., load conditions in transformers) and generate multiple outputs (e.g., both the winding loss and core loss).
The ultimate objective is to improve loss measurement approaches to enable a quick and accurate estimation, which is beneficial for both commercial magnetic component manufacturers and end-users.

Date of Award	3 Oct 2023
Original language	English
Awarding Institution	University of Bristol
Supervisor	Xibo Yuan (Supervisor) & Jun Wang (Supervisor)