Multi-objective optimization and meta-modeling of tape-wound transformers

Ahmed Taher, Purdue University

Abstract

In the research presented herein, design models of tape-wound transformers to support component and system-level optimization are considered. As a basis for component optimization, a magnetic equivalent circuit (MEC) model is derived. The key components of the MEC model are the leakage permeances, which have been established using analytical techniques and validated using both 2D and 3D finite element analysis. To enable high frequency design, expressions that predict the winding AC resistance and the proximity effect loss are derived. In addition, a thermal equivalent circuit (TEC) model is established to predict the temperature throughout the transformer and to account for temperature impact on winding resistances. To predict transformer performance, the T-equivalent, MEC, and thermal models are coupled to determine the magnetic operating point and establish core loss, winding loss, voltage regulation, and inrush current given the core and winding geometries, material properties, input voltage, and rated load. The coupled MEC/T-equivalent and the TEC circuit-based performance evaluation is demonstrated within an optimization in which the goals are to minimize mass and minimize Loss. To support system-level optimization, a scaling technique is derived in which transformer size/mass is predicted based upon rated power, specified current density, and frequency. Curve-fitting techniques are used to derive a meta-model for scaled mass and power loss. The meta-model is compared to designs obtained using detailed design code. A strong agreement between the results from the detailed design code and that predicted by the meta-model is achieved.

Degree

Ph.D.

Advisors

Pekarek, Purdue University.

Subject Area

Electrical engineering

COinS