Inductor Design for Nonisolated Critical Soft Switching Converters Using Solid and Litz PCB and Wire Windings Leveraging Neural Network Model

High-frequency (300 kHz–1 MHz) and high-efficiency (
$\ge\! \! 99.7\%$
) inductor design method is proposed for the application of critical soft switching with large current ripple (
$\triangle i_L \geq 200\%$
) to improve the power density and efficiency of electrified energy conversion systems. First, for the theoretical design section, the core/coil size, number of turns, airgap, and inductance are optimized using analytical model to minimize the inductor losses. Second, for the structural design section of coil, a 3-D routing method of litz printed circuit board (PCB) winding is developed to reduce the high-frequency copper losses. The strand number, trace width, thickness, and layout of litz PCB are analyzed in detail. Different types of coil, including solid/litz types of PCB/wire windings, are compared and analyzed based on the high-frequency copper loss reduction and space utilization. For the structural design section of core, E and I cores are optimized to reduce the volume and core losses. Specifically, E-E, E-I, I-I, and E-Air types of core structures are designed and compared considering the core loss, volume reduction, and airgap limitation for the stack of turns. Two four-layer neural network models are designed to analyze the ac losses of the proposed solid/litz PCB winding. The parameters of PCB winding can be optimized to reduce the inductor losses. Ten derived prototypes are built to benchmark between the proposed design and the commercial inductors. Power losses, temperature rise, and cost are reduced by factors of 10, 2.5, and 3, respectively, with the proposed core and coil structures for the high-frequency, high-current ripple critical soft-switching applications.