Flyback Converter Transformer Design Optimization by AI

The paper proposes an AI converter design method to cope with high power density requirements based on the Bayesian Optimization (BO) algorithm, aiming to achieve high-efficiency selection of converter parameters. An active clamp flyback (ACF) converter with 240W output power is taken as an example, the application effect of the BO algorithm in actual converter design is introduced in detail. The research process covers key steps such as initialization, transformer characteristic simulation, circuit simulation, transformer core loss simulation, efficiency calculation, and BO algorithm iterative optimization. The BO algorithm takes efficiency as the objective function and performs 10 ~ 20 iterative optimizations to obtain the best efficiency value. The experimental results verified the reliability of the simulation results and further confirmed the practical application value of the proposed method.

I. Introduction

With the rapid advancement of artificial intelligence (AI), machine learning has sparked a technological revolution across various domains, including the field of power electronics [1]. The paper aims to address the high power density requirements in current converter designs, imposing stringent demands on the accurate parameter design. However, traditional design methods shown in Fig. 1(a) require a lot of time in the iteration process of design, simulation, and implementation.