电工技术学报2026，Vol.41Issue(7)：2468-2483,16.DOI:10.19595/j.cnki.1000-6753.tces.250568
基于贝叶斯优化与CRITIC方法的中压高频变压器绝缘结构优化设计
Optimization Design of Insulation Structure of Medium Voltage and High Frequency Transformer Based on Bayesian Optimization and CRITIC Method
摘要
Abstract
High-frequency transformer(HFT)plays an important role in power conversion devices such as power electronic converters,facilitating voltage conversion and isolation.The corresponding optimized design is a crucial aspect in achieving high power density,high efficiency,and high reliability.HFT is characterized by compact size,high power density,and constrained insulation design margins,faces significant insulation challenges.This paper proposes an insulation structure and corresponding insulation design methodologies for medium-voltage,high-frequency transformer used in isolated DC-DC converters. A sandwich-type transformer structure is designed that employs an insulation package incorporating an internal air gap.By inserting an air gap between the insulating bobbin and the epoxy potting resin,the peak electric-field strength is reduced while heat dissipation is improved.This structure transfers the region of maximum field into the bulk insulating material,thereby protecting the windings and magnetic cores and mitigating the effects of partial discharge.And it was verified using finite element simulation. A detailed analysis is presented of the fundamentals of the method of the mirror charge for solving two-dimensional electrostatic field problems.The maximum electric field strength in the high-frequency transformer's 2-D domain is extracted using both the finite-element method and the mirror charge method,the discrepancy between the two is within 3%,while the mirror charge delivers roughly a tenfold speedup.Based on this,the mirror charge is employed to compute and constrain the insulation clearances of transformer components embedded in a single dielectric medium. A multi-objective optimization model is constructed using the CRITIC method,which determines the information entropy weight coefficients by calculating the comparative intensity and conflict analysis of the maximum electric field strength and insulator volume dual indicators.Based on this,the Bayesian optimization method is introduced to achieve the collaborative optimization of reducing transformer electric field strength and volume reduction within the constraints of evaluation times by constructing a Gaussian process surrogate model and an improved expected improvement acquisition function.Simulation results show that,compared to conventional optimization schemes,the proposed design reduces the maximum field strength by approximately 18%,and reduces the volume by about 10%compared to conservative schemes. Insulation tests were conducted to verify the insulation performance and reliability of the transformer prototype,including an impulse voltage test at 10 kV,a short-time withstand voltage test at 8 kV,and a partial discharge test at 7.2 kV.All the results met the relevant requirements.关键词
绝缘结构/高频变压器/镜像电荷法/贝叶斯优化/最大电场强度Key words
Insulation structure/high frequency transformer/mirror charge method/Bayesian optimization/maximum electric field strength分类
信息技术与安全科学引用本文复制引用
赵志刚,刘永军,陈天缘,张时,张子洁..基于贝叶斯优化与CRITIC方法的中压高频变压器绝缘结构优化设计[J].电工技术学报,2026,41(7):2468-2483,16.基金项目
国家自然科学基金项目(52377008)、河北省燕赵黄金台聚才计划骨干人才项目(HJZD202501)和河北省中央引导地方科技专项(254Z2101G)资助. （52377008）
