基于人工神经网络的差模EMI滤波器插入损耗预测  

Insertion Loss Prediction of Differential-mode EMI Filter Based on Artificial Neural Networks

作　　者：陈荣亮 梁海燕 刘艺涛 CHEN Rongliang;LIANG Haiyan;LIU Yitao(College of Mechatronics and Control Engineering,Shenzhen University,Shenzhen 518060,China)

机构地区：[1]深圳大学机电与控制工程学院,深圳518060

出　　处：《电源学报》2024年第5期67-73,共7页Journal of Power Supply

基　　金：深圳市科技计划资助项目(JCYJ20230808104910021)。

摘　　要：在电力电子设备中,高速开关经常会导致严重的电磁干扰EMI(electromagnetic interfere),严重影响电力电子系统的可靠性。为解决EMI问题,EMI滤波器是1种常用的解决方案。插入损耗作为噪声衰减能力的评价指标,其模型的准确性直接影响EMI滤波器的参数设计精度。为了提高EMI滤波器插入损耗模型的预测精度,首先准确描述系统行为并预测EMI滤波器滤波性能,提高EMI滤波器设计效率,然后利用反向传播神经网络对单级差模EMI滤波器的插入损耗进行建模。相较于理想模型和高频电路的行为模型,所提神经网络模型预测精度具有更好的实际应用价值,可以快速评估EMI滤波器的实际插入损耗,提高EMI滤波器设计效率,并为EMI滤波器的设计和优化提供指导。In power electronic devices,high-speed switching will often lead to serious electromagnetic interference(EMI)problems,which seriously affects the reliability of power electronic systems.To solve these EMI problems,EMI filters are a common solution.The insertion loss is an evaluation index for the noise attenuation capability,and the accuracy of its model directly affects the parameter design accuracy of EMI filters.To improve the prediction accuracy of the EMI filter insertion loss model,accurately describe the system behavior and predict the filtering performance of the EMI filter,and improve the design efficiency of the EMI filter,the insertion loss of a single-stage differential-mode EMI filter is modeled using a back propagation neural network.The proposed neural network model has better practical application value than the ideal model and the behavioral model of a high-frequency circuit,aiming to provide guidance for the design and optimization of EMI filters.This model can quickly evaluate the actual insertion loss of EMI filters to improve their design efficiency.

关 键 词：寄生参数 差模EMI滤波器 插入损耗 人工神经网络 

分 类 号：TM464[电气工程—电器]