Research on deconvolution methods for thermal network of power devices

The structure function method is critical for obtaining Cauer thermal network models for power devices. However, in its deconvolution step, different calculation methods have a large impact on the results, which affects the accuracy of the thermal network model. The underlying mechanism of each calc...

Full description

Saved in:
Bibliographic Details
Main Authors: DENG Erping, YANG Ying, WANG Yanhao, CHANG Guiqin, HUANG Yongzhang, DING Lijian
Format: Article
Language:zho
Published: Editorial Department of Electric Drive for Locomotives 2023-09-01
Series:机车电传动
Subjects:
Online Access:http://edl.csrzic.com/thesisDetails#10.13890/j.issn.1000-128X.2023.05.013
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The structure function method is critical for obtaining Cauer thermal network models for power devices. However, in its deconvolution step, different calculation methods have a large impact on the results, which affects the accuracy of the thermal network model. The underlying mechanism of each calculation method, the amplification effect of different calculation methods on noise, and the reasonable selection of calculation methods are very important and need to be solved urgently at present. In this paper, by studying the core aspect of deconvolution in the structure function method, the theoretical analysis of three deconvolution calculation methods was carried out from three aspects: calculation method input, calculation method core, and error analysis; data analysis was carried out from six aspects: number of iterations, calculation time, sampling frequency, accuracy of calculation methods, selection of calculation methods and practical application. The suitable data characteristics were also summarized for each of the three methods according to their calculation characteristics, which provide reference and guidance for selecting appropriate fast calculation methods of thermal network models for different types of input data.
ISSN:1000-128X