Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer

Yifan Chen; Qingxin Yang; Yongjian Li; Hao Zhang; Changgeng Zhang

doi:10.1109/ASEMD59061.2023.10369294

Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer

Yifan Chen
, Qingxin Yang
, Yongjian Li
, Hao Zhang
, Changgeng Zhang^*

^*Corresponding author for this work

Hebei University of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

This paper aims to estimate the electromagnetic field distribution in a simplified transformer through two-dimensional (2-D) finite element analysis. We create two datasets, namely the original dataset (OD) to simulate grayscale images and the new dataset (ND) which incorporates distinct physical properties of materials. These datasets are helpful to investigate the impact of different data types on deep learning. In order to enhance the accuracy of estimation, we compare the performances of two convolutional neural network (CNN) architectures: U-net and its improved version, U-Resnet (which incorporates residual blocks from ResNet). Additionally, we introduce a specialized loss function, Add-RMSE, which is better suited for dense regression problems, thus improving the prediction accuracy. The effectiveness of our proposed method is validated through test experiments, where we analyze the estimation results obtained.

Original language	English
Title of host publication	2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9798350301571
DOIs	https://doi.org/10.1109/ASEMD59061.2023.10369294
State	Published - 2023
Externally published	Yes
Event	2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023 - Tianjin, China Duration: 27 Oct 2023 → 29 Oct 2023

Publication series

Name	2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023

Conference

Conference	2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023
Country/Territory	China
City	Tianjin
Period	27/10/23 → 29/10/23

Keywords

CNN
Electromagnetic field distribution
FEM
ResNet
U-net

Access to Document

10.1109/ASEMD59061.2023.10369294

Cite this

Chen, Y., Yang, Q., Li, Y., Zhang, H., & Zhang, C. (2023). Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer. In 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023 (2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ASEMD59061.2023.10369294

Chen, Yifan ; Yang, Qingxin ; Li, Yongjian et al. / Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer. 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023. Institute of Electrical and Electronics Engineers Inc., 2023. (2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023).

@inproceedings{96aad4788ab44f70b87a9502c35f16df,

title = "Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer",

abstract = "This paper aims to estimate the electromagnetic field distribution in a simplified transformer through two-dimensional (2-D) finite element analysis. We create two datasets, namely the original dataset (OD) to simulate grayscale images and the new dataset (ND) which incorporates distinct physical properties of materials. These datasets are helpful to investigate the impact of different data types on deep learning. In order to enhance the accuracy of estimation, we compare the performances of two convolutional neural network (CNN) architectures: U-net and its improved version, U-Resnet (which incorporates residual blocks from ResNet). Additionally, we introduce a specialized loss function, Add-RMSE, which is better suited for dense regression problems, thus improving the prediction accuracy. The effectiveness of our proposed method is validated through test experiments, where we analyze the estimation results obtained.",

keywords = "CNN, Electromagnetic field distribution, FEM, ResNet, U-net",

author = "Yifan Chen and Qingxin Yang and Yongjian Li and Hao Zhang and Changgeng Zhang",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023 ; Conference date: 27-10-2023 Through 29-10-2023",

year = "2023",

doi = "10.1109/ASEMD59061.2023.10369294",

language = "英语",

series = "2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023",

address = "美国",

}

Chen, Y, Yang, Q, Li, Y, Zhang, H & Zhang, C 2023, Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer. in 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023. 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023, Institute of Electrical and Electronics Engineers Inc., 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023, Tianjin, China, 27/10/23. https://doi.org/10.1109/ASEMD59061.2023.10369294

Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer. / Chen, Yifan; Yang, Qingxin; Li, Yongjian et al.
2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023. Institute of Electrical and Electronics Engineers Inc., 2023. (2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer

AU - Chen, Yifan

AU - Yang, Qingxin

AU - Li, Yongjian

AU - Zhang, Hao

AU - Zhang, Changgeng

PY - 2023

Y1 - 2023

N2 - This paper aims to estimate the electromagnetic field distribution in a simplified transformer through two-dimensional (2-D) finite element analysis. We create two datasets, namely the original dataset (OD) to simulate grayscale images and the new dataset (ND) which incorporates distinct physical properties of materials. These datasets are helpful to investigate the impact of different data types on deep learning. In order to enhance the accuracy of estimation, we compare the performances of two convolutional neural network (CNN) architectures: U-net and its improved version, U-Resnet (which incorporates residual blocks from ResNet). Additionally, we introduce a specialized loss function, Add-RMSE, which is better suited for dense regression problems, thus improving the prediction accuracy. The effectiveness of our proposed method is validated through test experiments, where we analyze the estimation results obtained.

AB - This paper aims to estimate the electromagnetic field distribution in a simplified transformer through two-dimensional (2-D) finite element analysis. We create two datasets, namely the original dataset (OD) to simulate grayscale images and the new dataset (ND) which incorporates distinct physical properties of materials. These datasets are helpful to investigate the impact of different data types on deep learning. In order to enhance the accuracy of estimation, we compare the performances of two convolutional neural network (CNN) architectures: U-net and its improved version, U-Resnet (which incorporates residual blocks from ResNet). Additionally, we introduce a specialized loss function, Add-RMSE, which is better suited for dense regression problems, thus improving the prediction accuracy. The effectiveness of our proposed method is validated through test experiments, where we analyze the estimation results obtained.

KW - CNN

KW - Electromagnetic field distribution

KW - FEM

KW - ResNet

KW - U-net

UR - https://www.scopus.com/pages/publications/85183599072

U2 - 10.1109/ASEMD59061.2023.10369294

DO - 10.1109/ASEMD59061.2023.10369294

M3 - 会议稿件

AN - SCOPUS:85183599072

T3 - 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023

BT - 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023

Y2 - 27 October 2023 through 29 October 2023

ER -

Chen Y, Yang Q, Li Y, Zhang H, Zhang C. Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer. In 2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023. Institute of Electrical and Electronics Engineers Inc. 2023. (2023 IEEE International Conference on Applied Superconductivity and Electromagnetic Devices, ASEMD 2023). doi: 10.1109/ASEMD59061.2023.10369294

Data-Driven Deep Convolutional Neural Networks for Electromagnetic Field Estimation of Dry-type Transformer

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this