Accurate Leakage Inductance Modeling Using an Artificial Neural Network Based on the Dowell Model

カテゴリ: 論文誌(論文単位)

グループ名: 【D】産業応用部門（英文）

発行日: 2023/05/01

タイトル(英語): Accurate Leakage Inductance Modeling Using an Artificial Neural Network Based on the Dowell Model

著者名: Yu-Hsin Wu (Electrical Engineering, Nagoya University), Koichi Shigematsu (Institute of Materials and Systems for Sustainability, Nagoya University), Yasumichi Omoto (Power Electronics Division, NIDEC MOBILITY CORPORATION), Yoshihiro Ikushima (Power Elect

著者名(英語): Yu-Hsin Wu (Electrical Engineering, Nagoya University), Koichi Shigematsu (Institute of Materials and Systems for Sustainability, Nagoya University), Yasumichi Omoto (Power Electronics Division, NIDEC MOBILITY CORPORATION), Yoshihiro Ikushima (Power Electronics Division, NIDEC MOBILITY CORPORATION), Jun Imaoka (Institute of Materials and Systems for Sustainability, Nagoya University), Masayoshi Yamamoto (Institute of Materials and Systems for Sustainability, Nagoya University)

キーワード: transformer，leakage inductance，dowell，modeling，machine learning，neural network

要約(英語): This study proposes strategies for improving the accuracy and usefulness of leakage inductance modeling using the Dowell model (DM). As the analytical method of modeling the leakage inductance model considers geometrical factors, it is vital for front-loading transformer design. DM is one such widely used one-dimensional magnetic field-based approach for analytically modeling both AC resistance and leakage inductance. It is more accurate and requires less computational work than other analytical approaches proposed in previous studies. However, some approximations may cause errors and eventually lead to inaccurate results. Therefore, this study aims to ascertain the conditions that result in inaccurate leakage inductance modeling. Additionally, a simple exponential-based model and an artificial neural network are developed to increase the accuracy of inaccurate modeling. The results clarify the conditions that result in a frequency-dependent and bias error. Moreover, the intended findings indicate that the proposed strategies effectively improve modeling accuracy while simultaneously providing some extra advantages for transformer design.

本誌: IEEJ Journal of Industry Applications Vol.12 No.3 （2023） Special Issue on “IPEC-Himeji 2022”

本誌掲載ページ: 334-344 p

原稿種別: 論文／英語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejjia/12/3/12_22007452/_article/-char/ja/