Efficient Design Method for a Forward-converter transformer based on a KNN–GRU–DNN Model
- Authors
- Lee, Gang Seok; Kim, Sanha; Bae, Sung Woo
- Issue Date
- Jan-2023
- Publisher
- IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
- Keywords
- Artificial intelligence; deep neural network (DNN); forward-converter transformer (FCT); gate-recurrent unit (GRU); K-nearest neighbors (KNN)
- Citation
- IEEE TRANSACTIONS ON POWER ELECTRONICS, v.38, no.1, pp.73 - 78
- Indexed
- SCIE
SCOPUS
- Journal Title
- IEEE TRANSACTIONS ON POWER ELECTRONICS
- Volume
- 38
- Number
- 1
- Start Page
- 73
- End Page
- 78
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/185356
- DOI
- 10.1109/TPEL.2022.3203480
- ISSN
- 0885-8993
- Abstract
- This letter proposes an efficient design method for a forward-converter transformer (FCT) with artificial intelligence (AI). Conventional FCT design is inefficient because it requires numerous repeated design processes. To solve this problem, this letter proposes FCT design by applying a KNN–GRU–DNN model. The design estimation accuracy of the proposed AI model was over 91% based on Google colaboratory validation. The proposed transformer design also satisfied the design requirements with less than 1,450 epochs. Once the learning process is completed, the proposed AI-based transformer design can obtain various FCT designs without further repeated training procedures. To verify the proposed design results, this study conducted finite-element method (FEM) simulations using ANSYS Electronics Desktop 2018.2 and hardware-in-the-loop (HIL) experiments using OPAL-RT with the transformer design values resulting from the AI-based design model. According to the FEM simulations and HIL experiments, it is verified that the secondary winding induced voltage of the transformer designed by the AI-based model satisfies the design requirements.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 공과대학 > 서울 전기공학전공 > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/185356)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.