Lightweight Res-CNN-Based Online ESR Estimation Method for Boost Converters

Citations

WEB OF SCIENCE

0
Citations

SCOPUS

0

초록

This article proposes a lightweight residual learning-based convolutional neural network (Res-CNN) for online real-time estimation of the equivalent series resistance (ESR) of electrolytic capacitors (ECs) in DC/DC boost converters, implemented on an onboard microcontroller unit (MCU) platform. The developed lightweight Res-CNN, which estimates ESR values based on the regression network, shows strong generalization capability under variations in operating and load conditions, together with lightweight model design. A ghost module-based design presented in this article reduces model complexity, including parameter size, model size, and floating-point operations (FLOPs) while preserving diagnostic accuracy. The main contributions can be summarized as follows: 1) regression-based ESR estimation algorithm with strong generalization performance even under untrained device and operating/load conditions; 2) lightweight model architecture designed with a ghost module, which reduces the model complexity; and 3) deployment of the proposed lightweight Res-CNN on an MCU-based online platform. In this article, online real-time ESR estimation is experimentally performed during boost converter operation, confirming consistency between the offline and online evaluation results. The effectiveness of the proposed method is verified with comparative results of previous methods, from the standpoints of estimation accuracy and model complexity.

키워드

Modeling; Accuracy; Training; Capacitors; Estimation; Convolutional neural networks; Loading; Printing; Converters; Testing; Boost converter; convolutional neural network; CNN; electrolytic capacitor; EC; equivalent series resistance estimation; ESR estimation; lightweight model; OUTPUT CAPACITOR
제목
Lightweight Res-CNN-Based Online ESR Estimation Method for Boost Converters
저자
Lee, Eungyu; Kim, Jihwan; Kwak, Sangshin
DOI
10.1109/TIM.2026.3738115
발행일
2026
유형
Article
저널명
IEEE Transactions on Instrumentation and Measurement
권
75