Scholarlink

<p>Model simplifications lead to low precision and labour intensive operations in traditional LLC converter design approach, when external variations cause nonlinearity, controlling a system is challenging. Hence this paper presents a modified deep neural network control of LLC resonant converter to automate control and increase accuracy, utilizing the learning and training capability of neural networks. Comparisons are drawn between system response using traditional PIDD2 controller and NARMA L2 deep neural network controller. The design and control procedure of this modified NARMA L2 deep neural network controller is explained along with EDF modeling in LLC resonant converter. To confirm the effectivity of proposed method, MATLAB/SIMULINK is employed to analyse the deep neural network and PIDD2 controller considering integral square error, settling time, overshoot and rise time. The proposed NARMA L2 controller technique provides 86% improvement in rise time, 63% in settling time, 45% in overshoot and 55% in integral square error for line variations then 99% improvement in rise time, 57% in settling time, 88% in overshoot and 58% in integral square error for load variations compared to PIDD2 controller.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep neural network control for LLC resonant converter in electric vehicles

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月3日

查看全文
  • K. Sathya,
  • K P Guruswamy

摘要

Model simplifications lead to low precision and labour intensive operations in traditional LLC converter design approach, when external variations cause nonlinearity, controlling a system is challenging. Hence this paper presents a modified deep neural network control of LLC resonant converter to automate control and increase accuracy, utilizing the learning and training capability of neural networks. Comparisons are drawn between system response using traditional PIDD2 controller and NARMA L2 deep neural network controller. The design and control procedure of this modified NARMA L2 deep neural network controller is explained along with EDF modeling in LLC resonant converter. To confirm the effectivity of proposed method, MATLAB/SIMULINK is employed to analyse the deep neural network and PIDD2 controller considering integral square error, settling time, overshoot and rise time. The proposed NARMA L2 controller technique provides 86% improvement in rise time, 63% in settling time, 45% in overshoot and 55% in integral square error for line variations then 99% improvement in rise time, 57% in settling time, 88% in overshoot and 58% in integral square error for load variations compared to PIDD2 controller.