Scholarlink

<p>This paper presents an exploration analysis of the performance of a TD3 (Twin Delayed Deep Deterministic Policy Gradient) based Deep Reinforcement Learning (DRL) controller, comparing the traditionally available control methods such as PI control and Model Predictive Control (MPC) for the output voltage regulation of a Dual Active Bridge (DAB) Converter under dynamic operating conditions. The circuit operation of the DAB converter is studied for single phase shift modulation, and the generalized average model of the converter is analysed for optimized control performance. Preceding with an elaborate discussion on Grey Wolf Optimized-PI control and the MPC control for voltage control of the DAB converter. The rising application of DRL controllers for power electronic converters urges for the performance analysis of TD3 agent-based DRL control for voltage regulation of a DAB converter using single phase shift modulation. MATLAB 2024b provides a Reinforcement Learning Designer App to create, train, and simulate a DRL agent for a predefined or customized environment available in the MATLAB directory files. Thus, paving the way for a simpler and faster control option available for online real-time implementation. The simulation results present the output voltage regulation of the DAB converter during dynamic conditions for individual control schemes implemented in MATLAB. The Hardware-in-Loop simulation results using RTDS technology validates the superior performance of TD3-based DRL control over the other traditional control schemes.</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

Exploration Analysis of Deep Reinforcement Learning Controller in Contrast to Traditional Control Methods for Dual Active Bridge Converter for Enactment in Microgrids

  • 期刊论文

  • 发表日期：

    2026年1月6日

查看全文
  • Stutee Patra,
  • Aditi Chatterjee,
  • Kanungo Barada Mohanty

摘要

This paper presents an exploration analysis of the performance of a TD3 (Twin Delayed Deep Deterministic Policy Gradient) based Deep Reinforcement Learning (DRL) controller, comparing the traditionally available control methods such as PI control and Model Predictive Control (MPC) for the output voltage regulation of a Dual Active Bridge (DAB) Converter under dynamic operating conditions. The circuit operation of the DAB converter is studied for single phase shift modulation, and the generalized average model of the converter is analysed for optimized control performance. Preceding with an elaborate discussion on Grey Wolf Optimized-PI control and the MPC control for voltage control of the DAB converter. The rising application of DRL controllers for power electronic converters urges for the performance analysis of TD3 agent-based DRL control for voltage regulation of a DAB converter using single phase shift modulation. MATLAB 2024b provides a Reinforcement Learning Designer App to create, train, and simulate a DRL agent for a predefined or customized environment available in the MATLAB directory files. Thus, paving the way for a simpler and faster control option available for online real-time implementation. The simulation results present the output voltage regulation of the DAB converter during dynamic conditions for individual control schemes implemented in MATLAB. The Hardware-in-Loop simulation results using RTDS technology validates the superior performance of TD3-based DRL control over the other traditional control schemes.