Scholarlink

<p>Power quality problems with dynamic load fluctuations, harmonic distortion and voltage sags occur when renewable sources like solar PV systems are integrated into existing distribution systems. Series and shunt power disturbances are reduced using UPQCs. Under rapidly changing solar and grid conditions, PQ-based PI controllers will fail. These issues are addressed by a Deep Actor–Critic reinforcement learning (RL)-based multi-objective optimal control method for a Solar PV-integrated UPQC. Unlike PI, fuzzy, and model-based controllers, the recommended method learns optimal control policies from system interactions without mathematical modelling. The Actor network continually regulates UPQC’s series, shunt converters, while Critic network uses a multi-objective reward function to reduce THD, voltage sag/swell, DC-link voltage and power factor. This control methodology applies to a nonlinear, dynamic grid with demand uncertainty and intermittent solar power. MATLAB/Simulink simulations show that the proposed controller outperforms PI and adaptive controllers in dynamic response, harmonic suppression, voltage regulation, and DC-link stability. A reliable, adaptable, and scalable Deep Actor–Critic RL approach for intelligent power quality enhancement in smart distribution systems with substantial renewable penetration was found. The proposed DA-CRL controller, source Voltage THD values are 0.42%, source Current THD values are 2.12%.</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 Actor–Critic Reinforcement Learning (DA-CRL) based multi-objective optimal control of solar PV integrated UPQC for power quality enhancement in smart distribution networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月21日

查看全文
  • E. Hima Bindu,
  • P. B. Guru Prasanna,
  • D. Krishna,
  • B. Srikanth Goud,
  • Praveen Kumar Balachandran

摘要

Power quality problems with dynamic load fluctuations, harmonic distortion and voltage sags occur when renewable sources like solar PV systems are integrated into existing distribution systems. Series and shunt power disturbances are reduced using UPQCs. Under rapidly changing solar and grid conditions, PQ-based PI controllers will fail. These issues are addressed by a Deep Actor–Critic reinforcement learning (RL)-based multi-objective optimal control method for a Solar PV-integrated UPQC. Unlike PI, fuzzy, and model-based controllers, the recommended method learns optimal control policies from system interactions without mathematical modelling. The Actor network continually regulates UPQC’s series, shunt converters, while Critic network uses a multi-objective reward function to reduce THD, voltage sag/swell, DC-link voltage and power factor. This control methodology applies to a nonlinear, dynamic grid with demand uncertainty and intermittent solar power. MATLAB/Simulink simulations show that the proposed controller outperforms PI and adaptive controllers in dynamic response, harmonic suppression, voltage regulation, and DC-link stability. A reliable, adaptable, and scalable Deep Actor–Critic RL approach for intelligent power quality enhancement in smart distribution systems with substantial renewable penetration was found. The proposed DA-CRL controller, source Voltage THD values are 0.42%, source Current THD values are 2.12%.