Scholarlink

<p>The global increase in photovoltaic (PV) installations has resulted in their substantial integration into the electric grid, leading to initiatives to optimize their performance. This paper introduces an artificial intelligence (AI) control strategy designed to enhance the functionality of grid-connected PV (GCPV) systems. This incorporates deep reinforcement learning (DRL) algorithms, specifically the proximal policy optimization (PPO) and twin-delayed deep deterministic (TD3), to ensure that the system control’s variables align with their desired references. Furthermore, a Crayfish optimization algorithm (COA) is employed for maximum power point tracking (MPPT) to optimally control a DC boost converter. Additionally, another DRL controller manages a buck-boost converter within an energy storage system (ESS) to mitigate the PV output fluctuations. Moreover, four DRL controllers are implemented for optimal control of the outer and inner loops of the voltage source inverter (VSI). The performance of the DRL controllers was evaluated using simulations of a 1&#xa0;MW PV plant connected to a 33-kV utility grid, for different solar irradiance and temperature profiles, conducted in a MATLAB/Simulink environment. The results were compared to those obtained from an incremental conductance-based proportional-integral (InC-PI) method, highlighting that the proposed control strategy is both effective and efficient, indicating its potential for future applications.</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

Performance Enhancement of Grid-Connected PV System Using Artificial Intelligence Techniques

  • 期刊论文

  • 发表日期：

    2025年7月15日

查看全文
  • Ghazi A. Ghazi,
  • Essam A. Al-Ammar,
  • Hany M. Hasanien

摘要

The global increase in photovoltaic (PV) installations has resulted in their substantial integration into the electric grid, leading to initiatives to optimize their performance. This paper introduces an artificial intelligence (AI) control strategy designed to enhance the functionality of grid-connected PV (GCPV) systems. This incorporates deep reinforcement learning (DRL) algorithms, specifically the proximal policy optimization (PPO) and twin-delayed deep deterministic (TD3), to ensure that the system control’s variables align with their desired references. Furthermore, a Crayfish optimization algorithm (COA) is employed for maximum power point tracking (MPPT) to optimally control a DC boost converter. Additionally, another DRL controller manages a buck-boost converter within an energy storage system (ESS) to mitigate the PV output fluctuations. Moreover, four DRL controllers are implemented for optimal control of the outer and inner loops of the voltage source inverter (VSI). The performance of the DRL controllers was evaluated using simulations of a 1 MW PV plant connected to a 33-kV utility grid, for different solar irradiance and temperature profiles, conducted in a MATLAB/Simulink environment. The results were compared to those obtained from an incremental conductance-based proportional-integral (InC-PI) method, highlighting that the proposed control strategy is both effective and efficient, indicating its potential for future applications.