Scholarlink

<p>The intermittent nature of renewable generation, stochastic EV charging (EVC) behaviour, and the dynamic behaviour of battery energy storage systems (BESS) pose challenges for energy management (EM) in photovoltaic (PV)-integrated electric vehicle (EV) parking lots. The paper introduces a new hybrid EM model that combines the Serval Optimization Algorithm (SOA) and the Multimodal Adaptive Spatio-Temporal Graph Neural Network (MASTGNN) to enable a reconfigurable microgrid (MG) to achieve adaptive, efficient control. The suggested SOA-MASTGNN solution aims to achieve lower total energy costs and better system operation by optimizing PV utilization, EVC scheduling, and grid-battery interactions. The dynamically assigned energy resources are applied via SOA to optimize PV energy consumption and EVC schedules, thereby enhancing power flow and system reliability. Simultaneously, MASTGNN predicts solar power production, EV demand, and battery State-of-Charge (SoC) through adaptive spatio-temporal learning, enabling planning and coordination of energy activities in advance. MATLAB-based simulated models are utilized to evaluate the integrated model as well as compare it with the previous techniques, such as Giza Pyramids Construction-Recalling-Enhanced Recurrent Neural Network (GPC-RERNN), Deep Reinforcement Learning (DRL), Gannet Optimization Algorithm-Tree Hierarchical Deep Convolutional Neural Network (GOA-THDCNN), Shell Game Optimization and RERNN (SGO-RERNN), and CNN-Long Short-Term Memory (CNN-LSTM). Results show that SOA-MASTGNN has the lowest total energy cost of 132.52 c€ and the best efficiency of 99.2%, which are better than all baseline models. This indicates the suitability of the method for maximizing energy utilization, reducing operational expenses, and integrating RE in PV-assisted EV parking lots within reconfigurable MG environments.</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

Adaptive Energy Optimization for PV-Integrated EV Parking Lots in a Reconfigurable Microgrid Based on the Serval Optimization and Multimodal Adaptive Spatio-Temporal Graph Neural Network

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Chinthalacheruvu Venkata Krishna Reddy,
  • Venkatesh Kumar Chandrasekaran,
  • Gagan Kumar Koduru,
  • L. Guganathan

摘要

The intermittent nature of renewable generation, stochastic EV charging (EVC) behaviour, and the dynamic behaviour of battery energy storage systems (BESS) pose challenges for energy management (EM) in photovoltaic (PV)-integrated electric vehicle (EV) parking lots. The paper introduces a new hybrid EM model that combines the Serval Optimization Algorithm (SOA) and the Multimodal Adaptive Spatio-Temporal Graph Neural Network (MASTGNN) to enable a reconfigurable microgrid (MG) to achieve adaptive, efficient control. The suggested SOA-MASTGNN solution aims to achieve lower total energy costs and better system operation by optimizing PV utilization, EVC scheduling, and grid-battery interactions. The dynamically assigned energy resources are applied via SOA to optimize PV energy consumption and EVC schedules, thereby enhancing power flow and system reliability. Simultaneously, MASTGNN predicts solar power production, EV demand, and battery State-of-Charge (SoC) through adaptive spatio-temporal learning, enabling planning and coordination of energy activities in advance. MATLAB-based simulated models are utilized to evaluate the integrated model as well as compare it with the previous techniques, such as Giza Pyramids Construction-Recalling-Enhanced Recurrent Neural Network (GPC-RERNN), Deep Reinforcement Learning (DRL), Gannet Optimization Algorithm-Tree Hierarchical Deep Convolutional Neural Network (GOA-THDCNN), Shell Game Optimization and RERNN (SGO-RERNN), and CNN-Long Short-Term Memory (CNN-LSTM). Results show that SOA-MASTGNN has the lowest total energy cost of 132.52 c€ and the best efficiency of 99.2%, which are better than all baseline models. This indicates the suitability of the method for maximizing energy utilization, reducing operational expenses, and integrating RE in PV-assisted EV parking lots within reconfigurable MG environments.