Scholarlink

<p>The rapid growth in Electric Vehicle (EV) adoption poses significant challenges to existing power grid infrastructure due to unpredictable, high-power charging demands that can cause grid stress, instability, and potential failures. Addressing these challenges requires accurate forecasting models that can support intelligent scheduling and load management. This research introduces a novel Hybrid Artificial Lizard Search Optimization–Squirrel Search Algorithm (HALSO-SSA) integrated with a Multi-Blend Bidirectional Recurrent Neural Network (RNN) to deliver highly precise EV charging demand predictions. The proposed framework begins with comprehensive data preprocessing, followed by intelligent feature selection using HALSO-SSA, which effectively identifies the most influential charging behavior features, reducing data redundancy and improving model generalization. This hybrid optimization significantly enhances the learning efficiency and stability of the deep neural network, enabling it to adapt to complex temporal patterns in EV charging profiles. By accurately forecasting charging demand, the proposed method supports proactive load balancing and grid stability, mitigating peak load stress and operational risks. Experimental results demonstrate outstanding prediction performance with Root Mean Squared Error (RMSE) of 0.083, Mean Absolute Error (MAE) of 0.076, Mean Squared Logarithmic Error (MSLE) of 0.0049 and Mean Squared Error (MSE) of 0.007, outperforming existing machine learning and deep learning baselines. This establishes HALSO-SSA–based predictive modeling as a promising strategy for sustainable EV-grid integration.</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

Addressing Power Grid Challenges with EV Charging Demand Prediction Using HALSO-SSA

  • 期刊论文

  • 发表日期：

    2025年12月26日

查看全文
  • P. Gayathri,
  • A. Ravi

摘要

The rapid growth in Electric Vehicle (EV) adoption poses significant challenges to existing power grid infrastructure due to unpredictable, high-power charging demands that can cause grid stress, instability, and potential failures. Addressing these challenges requires accurate forecasting models that can support intelligent scheduling and load management. This research introduces a novel Hybrid Artificial Lizard Search Optimization–Squirrel Search Algorithm (HALSO-SSA) integrated with a Multi-Blend Bidirectional Recurrent Neural Network (RNN) to deliver highly precise EV charging demand predictions. The proposed framework begins with comprehensive data preprocessing, followed by intelligent feature selection using HALSO-SSA, which effectively identifies the most influential charging behavior features, reducing data redundancy and improving model generalization. This hybrid optimization significantly enhances the learning efficiency and stability of the deep neural network, enabling it to adapt to complex temporal patterns in EV charging profiles. By accurately forecasting charging demand, the proposed method supports proactive load balancing and grid stability, mitigating peak load stress and operational risks. Experimental results demonstrate outstanding prediction performance with Root Mean Squared Error (RMSE) of 0.083, Mean Absolute Error (MAE) of 0.076, Mean Squared Logarithmic Error (MSLE) of 0.0049 and Mean Squared Error (MSE) of 0.007, outperforming existing machine learning and deep learning baselines. This establishes HALSO-SSA–based predictive modeling as a promising strategy for sustainable EV-grid integration.