Scholarlink

<p>Electric vehicles (EVs) are becoming increasingly vital to modern power systems due to their substantial social and economic benefits. However, real-time Electric vehicle charging control remains challenging because of environmental uncertainties and the difficulty of optimizing multiple objectives simultaneously, such as reliability improvement, cost reduction, auxiliary service provision, and effective integration of renewable energy. To address these issues, we propose a predictive framework based on a Long Short-Term Memory fused Gated Recurrent Unit (LSTM-GRU) model for fuel economy prediction and charging control recommendation. Firstly, the simulation of Electric vehicles is conducted, and thereafter, pre-processing is accomplished. Afterwards, feature fusion (FF) is carried out utilizing Deep Q-network (DQN) with the Jaccard coefficient (JC). Then, data augmentation is performed utilizing oversampling method to increase dimensionality. After that, fuel economic prediction is done employing Long-Short Term Memory fused Gated recurrent unit, which is an amalgamation of Deep Long-Short Term Memory (DLSTM) and Gated Recurrent Unit (GRU). Finally, the recommendation for charging control is executed. In addition, Long-Short Term Memory fused Gated recurrent unit acquired minimum Mean Square Error (MSE), Mean&#xa0;Absolute&#xa0;Error (MAE) and Root Mean Square Error (RMSE) of 0.118, 0.109 and 0.343, demonstrating its effectiveness in supporting intelligent and reliable EV charging control. Furthermore, the performance improvement of the proposed LSTM-GRU method over the traditional methods, such as Artificial Neural Network (ANN), Gated Recurrent Unit (GRU), ensemble Stacked Generalization (ESG), Q-learning technique, LightGBM, and Group Learning Algorithm- convolutional neural network with bidirectional long short-term memory (GLA-CNN–Bi-LSTM), are 83.09%, 82.07%, 72.10%, 58.88%, 34.08%, and 13.87%.</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

Long-Short Term Memory and Gated Recurrent Unit-Based Fuel Economic Prediction and Charging Control Management of Electric Vehicles

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • S. Sumathi,
  • K. Mekala,
  • R. Gandhi,
  • S. Shobana

摘要

Electric vehicles (EVs) are becoming increasingly vital to modern power systems due to their substantial social and economic benefits. However, real-time Electric vehicle charging control remains challenging because of environmental uncertainties and the difficulty of optimizing multiple objectives simultaneously, such as reliability improvement, cost reduction, auxiliary service provision, and effective integration of renewable energy. To address these issues, we propose a predictive framework based on a Long Short-Term Memory fused Gated Recurrent Unit (LSTM-GRU) model for fuel economy prediction and charging control recommendation. Firstly, the simulation of Electric vehicles is conducted, and thereafter, pre-processing is accomplished. Afterwards, feature fusion (FF) is carried out utilizing Deep Q-network (DQN) with the Jaccard coefficient (JC). Then, data augmentation is performed utilizing oversampling method to increase dimensionality. After that, fuel economic prediction is done employing Long-Short Term Memory fused Gated recurrent unit, which is an amalgamation of Deep Long-Short Term Memory (DLSTM) and Gated Recurrent Unit (GRU). Finally, the recommendation for charging control is executed. In addition, Long-Short Term Memory fused Gated recurrent unit acquired minimum Mean Square Error (MSE), Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) of 0.118, 0.109 and 0.343, demonstrating its effectiveness in supporting intelligent and reliable EV charging control. Furthermore, the performance improvement of the proposed LSTM-GRU method over the traditional methods, such as Artificial Neural Network (ANN), Gated Recurrent Unit (GRU), ensemble Stacked Generalization (ESG), Q-learning technique, LightGBM, and Group Learning Algorithm- convolutional neural network with bidirectional long short-term memory (GLA-CNN–Bi-LSTM), are 83.09%, 82.07%, 72.10%, 58.88%, 34.08%, and 13.87%.