Scholarlink

<p>Predicting irradiance for a solar electric vehicle is crucial for optimizing energy management and route planning, thereby enhancing the vehicle’s efficiency and autonomy by accounting for variations in weather conditions. This prediction was achieved using six machine learning algorithms: Support Vector Regressor, Random Forest, XGBoost, Multilayer Perceptron, a simple LSTM, and a novel hybrid model that integrates K-means clustering with a Stacked LSTM-BiLSTM Attention Network (SLA-Net). The model is trained and evaluated using various performance metrics, including Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), the coefficient of determination (<i>R</i><sup>2</sup>). Averaging the results across four regions, the hybrid model achieves an RMSE of 4.67, MAPE of 1.58, and an <i>R</i><sup>2</sup> of 0.99. These results demonstrate that the hybrid model outperforms the compared models, exhibiting superior prediction accuracy and significantly reduced error rates.</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

Advanced artificial intelligence model for solar irradiance forecasting for solar electric vehicles

  • 期刊论文

  • 发表日期：

    2025年2月11日

查看全文
  • Mohamed Abdellatif Khalfa,
  • Lazhar Manai,
  • Walid Mchara

摘要

Predicting irradiance for a solar electric vehicle is crucial for optimizing energy management and route planning, thereby enhancing the vehicle’s efficiency and autonomy by accounting for variations in weather conditions. This prediction was achieved using six machine learning algorithms: Support Vector Regressor, Random Forest, XGBoost, Multilayer Perceptron, a simple LSTM, and a novel hybrid model that integrates K-means clustering with a Stacked LSTM-BiLSTM Attention Network (SLA-Net). The model is trained and evaluated using various performance metrics, including Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), the coefficient of determination (R2). Averaging the results across four regions, the hybrid model achieves an RMSE of 4.67, MAPE of 1.58, and an R2 of 0.99. These results demonstrate that the hybrid model outperforms the compared models, exhibiting superior prediction accuracy and significantly reduced error rates.