Scholarlink

Accurate forecasting of electric energy consumption is crucial for effective energy management, particularly in countries with diverse energy mixes like Spain. This study presents a deep learning (DL) approach to forecast Spanish electric energy consumption using a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) architecture enhanced with an attention mechanism. We introduce a comprehensive taxonomy of DL models for energy consumption forecasting and conduct a comparative analysis of seven architectures: Recurrent Neural Network, LSTM, Gated Recurrent Unit, Stacked LSTM, CNN, CNN-LSTM, and our proposed CNN-LSTM with an attention mechanism. Using a 4-year dataset of Spanish electrical system data, we evaluate model performance using Root Mean Square Error (RMSE). Our findings reveal that the CNN-LSTM with an attention mechanism significantly outperforms all other models, achieving the lowest RMSE of 68.54. This superior performance underscores the model’s ability to capture complex temporal dependencies and relevant features effectively. The study contributes to research on advanced forecasting techniques in the energy sector and offers practical insights for optimizing energy distribution and consumption in systems with high renewable energy penetration.

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

Energy Forecasting Utilizing CNN-LSTM Attention Mechanism: Empirical Evidence from the Spanish Electricity Market

  • 章节

  • 发表日期：

    2025年

查看全文
  • Ekramul Haque Tusher,
  • Jalal Uddin Md Akbar,
  • Riadul Islam Rabbi,
  • Mahmudul Hasan

摘要

Accurate forecasting of electric energy consumption is crucial for effective energy management, particularly in countries with diverse energy mixes like Spain. This study presents a deep learning (DL) approach to forecast Spanish electric energy consumption using a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) architecture enhanced with an attention mechanism. We introduce a comprehensive taxonomy of DL models for energy consumption forecasting and conduct a comparative analysis of seven architectures: Recurrent Neural Network, LSTM, Gated Recurrent Unit, Stacked LSTM, CNN, CNN-LSTM, and our proposed CNN-LSTM with an attention mechanism. Using a 4-year dataset of Spanish electrical system data, we evaluate model performance using Root Mean Square Error (RMSE). Our findings reveal that the CNN-LSTM with an attention mechanism significantly outperforms all other models, achieving the lowest RMSE of 68.54. This superior performance underscores the model’s ability to capture complex temporal dependencies and relevant features effectively. The study contributes to research on advanced forecasting techniques in the energy sector and offers practical insights for optimizing energy distribution and consumption in systems with high renewable energy penetration.