Scholarlink

<p>Accurate electricity price forecasting is important for efficient energy market operations and strategic decision-making by power generators, retailers, and other market participants. Precise prediction of the electricity price can be allowed by applying advanced deep learning (DL) techniques, namely Long Short-Term Memory (LSTM) and multivariate Convolutional Neural Network-LSTM models. The innovations in this article are related to the comparative analysis between these two models, which showed the better performance of the CNN-LSTM model in the capture of complex temporal dependencies and patterns in the data of electricity prices. The scores for the LSTM model are 2.483 for RMSE, 6.166 for MSE, 1.951 for MAE, and 2.812% for MAPE, while for the CNN-LSTM model, which has improved metrics, the results are as follows: 2.443 for RMSE, 5.97 for MSE, 1.895 for MAE, and 2.739% for MAPE. The present research goes ahead and provides a detailed performance analysis of the models. This is through the use of loss plots, tracking training, and validation losses, hence ensuring model robustness and that overfitting does not take place. Identification of the significant wide-scale predictors of the price of electricity is determined using correlation analysis, hence a better comprehension of the dynamics of prices. Results of this kind do carry much constructive insight into developing more reliable and more accurate forecasting models in energy.</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

Using AI for Forecasting Electricity Price Forecasting: Challenges and Innovations of Deep Learning Methods

  • 期刊论文

  • 发表日期：

    2025年6月27日

查看全文
  • Kai Guo,
  • Yuting Zhang,
  • Entong Li

摘要

Accurate electricity price forecasting is important for efficient energy market operations and strategic decision-making by power generators, retailers, and other market participants. Precise prediction of the electricity price can be allowed by applying advanced deep learning (DL) techniques, namely Long Short-Term Memory (LSTM) and multivariate Convolutional Neural Network-LSTM models. The innovations in this article are related to the comparative analysis between these two models, which showed the better performance of the CNN-LSTM model in the capture of complex temporal dependencies and patterns in the data of electricity prices. The scores for the LSTM model are 2.483 for RMSE, 6.166 for MSE, 1.951 for MAE, and 2.812% for MAPE, while for the CNN-LSTM model, which has improved metrics, the results are as follows: 2.443 for RMSE, 5.97 for MSE, 1.895 for MAE, and 2.739% for MAPE. The present research goes ahead and provides a detailed performance analysis of the models. This is through the use of loss plots, tracking training, and validation losses, hence ensuring model robustness and that overfitting does not take place. Identification of the significant wide-scale predictors of the price of electricity is determined using correlation analysis, hence a better comprehension of the dynamics of prices. Results of this kind do carry much constructive insight into developing more reliable and more accurate forecasting models in energy.