Scholarlink

In the field of short-term electricity load forecasting for microgrids, accurate forecasting is crucial for the stable operation of electricity grids and the improvement of energy efficiency. The selection of suitable features is a crucial step in the forecasting process as it significantly influences the accuracy and efficiency of the model’s predictions. To improve the selection of features, the study proposes the DC-ATTN-LSTM model, which includes the distance correlation (DC), the attention mechanism (ATTN) and the Long Short-Term Memory (LSTM) to improve the accuracy of short-term load forecasts. First, the DC method philtres the multivariate time series data of the microgrid to identify important features related to the electrical loads. An attention mechanism then refines these features and focuses the model on forecast-relevant information. These prioritised features are proposed to a newly designed LSTM network, which is designed to learn the temporal dependencies and non-linear patterns in the load data using advanced deep learning techniques. Energy consumption data from Tetouan, Morocco, is used for empirical validation to compare the DC-ATTN-LSTM model with traditional LSTM models and other prediction techniques. The results show that the DC-ATTN-LSTM model outperforms the benchmark models on various assessment measures, confirming its effectiveness in short-term electrical load forecasting.

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

Short-Term Load Forecasting Using Hybrid Distance Correlation and LSTM

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhongpeng Li,
  • Xiaoliang Wang,
  • Hanjing Cheng,
  • Yehui Zhu,
  • Ke-cai Cao,
  • Hongjie Wu,
  • Juping Gu

摘要

In the field of short-term electricity load forecasting for microgrids, accurate forecasting is crucial for the stable operation of electricity grids and the improvement of energy efficiency. The selection of suitable features is a crucial step in the forecasting process as it significantly influences the accuracy and efficiency of the model’s predictions. To improve the selection of features, the study proposes the DC-ATTN-LSTM model, which includes the distance correlation (DC), the attention mechanism (ATTN) and the Long Short-Term Memory (LSTM) to improve the accuracy of short-term load forecasts. First, the DC method philtres the multivariate time series data of the microgrid to identify important features related to the electrical loads. An attention mechanism then refines these features and focuses the model on forecast-relevant information. These prioritised features are proposed to a newly designed LSTM network, which is designed to learn the temporal dependencies and non-linear patterns in the load data using advanced deep learning techniques. Energy consumption data from Tetouan, Morocco, is used for empirical validation to compare the DC-ATTN-LSTM model with traditional LSTM models and other prediction techniques. The results show that the DC-ATTN-LSTM model outperforms the benchmark models on various assessment measures, confirming its effectiveness in short-term electrical load forecasting.