Scholarlink

<p>Short-term electricity load forecasting plays a crucial role in maintaining the stable operation of the power system and optimizing resource allocation. To enhance forecast accuracy and effectively capture the temporal features of load sequences, this paper proposes a method based on Successive Variational Mode Decomposition (SVMD), optimized using the Hiking Optimization Algorithm (HOA), for extracting predictable components from load sequences. The method decomposes the complex load data into several relatively stable subsequences, revealing load features at different time scales. To mitigate high-frequency noise and improve the signal’s feature representation capability, we introduce a Multi-Band Spectral Block (MBSB) method based on Fourier analysis. Based on this method, we further propose a hybrid forecasting model that integrates MBSB with the Informer framework. The model thoroughly analyzes the temporal features of each load component, which enhances the accuracy of short-term load forecasting. Through multiple experiments on datasets from China, Belgium, and Germany, our method outperforms the comparison models on all three datasets. The results show that the MBSB-Informer model significantly reduces the MAE by 42.55%, 42.06%, 27.97%, 23.93%, and 14.11% compared to BiLSTM, GRU, DLinear, Informer, and iTransformer, respectively. These results substantiate that the model demonstrates high predictive accuracy and stability. The substantial improvement in forecasting accuracy is crucial for enhancing power system reliability, optimizing energy management, and reducing operational costs, thereby providing significant practical value in real-world applications.</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

Short-term load forecasting based on successive variational mode decomposition and multi-band feature extraction

  • 期刊论文

  • 发表日期：

    2025年11月27日

查看全文
  • Nanwei Ding,
  • Ming Li,
  • Cong Zhou,
  • Ruijun Tie

摘要

Short-term electricity load forecasting plays a crucial role in maintaining the stable operation of the power system and optimizing resource allocation. To enhance forecast accuracy and effectively capture the temporal features of load sequences, this paper proposes a method based on Successive Variational Mode Decomposition (SVMD), optimized using the Hiking Optimization Algorithm (HOA), for extracting predictable components from load sequences. The method decomposes the complex load data into several relatively stable subsequences, revealing load features at different time scales. To mitigate high-frequency noise and improve the signal’s feature representation capability, we introduce a Multi-Band Spectral Block (MBSB) method based on Fourier analysis. Based on this method, we further propose a hybrid forecasting model that integrates MBSB with the Informer framework. The model thoroughly analyzes the temporal features of each load component, which enhances the accuracy of short-term load forecasting. Through multiple experiments on datasets from China, Belgium, and Germany, our method outperforms the comparison models on all three datasets. The results show that the MBSB-Informer model significantly reduces the MAE by 42.55%, 42.06%, 27.97%, 23.93%, and 14.11% compared to BiLSTM, GRU, DLinear, Informer, and iTransformer, respectively. These results substantiate that the model demonstrates high predictive accuracy and stability. The substantial improvement in forecasting accuracy is crucial for enhancing power system reliability, optimizing energy management, and reducing operational costs, thereby providing significant practical value in real-world applications.