Scholarlink

To ensure the reliable and cost-effective operation of integrated energy systems, accurate multi-load forecasting is crucial. In this paper, a method that combines the Sparrow Search Algorithm with the SE-CNN-LSTM model is proposed for multi-load forecasting. Firstly, qualitative analysis was conducted to identify dynamic coupling relationships among various loads, and a data-driven approach was used to select the influencing factors of multi-loads. Secondly, Convolutional Neural Networks(CNN)were introduced to extract and merge high-dimensional features, followed by the utilization of Long Short-Term Memory networks(LSTM)to further explore the temporal characteristics of multi-load data. Simultaneously, an SE attention mechanism module was incorporated to enhance the significance of critical information between channels. Finally, the Sparrow Search Algorithm(SSA)was employed to optimize the model’s hyperparameters, further enhancing the overall predictive performance. To validate the effectiveness of the model proposed in this paper, a comparison was made with four commonly used models. The results indicate that this model provides more desirable multi-load forecasts for integrated energy systems, outperforming other comparative models, thus providing a scientific basis for subsequent energy scheduling decisions.

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

Research on Multi-Energy Load Forecasting for Integrated Energy Systems Incorporating Sparrow Search Algorithm and SE-CNN-LSTM

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Erchao Li,
  • Minrui Liao

摘要

To ensure the reliable and cost-effective operation of integrated energy systems, accurate multi-load forecasting is crucial. In this paper, a method that combines the Sparrow Search Algorithm with the SE-CNN-LSTM model is proposed for multi-load forecasting. Firstly, qualitative analysis was conducted to identify dynamic coupling relationships among various loads, and a data-driven approach was used to select the influencing factors of multi-loads. Secondly, Convolutional Neural Networks(CNN)were introduced to extract and merge high-dimensional features, followed by the utilization of Long Short-Term Memory networks(LSTM)to further explore the temporal characteristics of multi-load data. Simultaneously, an SE attention mechanism module was incorporated to enhance the significance of critical information between channels. Finally, the Sparrow Search Algorithm(SSA)was employed to optimize the model’s hyperparameters, further enhancing the overall predictive performance. To validate the effectiveness of the model proposed in this paper, a comparison was made with four commonly used models. The results indicate that this model provides more desirable multi-load forecasts for integrated energy systems, outperforming other comparative models, thus providing a scientific basis for subsequent energy scheduling decisions.