Scholarlink

<p>The problem of water shortage is becoming increasingly serious, especially in Henan, a major agricultural province, where the challenge of water demand is huge. The development of cloud water resources is an important measure to alleviate this problem, and artificial rain enhancement is the main approach. The precise grasp of the operation timing is crucial for improving the cloud-water conversion efficiency, and enhancing the prediction accuracy of PWV is an urgent problem to be solved. To this end, this paper proposes a novel multi-layer stacked prediction model, SVMD-TCN-LSTM-SA, which innovatively combines successive variational mode decomposition (SVMD), temporal convolutional network (TCN), long short-term memory network (LSTM), and self-attention mechanism (SA). Firstly, SVMD decomposes the PWV sequence into multiple subsequences to reduce the complexity of the data. Then, these sub-sequences are input into TCN to extract local features. Subsequently, long-term dependencies are captured through LSTM, and SA is utilized to automatically allocate weights based on the importance of time steps to enhance the focus on key information. After reconstructing the prediction results of all subsequences, the optimal predicted value is obtained. Through multi-step prediction experiments with multiple comparison models on different datasets, the results show that the prediction accuracy of the proposed model has been significantly improved. The MAE and RMSE have decreased by at least 25.3% and 20.2% respectively, verifying the applicability and stability of the model. It provides an efficient solution for PWV prediction.</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

Construction of a new multi-layer stacked model for precipitable water vapor prediction

  • 期刊论文

  • 发表日期：

    2025年8月11日

查看全文
  • Bingyan Li,
  • Zhenping Shao,
  • Kequan Zhang,
  • Mingjun Li,
  • Chenglin Yang,
  • Wenyu Zhang

摘要

The problem of water shortage is becoming increasingly serious, especially in Henan, a major agricultural province, where the challenge of water demand is huge. The development of cloud water resources is an important measure to alleviate this problem, and artificial rain enhancement is the main approach. The precise grasp of the operation timing is crucial for improving the cloud-water conversion efficiency, and enhancing the prediction accuracy of PWV is an urgent problem to be solved. To this end, this paper proposes a novel multi-layer stacked prediction model, SVMD-TCN-LSTM-SA, which innovatively combines successive variational mode decomposition (SVMD), temporal convolutional network (TCN), long short-term memory network (LSTM), and self-attention mechanism (SA). Firstly, SVMD decomposes the PWV sequence into multiple subsequences to reduce the complexity of the data. Then, these sub-sequences are input into TCN to extract local features. Subsequently, long-term dependencies are captured through LSTM, and SA is utilized to automatically allocate weights based on the importance of time steps to enhance the focus on key information. After reconstructing the prediction results of all subsequences, the optimal predicted value is obtained. Through multi-step prediction experiments with multiple comparison models on different datasets, the results show that the prediction accuracy of the proposed model has been significantly improved. The MAE and RMSE have decreased by at least 25.3% and 20.2% respectively, verifying the applicability and stability of the model. It provides an efficient solution for PWV prediction.