| 引用本文: | 王子轩,欧 斌,陈德辉,等.基于深层挖掘变形时间序列的大坝预测模型[J].灌溉排水学报,2025,44(7):70-79. |
| WANG Zixuan,OU Bin,CHEN Dehui,et al.基于深层挖掘变形时间序列的大坝预测模型[J].灌溉排水学报,2025,44(7):70-79. |
|
| 摘要: |
| 【目的】针对大坝变形中存在的系统噪声及强非线性影响，提出一种深层挖掘变形时间序列的大坝变形预测模型。【方法】利用样本熵重构与K-means聚类算法优化自适应噪声完全集合经验模态分解（CEEMDAN）过程，得到若干模态分量函数（IMF），针对高频模态分量采用变分模态分解（VMD）进行二次分解处理，从而提取出最佳IMF分量。结合改进的共生生物搜索算法（ISOS）和双向门控循环单元（BiGRU）进行大坝变形的精确预测。【结果】与传统预测模型相比，本研究提出的模型均方根误差（RMSE）、平均绝对误差（MAE）、平均绝对百分误差（MAPE）和决定系数（R2）分别为0.031 9 mm、0.015 3 mm、2.51%和0.971 2。【结论】本研究提出的预测模型可准确模拟大坝形变过程，具有更高的预测精度和更强的泛化能力。 |
| 关键词: 大坝变形；自适应噪声完全集合经验模态分解；样本熵重构；K-means聚类算法；共生生物搜索算法；变分模态分解 |
| DOI：10.13522/j.cnki.ggps.2024213 |
| 分类号: |
| 基金项目: |
|
| Prediction of dam deformation using adaptive noise CEEMDAN and BiGRU time series modeling |
|
WANG Zixuan, OU Bin, CHEN Dehui, YANG Shiyong, ZHAO Dingzhu, FU Shuyan
|
|
1. College of Water Conservancy, Yunnan Agricultural University, Kunming 650201, China;
2. The National Key Laboratory of Water Disaster Prevention, Nanjing 210098, China;
3. Research Center for Smart Management and Maintenance of Small and
Medium-sized Water Conservancy Projects in Yunnan Province, Kunming 650201, China
|
| Abstract: |
| 【Background and Objective】Accurate prediction of dam deformation is crucial for ensuring the safety of dam structures in engineering monitoring. Dam deformation is influenced by multiple factors, including water pressure, temperature, and material aging, which often exhibit nonlinear and dynamic relationships. During monitoring, system noise and observation errors frequently interfere with data quality, posing additional challenges for analysis. To address the challenges posed by system noise and strong nonlinear effects in dam deformation, this paper proposes a dam deformation monitoring model based on multi-layer integrated signal processing technology.【Method】The model uses sample entropy reconstruction and the K-means clustering algorithm to optimize the adaptive noise complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) process, generating multiple intrinsic mode functions (IMF). High-frequency modal components undergo secondary decomposition using variational mode decomposition (VMD) to extract the optimal intrinsic mode function. Finally, an improved symbiotic biological search algorithm combined with a Bidirectional Gated Recurrent Unit (BiGRU) is used to accurately predict dam deformation.【Result】Case analysis demonstrates that, compared to traditional prediction models, the proposed model achieves a root mean square error (RMSE) of 0.031 9 mm, mean absolute error (MAE) of 0.015 3 mm, mean absolute percentage error (MAPE) of 2.51%, and determination coefficient (R2) of 0.971 2.【Conclusion】 The results verify that the proposed model captures and simulates the dam deformation process more accurately, exhibiting higher prediction accuracy and stronger generalization ability. |
| Key words: dam deformation; complete ensemble empirical mode decomposition of adaptive noise; sample entropy reconstruction; K-means clustering algorithm; symbiotic search algorithm; variational mode decomposition |
