基于深度学习的尾矿坝位移预测模型研究与应用

现代矿业 ›› 2025, Vol. 41 ›› Issue (11): 157-161，168.

基于深度学习的尾矿坝位移预测模型研究与应用

刘常诺黄明轩程子煜杨玉好杨斌

辽宁科技大学土木工程学院

出版日期:2025-11-25 发布日期:2025-12-23

Research and Application on Tailings Dam Displacement Prediction Model Based on Deep Learning

School of Civil Engineering，University of Science and Technology Liaoning

Online:2025-11-25 Published:2025-12-23

摘要/Abstract

摘要： 为掌握坝体变形趋势，避免尾矿库溃坝，构建基于 N-BEATSx 的尾矿坝位移预测模型。首先，对原始监测数据进行预处理，采用 Z-Score 法识别并剔除异常值，利用 Savitzky-Golay 滤波器对剔除异常值的数据进行平滑处理，并通过线性插值方法填补缺失数据，保证数据的完整性；然后，利用灰色关联度（GRA）分析坝体变形的影响因素，确定温度、时间、库水压、降雨等 10种影响因素作为输入变量；最后，使用 N-BEATSx 模型对坝体未来变形趋势进行预测。结果表明：该模型在使用 30 d 的历史数据，对未来 30 d 的坝体位移变化趋势进行预测时，取得了较高的预测精度，水平位移预测的均方根误差平均为 0.098 mm，均方误差平均为 0.012 mm²；竖直位移预测的均方根误差平均为 0.015 mm，均方误差平均为 0.025 mm²。该研究可以为尾矿坝位移预测与管理提供一定的参考。

关键词: N-BEATSx模型, 灰色关联度, 坝体位移预测

Abstract: To grasp the deformation trend of the dam body and avoid the dam failure of the tailings pond，a displacement prediction model of the tailings dam based on N-BEATSx is constructed. Firstly，the original monitoring data is preprocessed. The Z-Score method is adopted to identify and eliminate outliers. The Savitzky-Golay filter is used to smooth the data with excluded outliers，and the missing data is filled in through the linear interpolation method to ensure the integrity of the data. Then，the influencing factors of dam body deformation were analyzed by using the Grey Relational Degree（GRA），and ten influencing fac⁃ tors such as temperature，time，reservoir water pressure and rainfall were determined as input variables. Fi⁃ nally，the N-BEATSx model is used to predict the future deformation trend of the dam body. The results show that when this model uses 30 days of historical data to predict the displacement change trend of the dam body in the next 30 days，it achieves high prediction accuracy. The average root mean square error of the horizontal displacement prediction is 0.098 mm，and the average mean square error is 0.012 mm². The average root mean square error of vertical displacement prediction is 0.015 mm，and the average mean square error is 0.025 mm². This research can provide certain references for the displacement prediction and management of tailings dam.

Key words: N-BEATSx model, Grey Relational Degree, dam displacement prediction

刘常诺, 黄明轩, 程子煜, 杨玉好, 杨斌. 基于深度学习的尾矿坝位移预测模型研究与应用[J]. 现代矿业, 2025, 41(11): 157-161，168.

LIU Changnuo HUANG Mingxuan CHENG Ziyu YANG Yuhao YANG Bin. Research and Application on Tailings Dam Displacement Prediction Model Based on Deep Learning[J]. Modern Mining, 2025, 41(11): 157-161，168.

扫码分享