人民黄河2025，Vol.47Issue(3)：135-140,145,7.DOI:10.3969/j.issn.1000-1379.2025.03.021
考虑滞后效应的CNN-BIGRU-Attention预测降水型滑坡位移
CNN-BIGRU-Attention Prediction of Precipitation Type Landslide Displacement Considering Lag Effect
摘要
Abstract
In order to study the impact of precipitation on landslides,based on the daily precipitation and displacement data of Dashawao landslide,the displacement was decomposed into trend term displacement and periodic term displacement by using the moving average meth-od.Convolutional Neural Network(CNN)was used to predict trend term displacement.Convolutional Neural Network Bidirectional Gated Re-current Unit(CNN-BIGRU)model with Attention mechanism was used to predict periodic term displacement,and the final predicted dis-placement results were obtained by overlaying trend term displacement and periodic term displacement.Using Spearman correlation coefficient combined with lagged research to analyze the lagged relationship between variables.Using BIGRU-Attention,Gated Recurrent Unit(GRU),and Long Short Term Memory Network(LSTM)model as controls,the accuracy of the CNN-BIGRU-Attention model in predicting periodic term displacement was compared.The results show that the R2 values of the CNN model for predicting trend term displacement at 3,6,and 12 hours step length are 0.992,0.977,and 0.965,respectively.The R2 values of the CNN-BIGRU-Attention model for predicting the dis-placement of the 3,6,and 12 hours step length periodic term are 0.963,0.939,and 0.896,respectively,with higher prediction accuracy than the BIGRU-Attention,GRU,and LSTM models.Based on the landslide monitoring data of Xiarenyi Village,the generalization of CNN-BIGRU-Attention model was verified.关键词
位移预测/CNN/BIGRU/Attention/大沙窝滑坡/呷任依村滑坡Key words
displacement prediction/CNN/BIGRU/Attention/Dashawao landslide/Xiarenyi Village landslide分类
天文与地球科学引用本文复制引用
肖金涛,王自法,王超,赵登科,李兆焱..考虑滞后效应的CNN-BIGRU-Attention预测降水型滑坡位移[J].人民黄河,2025,47(3):135-140,145,7.基金项目
国家自然科学基金面上项目(52378544) （52378544）
中国地震局工程力学研究所基本科研业务费专项(2021B09) （2021B09）
