Research on predicting surface displacement of landslides based on CNN-BiLSTM-Attention in the Three Gorges reservoir area
-
摘要:
地表位移预测在滑坡监测预警中具有重要意义，建立稳定可靠的滑坡位移预测模型是关键。本文基于卷积神经网络和注意力机制的滑坡位移预测方法，并以三峡库区黄泥巴蹬坎滑坡为例进行了验证。本文综合分析了该滑坡长达8年的降雨量、库水位和地表位移等监测数据，建立了结合卷积神经网络（convolutional neural network, CNN）、双向长短期记忆（bidirectional long short-term memory, BiLSTM）网络和注意力机制（attention）的CNN-BiLSTM-Attention深度学习组合预测模型，采用了适应性学习率和正则化技术进行模型训练，提高了模型的泛化能力同时避免过拟合，并与传统LSTM模型进行对比验证。结果表明：相较于传统的机器学习和神经网络方法，该模型在滑坡位移预测精度上取得了显著提升，预测模型拟合优度（R2）达0.989，平均绝对百分比误差（MAPE）仅为0.059。
Abstract:Surface displacement prediction is of great significance in landslide monitoring and early warning, and establishing a stable and reliable landslide displacement prediction model is crucial. This paper utilizes a convolutional neural network (CNN) and attention mechanism to predict landslide displacement, and takes the Huangniba Dengkan landslide in the Three Gorges reservoir area as an example for verification. This paper comprehensively analyzes the landslide's monitoring data on rainfall, reservoir water level, and surface displacement for 8 years. It establishes a CNN-BiLSTM-Attention deep learning combination prediction model, and uses adaptive learning rate and regularization techniques for model training, improving the generalization ability of the model while avoiding overfitting. Additionally, the model is subjected to comparative validation with the traditional long short-term memory (LSTM) model. The results show that the model's landslide displacement prediction accuracy has been significantly enhanced compared to traditional machine learning and neural network methods. The prediction model's goodness of fit (R2) reaches 0.989, and the mean absolute percentage error (MAPE) is merely 0.059.
-
-
表 1 模型参数设置表
Table 1. Configuration parameters for CNN-BiLSTM-Attention
参数名称 参数值 核心参数 输入维度（input_dim） 3 输出维度（units） 128 输入序列长度（input_length） 8 随机失活（dropout） 0.2 训练参数 迭代次数（epochs） 20 学习率（learning rate） 0.001 每次学习样本数量（batch_size） 32 验证集占比（validation_split） 0.2 
下载: 导出CSV
表 2 CNN-BiLSTM-Attention预测模型性能指标
Table 2. Performance index of the CNN-BiLSTM-Attention prediction model
考核指标 GPS1点位 GPS4点位 拟合优度（R2） 0.989 0.995 平均绝对百分比误差（MAPE） 0.059 0.059 均方根误差（RMSE） 0.030 0.021 平均绝对误差（MAE） 0.059 0.092
下载: 导出CSV
表 3 LSTM预测模型性能指标
Table 3. Performance index of the LSTM prediction model
考核指标 GPS1点位 GPS4点位 拟合优度（R2） 0.720 0.719 平均绝对百分比误差（MAPE） 0.224 0.084 均方根误差（RMSE） 14.056 35.529 平均绝对误差（MAE） 11.929 33.096
下载: 导出CSV
-
[1] Alizadeh B,Bafti A G,Kamangir H,et al.,2021. A novel attention-based LSTM cell post-processor coupled with bayesian optimization for streamflow prediction[J]. Journal of Hydrology,601:126526. doi: 10.1016/j.jhydrol.2021.126526
[2] Corominas J,Moya J,Ledesma A,et al.,2005. Prediction of ground displacements and velocities from groundwater level changes at the Vallcebre landslide (Eastern Pyrenees,Spain)[J]. Landslides, 2(2):83 − 96.
[3] 邓国仕,郑万模,杨桂花,等,2011. 四川丹巴县甲居滑坡GPS监测结果及分析[J]. 沉积与特提斯地质,31(2):99 − 104. doi: 10.3969/j.issn.1009-3850.2011.02.014
Deng G S,Zheng W M,Yang G H,et al.,2011. GPS monitoring of the Jiaju landslide in Danba,Sichuan[J]. Sedimentary Geology and Tethyan Geology,31(2):99 − 104 (in Chinese with English abstract). doi: 10.3969/j.issn.1009-3850.2011.02.014
[4] Hochreite S,Schmidhuber J,1997. Long short-term memory[J]. Neural Computer, 9(8):1735 − 1780.
[5] Huang F M,Huang J S,Jiang S H,et al.,2017. Landslide displacement prediction based on multivariate chaotic model and extreme learning machine[J]. Engineering Geology,218:173 − 186. doi: 10.1016/j.enggeo.2017.01.016
[6] Kayacan E,Ulutas B,Kaynak O,2010. Grey system theory-based models in time series prediction[J]. Expert Systems with Applications,37(2):1784 − 1789. doi: 10.1016/j.eswa.2009.07.064
[7] Li H J,Xu Q,He Y S,et al.,2018. Prediction of landslide displacement with an ensemble based extreme learning machine and copula models[J]. Landslides,15:2047 − 2059. doi: 10.1007/s10346-018-1020-2
[8] Lian C,Zeng Z G,Yao W,et al.,2015. Multiple neural networks switched prediction for landslide displacement[J]. Engineering Geology,186:91 − 99. doi: 10.1016/j.enggeo.2014.11.014
[9] Liu G,Guo J B,2019. Bidirectional LSTM with attention mechanism and convolutional layer for text classification[J]. Neurocomputing,337(C):325 − 338.
[10] Ma T S,Xiang G F,Shi Y F,et al.,2022. Horizontal in situ stresses prediction using a cnn-bilstm-attention hybrid neural network[J]. Geomechanics and Geophysics for Geo-Energy and Geo-Resources, 8(5):1 − 26.
[11] Medsker L R,Jain L C,2001. Recurrent neural networks[J]. Design and Applications,5:64 − 67.
[12] Miao F S,Wu Y P,Xie Y H,et al.,2018. Prediction of landslide displacement with step-like behavior based on multialgorithm optimization and a support vector regression model[J]. Landslides, 15(3):475 − 488.
[13] Mufundirwa A,Fujii Y,Kodama J,2010. A new practical method for prediction of geomechanical failure-time[J]. International Journal of Rock Mechanics And Mining Sciences,47(7):1079 − 1090. doi: 10.1016/j.ijrmms.2010.07.001
[14] Neil D,Pfeiffer M,Liu S C,2016. Phased LSTM:Accelerating Recurrent Network Training for Long or Event-based Sequences[C]. NIPS'16:Proceedings of the 30th International Conference on Neural Information Processing Systems,3889 − 3897.
[15] Peter Naur,1975. Concise survey of computer methods [M]. New York:Petrocelli Books.
[16] Reichenbach P,Rossi M,Malamud B,et al.,2018. A review of statistically based landslide susceptibility models[J]. Earth-Science Reviews,180:60 − 91. doi: 10.1016/j.earscirev.2018.03.001
[17] Saito M,1965. Forecasting the time of occurrence of a slope failure. In:Proceedings of the 6th International Mechanics and Foundation Engineering [C]. 537 − 541.
[18] Thiebers H,2014. Integration of a limit-equilibrium model into a landslide early warning system[J]. Landslides,11(5):859 − 875. doi: 10.1007/s10346-013-0416-2
[19] 王家柱,高延超,铁永波,等,2023. 基于斜坡单元的山区城镇滑坡灾害易发性评价:以康定为例[J]. 沉积与特提斯地质,43(3):640 − 650.
Wang J Z,Gao Y C,Tie Y B,et al.,2023. Landslide susceptibility assessment based on slope units of mountainous cities and towns:A case study of Kangding city[J]. Sedimentary Geology and Tethyan Geology,43(3):640 − 650 (in Chinese with English abstract).
[20] 许强,陈建君,冯文凯,等,2009. 斜坡地震响应的物理模拟试验研究[J]. 四川大学学报:工程科学版,41(3):266 − 272.
Xu Q,Chen J J,Feng W K,et al.,2009. Study of the Seismic Response of Slopes by Physical Modeling[J]. Journal of Sichuan University(Engineering Science Edition),41(3):266 − 272 (in Chinese with English abstract).
[21] 杨背背,殷坤龙,杜娟,2018. 基于时间序列与长短时记忆网络的滑坡位移动态预测模型[J]. 岩石力学与工程学报,37(10):2334 − 2343.
Yang B B,Yin K L,Du J,et al.,2018. A model for predicting landslide displacement based on time series and long and short term memory neural network[J]. Chinese Journal of Rock Mechanics and Engineering,37(10):2334 − 2343 (in Chinese with English abstract).
[22] Yang B B,Yin K L,Lacasse S,et al.,2019. Time series analysis and long short-term memory neural network to predict landslide displacement[J]. Landslides,16:677 − 694. doi: 10.1007/s10346-018-01127-x
[23] 殷坤龙,晏同珍,1996. 滑坡预测及相关模型[J]. 岩石力学与工程学报,15(1):1 − 8. doi: 10.3321/j.issn:1000-6915.1996.01.001
Yin K L,Yan T Z,1996. Landslide prediction and relevant models[J]. Chinese Journal of Rock Mechanics and Engineering,15(1):1 − 8 (in Chinese with English abstract). doi: 10.3321/j.issn:1000-6915.1996.01.001
[24] Zhang W Y,Zhou H Y,Bao X H,et al.,2023. Outlet water temperature prediction of energy pile based on spatial-temporal feature extraction through CNN–LSTM hybrid model[J]. Energy,264(1):126 − 190.
[25] Zhou C,Yin K L,Cao Y,et al.,2016. Application of time series analysis and PSO-SVM model in predicting the Bazimen landslide in the Three Gorges Reservoir, China[J]. Engineering Geology,204:108 − 120. doi: 10.1016/j.enggeo.2016.02.009
-
图(9)
表(3)
计量
- 文章访问数: 837
- PDF下载数: 187
- 施引文献: 0