| 深度学习在大坝变形预测中的应用研究 |
| |
| 引用本文： | 刘琼,李能.深度学习在大坝变形预测中的应用研究[J].测绘与空间地理信息,2020(3):201-203,207,210. |
| |
| 作者姓名： | 刘琼 李能 |
| |
| 作者单位： | 海南水文地质工程地质勘察院;陕西地矿区研院有限公司 |
| |
| 摘 要： | 大坝时间序列变形的高精度预测对于大坝运行维护及保护人民生命安全显得尤为重要。本文以某大坝113期变形时间序列数据为实验,提出了一种深度学习中的循环神经网络(LSTM)方法来进行大坝变形预测,将实验的结果与机器学习中NAR神经网络和ARIMA自回归移动平均模型的预测结果进行对比,LSTM、NAR和ARIMA模型的均方根误差(RMSE)分别为0.392 5、0.573 7、1.298 7;平均相对误差(MRE)分别为0.0498、0.1046、0.1878;R^2系数分别为0.932 3、0.822 1、0.247 7。从上述结果对比可知,LSTM时间序列预测模型的精度更高且稳定性更好,可作为后续大坝变形预测的一种新的思路和探索。
|
| 关 键 词： | 神经网络 LSTM NAR ARIMA |
Application of Dam Deformation Prediction Based on Deep Learning |
| |
| LIU Qiong,LI Neng.Application of Dam Deformation Prediction Based on Deep Learning[J].Geomatics & Spatial Information Technology,2020(3):201-203,207,210. |
| |
| Authors: | LIU Qiong LI Neng |
| |
| Institution: | (Hainan Hydrogeological Engineering Geological Survey Institute,Haikou 571100,China;Shaanxi Geology and Mining Area Research Institute Co.,Ltd.,Xianyang 712000,China) |
| |
| Abstract: | Accurate prediction of dam deformation is of great significance to the maintenance of dam operation and the protection of human lives. By using a 113 time series data of dam deformation in the experiment,this paper proposes a cycle of deep learning neural network( LSTM) method for dam deformation prediction. The prediction result of the LSTM was compared with NAR auto-regressive moving average model of neural network and ARIMA and the root mean square errors of LSTM,NAR and ARIMA models were 0.392 5,0.573 7,1.298 7. The mean relative errors were 0.049 8,0.104 6 and 0.187 8 and the fitting R^2 coefficients were 0.932 3,0.822 1 and 0.247 7,respectively. The comparison results show that the LSTM time series prediction model has higher accuracy and better stability,which can be used as a method for subsequent dam deformation prediction. |
| |
| Keywords: | neural network LSTM NAR ARIMA |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
