基于深度学习的地表形变预测

Prediction of Surface Deformation Based On Deep Learning

地表沉降属于一种动态且长期存在的地质变动现象,唐山南湖生态城区域在发展过程中出现了显著的地表沉降问题。为掌握研究区地面沉降趋势,通过构建ARIMA模型、SVM模型、LSTM模型,揭示变形监测数据蕴含的规律,从而针对监测区域的形变情况进行预测。采用RMSE、MAE和MAPE 3个评价指标对模型的预测结果进行精度评估。研究结果表明:与其他2个模型相比,LSTM模型的预测结果与原始值较为一致,LSTM模型的RMSE、MAE、MAPE值均为最小,模型预测精度高,满足研究区地表形变监测工程需求,可应用于实际生产中。该研究成果对当地的防灾减灾和城市发展建设具有非常重要的意义。

Land surface subsidence is a dynamic and long-term geological phenomenon,and significant land surface subsidence has been appeared in Nanhu Eco-city area of Tangshan during its development.In order to grasp the trend of land subsidence in the study area,ARIMA model,SVM model and LSTM model were constructed,and the law of deformation monitoring data was revealed to predict the subsidence in the monitoring area.RMSE,MAE and MAPE as three evaluation indexes,were used to evaluate the accuracy of the prediction results.The results show that LSTM model are consistent with the original value monitoring results.Compared with the other two models,the RMSE,MAE and MAPE values of LSTM model are the minimum,and the prediction accuracy of LSTM model is higher,which meets the demand of surface subsidence monitoring project in the study area and it can be applied to actual production.It is of great significance to local disaster prevention and reduction as well as urban development and construction.