基于数据驱动模型的深圳填海区地面沉降预测分析  

PREDICTIVE ANALYSIS OF LAND SUBSIDENCE IN SHENZHEN RECLAMATION AREA BASED ON DATA-DRIVEN MODELS

作　　者：王艳 邓英尔[2] 高延超[3] WANG Yan;DENG Ying-er;GAO Yan-chao(Shenzhen Survey and Research Institute Co.,Ltd.,Guangdong,Shenzhen 518000,China;State Key Laboratory of Geohazard Prevention and Geoenvironment Protection,Chengdu University of Technology,Chengdu 610059,China;Chengdu Center,China Geological Survey,Chengdu 610081,China)

机构地区：[1]深圳市勘察研究院有限公司,深圳518000 [2]成都理工大学地质灾害防治与地质环境保护国家重点实验室,成都610059 [3]中国地质调查局成都地质调查中心,成都610081

出　　处：《地质灾害与环境保护》2024年第3期26-31,共6页Journal of Geological Hazards and Environment Preservation

基　　金：四川省自然科学基金(23NSFSC0297)。

摘　　要：为解决土地资源问题,深圳市的迅猛发展伴随着填海造陆工程。填海区海相淤泥层属于高饱和软土,高压缩性,上部荷载作用及回填土的自重固结作用易导致地面沉降地质灾害的发生,对填海区地面沉降问题的预测研究也变得至关重要。以空港新城区域2019年5月至2022年4月的监测数据为例,选择多种数学模型对地面沉降进行预测,包括NAR神经网络模型、GM(1,1)模型、多项式回归模型和ARIMA模型。预测表明:4种模型均能较好地反应出累计沉降量的未来发展趋势,可为后期地面沉降防治方案及地基处理等问题提供基础依据,有效降低地面沉降地质灾害的危险性。综合其预测结果的稳定性及精确度等因素,4种模型的优劣可按ARIMA模型、NAR神经网络模型、多项式回归模型、GM(1,1)模型顺次排序。To solve the problem of land resources, the rapid development of Shenzhen was accompanied by reclamation projects. The marine silt layer in the reclamation area is a highly saturated soft soil with high compressibility. The effect of upper loads and the self-weight consolidation of backfill can easily lead to the occurrence of geological hazards of ground settlement, and the prediction study of ground settlement problems in reclamation areas has become crucial. Taking the monitoring data from May 2019 to April 2022 in the Airport New City area as an example, various mathematical models were selected to predict the ground subsidence, including the NAR neural network model, the GM(1,1) model, the polynomial regression model, and the ARIMA model. The prediction shows that all four models can better reflect the future development trend of the cumulative settlement, which can provide a basic basis for the prevention and control program of ground settlement and foundation treatment in the later stage, and effectively reduce the risk of geological disasters of ground settlement. Combining the stability and accuracy of their prediction results and other factors, the advantages and disadvantages of the four models can be ranked in order of the ARIMA model, NAR neural network model, polynomial regression model, and GM(1,1) model.

关 键 词：地面沉降 数据驱动模型 预测 填海区 

分 类 号：P642.26[天文地球—工程地质学] P694[天文地球—地质矿产勘探] TU434[天文地球—地质学]