摘要
以鄂尔多斯东部某碳矿的3014工作面作为研究对象,利用2018-09-15~2019-02-18的14景SAR数据对矿区进行监测,将小基线集技术(SBAS-InSAR)得到的时序累积沉降量作为训练样本,并与网格搜索算法(GS)优选超参数的长短期记忆网络(LSTM)进行结合,对矿区开采沉陷做出预测。结果表明:SBAS-InSAR技术反演值能够满足矿区开采沉陷预测,结合LSTM-GS模型对矿区开采沉陷进行预测,得到的最大均方根误差为3.569 3 mm,最大平均绝对误差为3.252 4 mm,最小决定系数为0.57,说明该模型预测精度能够满足工程需求。
Taking the 3104 working face of a coal mine in eastern Ordos as the research object,we used 14 SAR data from September 15,2018 to February 18,2019 to monitor the mining area.Taking the time series accumulated subsidence obtained by the small baseline set technology(SBAS-InSAR)as the training sample,and combining with the long short-term memory network(LSTM)of the grid search algorithm(GS)to opti-mize the hyperparameters,we predicted the mining subsidence.The results show that the inversion value of SBAS-InSAR technology can meet the mining subsidence prediction needs.Combined with the LSTM-GS model to predict the mining subsidence,the maximum root mean square error is 3.5693 mm,the maximum average absolute error is 3.2524 mm,and the minimum decision coefficient is 0.57,indicating that the predic-tion accuracy of model can meet engineering needs.
作者
惠甜甜
刘长星
王圣杰
郭一帆
HUI Tiantian;LIU Changxing;WANG Shengjie;GUO Yifan(College of Geomatic Science and Technology,Xi’an University of Science and Technology,Xi’an 710054,China)
出处
《地理空间信息》
2022年第8期13-17,共5页
Geospatial Information
