融合SBAS-InSAR与CS-SVM的矿区地表残余沉降预测模型

Surface Residual Subsidence Prediction Model for Mining Area Based on the Fusion of SBAS-InSAR and CS-SVM

煤矿开采地表残余形变可能对地表建(构)筑物、道路、地下管线等基础设施造成潜在威胁,有必要对其进行准确预测。融合SBAS-InSAR监测方法,提出了一种布谷鸟搜索算法改进支持向量机回归(CS-SVM)的预测模型,利用2017年11月—2020年6月的60景Sentinel-1A SAR影像对安徽省某矿7221工作面进行开采沉陷长时序监测,获取了该工作面回采过程中与停采后2 a内地表年均形变速率与累计形变。结果表明:该工作面最大年均形变速率为-56 mm/a,最大累计沉降为151 mm。利用水准测量数据对InSAR结果进行验证,两者残差均小于5 mm,证明了两者具有较好的一致性。为比较优化前后SVM预测模型的精度,引入平均绝对误差和均方根误差进行精度评价。结果显示:优化模型的2种误差均在4 mm以内,相比传统模型,误差分别降低了59%和60%,预测精度明显提高。研究反映出,所提模型具有较好的预测能力,可为废弃煤矿区防灾减灾提供参考。

The surface residual deformation caused by coal mining activities may pose a potential threat to surface structures,roads,underground pipelines,and other infrastructure.Accurate prediction of this deformation is necessary.This study proposes an improved predictive model using the Cuckoo Search algorithm to enhance Support Vector Machine regression(CSSVM)based on the monitoring results of SBAS-InSAR.Sixty Sentinel-1A SAR images from November 2017 to June 2020 were utilized to monitor the long-term subsidence of the 7221 mining face in a coal mine in Anhui Province.The surface average annual deformation rates and cumulative deformation within 2 years after cessation of mining were obtained.The results indicate that the maximum average annual deformation rate of the mining face is-56 mm/a,and the maximum cumulative subsidence is 151 mm.Validation of the InSAR results was performed using leveling measurement data,and the residual values were both less than 5 mm,demonstrating good consistency between the two methods.To evaluate the prediction model′s accuracy before and after optimization,the average absolute error and root mean square error were introduced as evaluation criteria.The results demonstrate that both errors of the optimized model were within 4 mm,representing a 59%and 60%reduction in errors compared to the traditional model,resulting in a significantly improved prediction accuracy.This study shows that the proposed method exhibits strong predictive capability and can serve as a reference for disaster prevention and mitigation in abandoned coal mining areas.