摘要
残余变形对下伏老采空区的重大建(构)筑物安全运维带来潜在风险,可靠的变形预测对保障地下采煤区国土空间安全再利用具有重要意义。针对残余变形预测模型因建模数据获取困难或模型自身不足而对预测结果影响显著的问题,该文提出一种基于数据结构分解的GM(1,1)-BP-RCRM-CM模型。首先采用牛顿插值法对模型训练数据进行插值处理使其等间距;然后利用对称滑动平均法将老采空区残余沉降监测数据分解为趋势项和扰动项,采用自回归模型进行监测数据序列分解后前端丢失值的端值补充;其次分别采用GM(1,1)-RCRM模型和BP-RCRM模型进行趋势项预测和扰动项预测;最后组合趋势项与扰动项预测结果作为总体预测结果,并采用平均绝对百分比误差进行模型拟合精度与预测精度评价。实验结果表明:GM(1,1)-BP-RCRM-CM模型在不同监测点位的残余沉降预测中结果一致性较高,符合预测模型随周期递进时预测精度渐次下降的客观规律,预测结果更贴合实际情况,较GM(1,1)-RCRM模型和BP-RCRM模型更稳健。
Residual deformation brings potential risks to the safe operation and maintenance of major structures in the underlying old goaf.Reliable deformation prediction is of great significance to ensure the safe reuse of territorial space in underground coal mining areas.Aiming at the problem that the prediction model of residual deformation has a significant impact on the prediction result due to the difficulty in obtaining modeling data or the deficiency of the model itself,a GM(1,1)-BP-RCRM-CM model based on data structure decomposition is proposed in this paper.Firstly,Newton interpolation method was used to interpolate the model training data to make them equally spaced.Then the monitoring data of the old goaf residual subsidence was decomposed into trend and disturbance terms by symmetric sliding average method,and the end value of the front-end lost value after the decomposition of monitoring data series was supplemented by an autoregressive model.Secondly,GM(1,1)-RCRM model and BP-RCRM model were used to predict the trend term and the disturbance term respectively.Finally,the prediction results of trend item and disturbance item were combined as the overall prediction results,and the average absolute percentage error was used to evaluate the model fitting accuracy and prediction accuracy.The experimental results show that:The GM(1,1)-BP-RCRM-CM model has high consistency in predicting the residual settlement at different monitoring locations,which is in line with the objective law that the prediction accuracy of the prediction model decreases gradually as the period progresses.The predicted results are more in line with the actual situation,and more robust than GM(l,1)-RCRM model and BP-RCRM model.
作者
薛永安
邹友峰
冀哲
张文志
XUE Yong'an;ZOU Youfeng;JI Zhe;ZHANG Wenzhi(School of Surveying and Land Information Engineering,Henan Polytechnic University,Jiaozuo,Henan 454000,China;College of Mining Engineering,Taiyuan University of Technology,Taiyuan 030024,China;Henan Engineering Research Center for Ecological Restoration and Construction Technology of Goaf Sites,Jiaozuo,Henan 454000,China)
出处
《测绘科学》
CSCD
北大核心
2024年第4期43-56,共14页
Science of Surveying and Mapping
基金
国家自然科学基金山西煤基低碳联合基金重点项目(U1810203,U22A20620)
教育部产学合作协同育人资助项目(202102245009,22087106262449)。
关键词
老采空区
残余变形
组合预测
数据分解
残差修正
old goaf
residual deformation
combination prediction
data decomposition
residual correction
