老采空区地表沉降预测合理监测模式分析

Analysis on the monitoring model of surface subsidence prediction in old mined-out area

为研究老采空区沉降监测数据时间间隔对预测精度的影响,本文利用某老采空区地表沉降监测点实测沉降数据,在等时间间隔、非等时间间隔两种情况下建立两种方案、六种沉降预测模式,采用长短期记忆神经网络(LSTM)预测模型对老采空区地表沉降进行预测,以平均绝对误差(MAE)和平均绝对百分比误差(MAPE)作为评价指标,分析不同时间间隔监测数据对预测精度的影响。结果表明,在总监测时长不变的情况下,预测精度随平均监测间隔时长的增长呈先增高后降低的趋势,即并非监测间隔越短,预测精度越高,而是在相应监测间隔范围内存在预测精度最优值。研究成果可为老采空区监测方案设计及沉降预测模式提供借鉴和指导。

In order to study the influence of the time interval of the old goaf subsidence monitoring data on the prediction accuracy,this paper uses the measured subsidence data of a certain old goaf surface subsidence monitoring points to establish two schemes and six subsidence prediction models with equal and unequal time intervals.Long short-term memory neural network(LSTM)prediction model is used to predict the surface subsidence of the old goaf.The mean absolute error(MAE)and the mean absolute percentage error(MAPE)are used as evaluation indicators to analyze the influence of time interval monitoring data on prediction accuracy.The results show that,under the condition of constant total monitoring duration,the prediction accuracy increases first and then decreases with the increase of the average monitoring interval,that is,it is not that the shorter the monitoring interval,the higher the prediction accuracy,but that there is an optimal value of prediction accuracy within the corresponding monitoring interval.The research results can provide reference and guidance for the monitoring scheme design and subsidence prediction model of the old goaf.