基于时间序列模型的隧道涌水量反演与预测
Inversion and predication of inflow in tunnel based on time-series model
-
摘要: 鉴于使用确定性模型预测隧道涌水量时存在难以准确获取水文地质参数等诸多不便,本文将地下水系统视为"黑箱"模型,通过提取隧道涌水量历史观测数据本身蕴涵的趋势、周期和随机变化规律,建立了隧道涌水的时间序列预测模型。经用于铜锣山隧道实例,反演系列的平均绝对误差为14.67%,预测序列的平均绝对误差为14.34%,表明模型的整体预测精度较高。Abstract: Considering inconvenience of obtaining accurate hydrogelogical parameters when using determined model to predict the inflow of tunnel,this paper takes the groundwater system as a "black box" and establishes a time-series model to forecast the inflow in tunnel,based on analyzing and extracting the trend component,the periodic component and the stochastic component of the historical observed data.The method is applied to the Tongluoshan tunnel.The Mean Absolute Percent Error(MAPE) of the inversion series is 14.67%,while the MAPE of the prediction series is 14.34%.The results show that this model has a high prediction accuracy.