利用趋势化随机参数场的地下水流数值模拟优化方法

Optimized Groundwater Numerical Simulation Model with Trending Parameter Field

水文地质参数场的刻画是建立地下水流数值模拟模型的关键问题和难点问题。通常来讲,参数场合理性程度越高,模型拟合精度越高。本次研究将随机方法和参数空间分布表达进行结合,提出了趋势化随机参数场的构建方法。以渗透系数为研究对象,首先利用MCMC采样和样本数据特征确定水文地质参数的基本数据结构,进而根据样本空间分布特征对其进行趋势化处理,最终形成趋势化的渗透系数场。通过算例分析,利用趋势化处理后的渗透系数场能够大幅提高模拟精度,相比传统赋均值方法其误差可降至原来的1/3。在北京大兴跌隆起地区进行的实例应用说明,趋势化渗透系数场对提升岩性粒径较大(中砂以上)地区模拟精度效果显著,案例中粗砂区域渗透系数经趋势化处理后平均拟合误差由2.76 m下降至0.64 m;而对岩性以细砂及以下粒径为主的区域模拟精度提升并不明显。总体来说,该方法可为地下水流数值模型的优化提供借鉴,提升模型拟合精度,从而更加合理地刻画地下水流系统。

Hydrogeological parameter field is the key and challenging topic in groundwater numerical simulation. Generally, high-precision simulation results are derived from more reasonable parameter field. In this study, a trending stochastic parameter field construction method was proposed with stochastic and spatial distribution simulation technology. We took the hydraulic conductivity field as an example, and used the MCMC sampling method and sample feature analysis to extract data structure first. After that, the stochastic parameter field was reshaped with the trend features. For the case study, we compared the traditional parameter field, which filled blocks with the regional mean value. The trending parameter field has significantly improved the model accuracy with the mean error reduced by about 2 times. Practically, the mean simulation error comes down from 2.76 to 0.64 m in the coarse-grained area. In contrast, there is little effect in the low permeability area. To conclude, our method can provide reference for optimizing groundwater numerical simulation, improve the model fitting accuracy, and better describe the groundwater flow system.