基于GAMLSS-GLO及随机森林算法的土石坝渗流监测模型

根据土石坝渗流原型观测，厘清水位、降水等因素的影响，模拟大坝渗流的真实状态，合理评估其渗流监测结果，是土石坝安全监控亟待解决的关键问题。基于此，根据数理统计原理，采用随机森林算法构建无降水条件下渗流量与上下游水位的回归模型；考虑渗流量受前期累积降水的综合影响，引入广义可加模型（GAMLSS-GLO），模拟降水影响下土石坝渗流监测值的波动区间，并将其与渗流-水位回归模型叠加，预测土石坝渗流监测的可靠区间；最后，将该方法应用于糯扎渡心墙堆石坝的渗流监测。结果表明：所提模型方法对渗流的水位、降水响应表现出良好的适用性，显著提高了渗流模拟预测质量。同时求解了渗流量置信区间，有利于土石坝的运行工况判断及安全监控。

Seepage monitoring model of earth-rock dam based on GAMLSS-GLO and random forest algorithm

The key problems of the seepage flow monitoring of earth-rock dams are how to clarify the influence of water level, rainfall and other factors, how to simulate the real seepage flow and evaluate the results of the seepage monitoring based on the prototype observation. In view of this, a regression model of seepage flow and upstream and downstream water levels is established under the condition of no rainfall using the random forest algorithm and the principles of mathematical statistics. Regarding to the comprehensive influence of the seepage flow on the accumulated rainfall in the earlier period, the Generalized Additive Models for Location, Scale and Shape of Generalized Logistic Distribution (GAMLSS-GLO) is introduced to simulate the fluctuation intervals of the monitoring values of seepage flow under the influence of rainfall. Then, GAMLSS-GLO is superimposed with the seepage flow-water level regression model to predict the reliable intervals of the seepage flow. Finally, the method is applied to the seepage monitoring of the core wall earth-rock dam in Nuozhadu Reservoir. The results show that the proposed method shows good applicability to the water level and rainfall response of seepage flow, it can significantly improve the quality of seepage simulation prediction, and solve the confidence intervals of seepage flow, which is conducive to judging the operation condition and safety monitoring of earth-rock dams.