Abstract

During the operation period of the high arch dam in some canyon area, the valley deformation is obvious under complex geological environment. Several possible factors may influence this phenomenon. In this article, influencing factors are considered from the aspects of precipitation, temperature, reservoir water level elevation and the rate of reservoir water level variation. Two machine learning methods are employed to investigate these factors, namely, Lasso and Random forest, and serve as a basis for further studying the formation mechanism of valley deformation. Using the Lasso method, four variations are selected from 74 variations which are representative of precipitation and temperature, e.g. daily precipitation, 10-day antecedent precipitation, 10-day maximum rainfall difference, and 25-day maximum temperature difference. Combined with the reservoir water level elevation and the rate of reservoir water level variation, these six key factors are analyzed by the Random forest method so to carry out a quantitative analysis of their influence on valley deformation. Results show that the valley deformation is mainly affected by the rate of reservoir water level variation and the reservoir water level elevation among the selected factors, indicating valley deformation might be induced with hydrodynamic force.

摘要

某峡谷区高拱坝运营期，地质环境复杂，河谷变形明显。几个可能的因素可能会影响这种现象。本文从降水量、温度、水库水位高程和水库水位变化率等方面考虑影响因素。采用两种机器学习方法来研究这些因素，即Lasso和随机森林，并作为进一步研究山谷变形形成机制的基础。利用Lasso方法，从74个代表降水量和温度的变量中选出4个变量，如日降水量、前期10天降水量、10天最大降水差、25天最大温差。结合水库水位高程和水库水位变化率，采用随机森林法对这6个关键因素进行分析，定量分析其对河谷变形的影响。结果表明，在所选因素中，河谷变形主要受水库水位变化率和水库水位高程的影响，表明河谷变形可能是由水动力引起的。