基于经验模式分解的边坡位移预测模型研究  

Study of slope displacement prediction model based on empirical mode decomposition

作　　者：张永磊[1] 何秀凤[1] 刘志平[2] 

机构地区：[1]河海大学卫星及空间信息应用研究所,江苏南京210098 [2]中国矿业大学国土灾害与环境监测国家测绘地理信息局重点实验室,江苏徐州221116

出　　处：《人民长江》2013年第3期48-51,92,共5页Yangtze River

基　　金：国家自然科学基金项目(40974001)

摘　　要：高边坡监测数据分析是研究边坡变形演化机制及稳定性评价的主要方法,是工程施工决策的基础。根据边坡位移时间序列呈现趋势性与随机性相耦合的特点,运用经验模式分解方法将边坡位移量分解为趋势项和随机项部分,对趋势项采用支持向量机回归模型预测,并基于马尔可夫状态切换对随机项建立自回归模型进行预测。以小湾水电站高边坡为例,将该模型应用到边坡位移变形预报中,取得了较好的效果。High slope monitoring data analysis is a main method to study the deformation mechanism of a slope and evaluate its stability, laying foundation for decision - making during construction. According to the characteristics that the tendency and ran-domness of the slope displacement time series have obvious coupling effect, we decompose the displacement series into trend item and random item by empirical mode decomposition （EMD）. The trend item is predicted by support vector regression （SVR） , and the random item is predicted by autoregressive model based on Markov regime switching. The combined model is applied to predict the displacement of a slope in Xiaowan Hydropower Station, and the calculated results are accurate.

关 键 词：边坡位移 经验模式分解 支持向量机回归 马尔可夫状态切换 自回归模型 

分 类 号：P642[天文地球—工程地质学]