基于集合经验模态分解与支持向量机回归的位移预测方法:以三峡库区滑坡为例

发布时间：2019-02-20 18:00

【摘要】：三峡库区滑坡地表位移-时间曲线多呈台阶型特征。基于位移响应成分模型的滑坡位移预测方法是该类滑坡位移预测的主要方法之一。针对目前水库滑坡波动项位移预测工作中尚未考虑主要诱发因素的高频成分与低频成分的问题,提出了基于时间序列集合经验模态分解(EEMD)与重构的粒子群优化-支持向量机回归(PSO-SVR)位移预测方法。以白水河滑坡2003年7月至2013年3月117个地表位移数据为例,采用EEMD法将位移时间序列分解为趋势项位移和波动项位移,该趋势项位移用最小二乘法的二次多项式拟合预测;根据EEMD和t检验法,确定高频降雨量、低频降雨量、高频库水位和低频库水位,结合其他常用因素,采用灰色关联分析确定白水河滑坡影响波动项位移的优势因素为高频降雨量和月间库水位变化,基于优势因素建立PSO-SVR模型预测波动项位移。结果表明,总预测值的平均相对误差为0.009 8,方差比为0.023 9,小误差概率为1,预测效果较好。利用该方法对三峡库区其他5个台阶型滑坡进行了预测,预测位移与实测位移较吻合,进一步证明了该方法的有效性,对同类滑坡的预测预报具有一定的借鉴意义。
[Abstract]:The surface displacement-time curves of landslide in the three Gorges Reservoir area are mostly of step type. Landslide displacement prediction method based on displacement response component model is one of the main methods for landslide displacement prediction. Aiming at the problem that the high frequency and low frequency components of the main induced factors have not been considered in the displacement prediction of the landslide fluctuation term in the reservoir at present, A particle swarm optimization-support vector machine (PSO-SVR) displacement prediction method based on empirical mode decomposition (EEMD) and reconstruction of time series is proposed. Taking 117 surface displacement data of Baishuihe landslide from July 2003 to March 2013 as an example, the displacement time series is decomposed into trend term displacement and fluctuation term displacement by EEMD method. The trend term displacement is predicted by quadratic polynomial fitting of least square method. According to EEMD and t test method, high frequency rainfall, low frequency rainfall, high frequency reservoir water level and low frequency reservoir water level are determined, combined with other commonly used factors, Grey correlation analysis is used to determine that the dominant factors for the displacement of the fluctuation term of the Baishui River landslide are high frequency rainfall and monthly water level change. Based on the dominant factors, the PSO-SVR model is established to predict the fluctuation term displacement. The results show that the average relative error of the total prediction value is 0.009 8, the variance ratio is 0.023 9, the probability of small error is 1, and the prediction effect is better. The method is used to predict the other five step landslides in the three Gorges Reservoir area. The predicted displacement is in good agreement with the measured displacement, which further proves the validity of this method and has some reference significance for the prediction of similar landslides.
【作者单位】：中国地质大学(武汉)工程学院;
【基金】：国家自然科学基金项目(No.41372310) 中国地质大学(武汉)中央高校基本科研业务费专项资金资助项目(No.1610491T07)~~
【分类号】：P642.22

【相似文献】