BP神经网络模型在库区高边坡变形预测中的应用  

Application of BP Neural Network in Deformation Predication of High Slope in Reservoir Area

作　　者：范新宇[1,2] 杨天俊[1,2] 高凯[1] 贾志献[1] FAN Xinyu YANG Tianjun GAO Kai JIA Zhixian(Northwest Engineering Corporation Limited, Xi＇an 710065, China High Slope and Geological Hazard Research Treatment Division of China Hydropower Technology Research and Development Center, Xi＇an 710065 ,China)

机构地区：[1]中国电建集团西北勘测设计研究院有限公司,西安710065 [2]国家能源水电工程技术研发中心高边坡与地质灾害研究治理分中心,西安710065

出　　处：《西北水电》2017年第1期72-74,共3页Northwest Hydropower

摘　　要：影响因素的多变及周期特征使高边坡变形表现出明显的非线性,线性模型难以对边坡变形做出准确预测。为研究库区高边坡变形发展趋势,以某水电站高边坡变形监测成果为依托,在对其变形影响因素分析基础上,选取半年库水位变化量、半年降雨量、前半年惯性位移量为输入参数,以监测点半年变形增量值为输出参数,构建起3层BP神经网络模型。利用该模型对2011年6月—2014年12月间监测数据进行训练,精度达到预设要求后,预测2015年1月—2016年6月测点变形值,结果表明:预测最大误差8.26%,平均误差7.2%,满足工程精度要求,说明该模型参数选取及设置适合,可为后续边坡变形趋势研究提供参考。Variation and periodic features of the impact factors have the high slope deformation present the obvious non-linear performance. The linear model can hardly and accurately predict the slope deformation. To study the development tendency of the high slope deformation in reservoir area and based on the monitoring data of the high slope deformation of one hydropower station, the input parameters such as the water level variation of a half year, precipitation of a half year and the inertia deformation of the first half year as well as the output parameters such as the increased deformation value at the monitored point in a half year are taken to build the BP neural network in three layer on the basis of the analysis on the impact factors to the deformation. By application of the model, the monitoring data from June 2011 to December 2014 are excised. When the precision reaches the pre-designed one, the deformation at the monitoring point from January 2015 to June 2016 is predicted. The prediction results show that the maximum prediction error is 8.26% and the average error is 7.2%. Both of them satisfy requirements on the engineering precision. These proves that the parameter selection and arrangement of the model is suitable. This can provides the sequent study of the slope deformation tendency with reference.

关 键 词：库区 高边坡 影响因素 变形预测 BP神经网络 

分 类 号：P642.22[天文地球—工程地质学] TP183[天文地球—地质矿产勘探]