基于多因素影响的BP-RBF神经网络渗流预测模型  

BP-RBF Neural Network Seepage Model Under the Influence of Various Factors

作　　者：李鹏犇 苏亮渊 贾亚杰[1] 孟弯弯 LI Pengben;SU Liangyuan;JIA Yajie;MENG Wanwan(College of Water Resources Science and Engineering,Taiyuan University of Technology,Taiyuan 030024,China)

机构地区：[1]太原理工大学水利科学与工程学院,山西太原030024

出　　处：《人民黄河》2018年第4期132-135,共4页Yellow River

基　　金：山西省国际合作项目(2013081034);2015年度山西省研究生教育创新项目(2015SY18)

摘　　要：为提高大坝坝基渗流的预测精度,把BP神经网络较强的模糊推理和自学能力与RBF神经网络在函数收敛中的快速性和绝对性相结合,以避免BP神经网络陷入局部最小或不收敛,构建了以水库大坝库水深、降雨量和温度三参数为主要影响因素,大坝渗流量为观测值的函数关系。通过与汾河水库实测资料对比分析表明,基于BP-RBF神经网络模型的坝基渗流预测模型预测效果良好,可以为大坝的安全监测与病险防护提供数据支持,为大坝原型观测资料处理提供了新途径。In order to improve the accuracy of dam safety prediction based on monitoring data,BP neural network,which has strong fuzzy reasoning and self-learning ability,and RBF neural network,which has fast and stable function convergence were combined to avoid BP neural network’s sinking into local minima and misconvergence.A functional relationship was established with reservoir level,rainfall and tem-perature as the three input parameters and the observed seepage value as the output parameter.Analysis of real data from the Fenhe Reservoir in Shanxi Province shows that the model based on BP-RBF has a high accuracy to predict dam foundation seepage.This model can be used to provide data support for dam safety monitoring and risk prevention and a new perspective to deal with observed real dam data.

关 键 词：大坝安全监测 渗流 预测 BP神经网络 RBF神经网络 汾河水库 

分 类 号：TV698.1[水利工程—水利水电工程]