基于IPSO-BP神经网络的坝基扬压力预测方法研究  

Study of Prediction Method of Dam Foundation Uplift Pressure Based on Improved Particle Swarm Optimization-BP Neural Network

作　　者：顾浩钦[1] 仲云飞[1] 程井[1] 邓同春[1] 李阳[1] 

机构地区：[1]河海大学水利水电学院,南京210098

出　　处：《三峡大学学报（自然科学版）》2013年第2期20-24,共5页Journal of China Three Gorges University：Natural Sciences

基　　金：国家自然科学基金青年项目(51009056)

摘　　要：针对坝基扬压力预测的传统BP神经网络模型初始权值和阈值随机性强、易陷入局部最优等局限,采用惯性权重动态调整的改进粒子群算法对BP网络的初始权值和阈值进行优化,建立了基于IPSO的BP神经网络坝基扬压力预测模型.通过算例验证算法的优越性及程序的准确性,并以某大坝多年扬压力监测数据进行工程实例应用,结果表明,IPSO-BP扬压力预测模型与传统BP模型相比,拟合相关系数大,统计误差小,预测精度更高.Initialized weights and thresholds of the traditional BP neural network alogorithm in prediction of dam foundation uplift pressure are random; and it＇s easily to converge to local optimum. According to this characteristics, particle swarm optimization（PSO） based on dynamic regulation of inertia weight,which has a strong capability of global searching,is utilized to optimize the initialized weights and thresholds of the BP neural network. The prediction model of dam foundation uplift pressure of BP neural network based on im- proved particle swarm optimization（IPSO） is established. The advantage and accuracy of this algorithm is ver- ified by a case study. And the years of uplift pressure monitoring data of a dam foundation is used for evalua- ting the IPSO-BP neural network model. The results show that compared with the traditional BP neural net- work, the prediction of dam foundation uplift pressure model based on IPSO-BP neural network has higher co- efficient correlation, smaller statistical error and better prediction accuracy.

关 键 词：扬压力 BP神经网络 改进粒子群算法 统计模型 

分 类 号：TV698.1[水利工程—水利水电工程]