摘要
针对传统BP神经网络存在收敛速度慢、难以脱离局部极小值的不足,利用改进粒子群算法(PSO)快速的收敛特性和强大的全局搜索能力,将改进的粒子群算法与BP神经网络结合起来,根据海堤特点分类比较渗压的影响因素,采用相关系数法选取主要影响因子构建模型输入层,对应渗压作为模型输出层,建立海堤渗压改进PSO-BP监测模型。采用浦东海堤实测信息作为实例进行分析,结果显示,与BP模型相比,改进PSO-BP模型在海堤渗压监测应用中具有更高的收敛速度和更强的预测能力,能更有效地揭示海堤渗压的变化规律。
On account of the traditional BP neural network is easy to fall into local minimum and has a slow convergence speed problem.This paper intends to use the rapid convergence property and strong global search ability of the modified Particle Swarm Optimization（PSO）,then combine the modified PSO with BP neural network.Classifying and comparing the influencing factors of seepage pressure on the basis of the characteristics of seawalls.Adopting correlation coefficient method for selecting the main influence factors which will be imported as a model,and the seepage pressure is used as the outputs of the model,then the modified PSO-BP monitor model of seawall seepage pressure is established.The analysis of the examples about the measured data from Pudong seawall indicates that in the application of seawall seepage pressure monitoring,by contrast with BP neural network monitor model,the modified PSO-BP monitor model has higher convergence speed and higher prediction accuracy.Besides it can reveal the change regulation of the seawall seepage pressure more effectively.
出处
《中国农村水利水电》
北大核心
2017年第1期148-151,共4页
China Rural Water and Hydropower
基金
国家自然科学基金资助项目(50979056)
水利部公益性行业专项经费资助项目(201401063-02)
三峡库区地质灾害教育部重点实验室(三峡大学)开放研究基金(2015KDZ03)
安徽省科技攻关计划项目(1604a0802106)
关键词
海堤渗压
改进粒子群算法
神经网络
因子选择
监测模型
seawall seepage pressure
modified particle swarm optimization
neural network
select impact factors
monitor model
