摘要
针对BP神经网络的初始化权值和阈值的随机性,易导致训练速度慢和落入局部极小等弱点,本文运用具有并行特性和全局优化能力的粒子群算法(PSO)对BP神经网络的权值和阈值进行优化,建立了基于粒子群-BP神经网络的大坝变形监测模型,并以丰满大坝多年监测的坝顶水平位移资料为例进行实证分析。与经典BP神经网络模型的预测结果相比,粒子群-BP神经网络模型的收敛速度更快、预测精度更高。
Initialized weights and thresholds of the BP neural network are random,which results in slow convergence and easily converging to local optima.According to these characteristics,Particle Swarm Optimization(PSO),which has a strong global searching ability,was utilized to optimize the weights and thresholds of the BP neural network in the paper.The model of dam displacement prediction based on PSO-BP neutral network was established and the transverse displacement monitoring data of Fengman Dam was used for evaluating the model.The experimental results showed that the PSO-BP neutral network model was faster in training and more accurate in prediction than the classic BP neural network mode.
出处
《测绘科学》
CSCD
北大核心
2012年第4期181-183,共3页
Science of Surveying and Mapping
关键词
变形监测
粒子群
BP神经网络
displacement prediction
particle swarm optimization
BP neutral network