摘要
提出一种基于经验模态分解(EMD)和遗传BP神经网络的大坝变形预测新算法。该算法首先通过EMD对变形序列进行分解,有效分离出非线性高频波动分量和低频趋势分量;然后应用遗传算法优化BP神经网络的权值和阈值,再对各分量进行建模预测;最后叠加各分量预测值得到预测结果。应用新算法与灰色GM(1,1)、回归模型、普通卡尔滤波和遗传BP神经网络算法进行对比分析。结果表明,该算法具有较强的自身内部环境优化和外部平台构建能力,自适应能力和非线性拟合能力较强,在一定程度上保证较优的局部预测值和较好的全局预测精度,在大坝变形预测中具有一定的实用价值。
A new algorithm based on EMD and genetic algorithm-BP neural network is proposed.First,to effec-tively separate the nonlinear trend of volatility of high frequency and low frequency components,the algorithm deformation sequence is decomposed by EMD.Then,genetic algorithm is used to optimize weights and thresh-old values of the BP neural network,to build a prediction model for each component.Finally,the predicted values of each component in the forecast is overlay.The calculation is analyzed and compared with grey GM （1 , 1 ）,regression analysis,common Carl filtering and GA-BP neural network.The results show that the method can build external and internal environment optimization platform.With generalization ability and an adaptive fitting,it ensures the optimal local prediction with higher precision forecasting,and can be applied to dam de-formation prediction practically.
出处
《桂林理工大学学报》
CAS
北大核心
2015年第1期111-116,共6页
Journal of Guilin University of Technology
基金
国家自然科学基金项目(41071294)
广西"八桂学者"专项经费项目
广西空间信息与测绘重点实验室项目(桂科能1207115-06)
广西矿冶与环境科学实验中心项目(KH2012ZD0004)
关键词
大坝变形
经验模态分解
遗传算法
BP神经网络
精度评定
dam deformation
empirical mode decomposition
genetic algorithm
BP neural network
precision evaluation
