摘要
水文时间序列预测对于水文水利决策有着重要的意义。鉴于水文时间序列的复杂性,提出了一种水文时间序列的混合核PSO-KELM预测模型:将极限学习机(extreme learning machine,ELM)模型应用于水文时序预测研究,基于多核学习思想,构造由径向基核函数和多项式核函数加权构成的混合核函数,其综合了径向基核函数和多项式核函数的优点,并通过粒子群算法(particle swarm optimization,PSO)对模型的参数进行寻优,避免了人工操作造成的繁琐性和主观性。兰州站年径流量和金沟河流域年径流量实测数据被用来验证新模型合理性。通过两个算例表明:新模型能够获取比BP模型、RBF模型更好的结果。
Hydrological time series forecasting is of great significance for hydrological and water conservancy decision-making. In view of the complexity of hydrological time series,the mixed kernel PSO-KELM models a prediction of hydrological time series： extreme learning machine（ ELM） method is used to study the prediction of hydrological time series,and in accordance with the idea of multiple kernel learning,the mixed kernel function is constructed by radial basis kernel function and polynomial kernel function weighted by a new mixed kernel function,the comprehensive advantages of RBF kernel function and polynomial kernel function,and the particle swarm algorithm（ PSO） to optimize the parameters of the model,avoiding the tedious manual operation. The annual runoff of Lan Zhou Railway Station and the measured data of annual runoff of Jingou River Basin are used to verify the rationality of the new model. Two examples show that the new model can obtain better results than the BP model and the RBF model.
作者
涂异
汪金能
朱曲平
安雪玮
梅艺
陈东祖
TU Yi;WANG Jin-neng;ZHU Qu-ping;AN Xue-wei;MEI Yi;CHEN Dong-zu(College of civil engineering and architecture,Chongqing Institute of Engineering,Chongqing,40000)
出处
《中国农村水利水电》
北大核心
2018年第7期21-24,共4页
China Rural Water and Hydropower
基金
国家自然科学基金面上项目(51279219)
重庆市教委科学技术研究项目(KJ1601005)
关键词
水文时间序列
极限学习机
粒子群算法
混合核函数
hydrological time series
extreme learning machine
particle swarm optimization algorithm
hybrid kernel function
