文献类型：期刊文献

中文题名：基于MI-KPCA与高斯回归过程的北汝河中长期径流预测

英文题名：Medium and Long Term Runoff Forecast Based on MI-KPCA and GPR in Beiru River

作者：周靖楠 刘振男 陆之洋 焦卫国

第一作者：周靖楠

机构：[1]贵州理工学院理学院,贵州贵阳550003;[2]贵州大学电气工程学院,贵州贵阳550025

第一机构：贵州理工学院理学院

年份：2021

卷号：39

期号：5

起止页码：42-45

中文期刊名：水电能源科学

外文期刊名：Water Resources and Power

收录：CSTPCD;;北大核心:【北大核心2020】；

基金：黔科合基础-ZK[2021]一般295。

语种：中文

中文关键词：互信息;核主成分分析;高斯回归过程;径流预测

外文关键词：MI;KPCA;GPR;runoff forecasting

摘要：针对径流预测因子筛选常用的线性分析方法无法识别与径流存在非线性关系因子的局限性,基于互信息与核主成分分析,提出一种能同时遴选与径流存在线性与非线性关系的因子识别方法,来筛选径流预测因子,且将高斯回归过程拓展应用于北汝河的中长期径流预测。结果表明,构建的MI-KPCA因子筛选质量优于相关分析法与互信息法,不但能提高预测精度,还能简化模型结构、缩短运行时间;同时,高斯回归过程适用于径流预测、可作为其他水文预测模型推广使用。
The common linear analysis method for selecting runoff predictors is unable to identify the nonlinear factors associated with runoff.Based on the mutual information and kernel principal component analysis,a factor identification method which can simultaneously select runoff with linear and nonlinear relationship was proposed to select runoff predictors.And then Gaussian regression process was extended and applied to the mid-and long-term runoff prediction of Beiru river.The results show that the quality of MI-KPCA is better than that of correlation analysis and mutual information method,which can not only improve the prediction accuracy,but also simplify the model structure and shorten the operation time.Meanwhile,the Gaussian regression process is suitable for runoff prediction and can be used as other hydrological prediction models.