一种新型耦合模型在月径流预测中的应用  

Application of New Hybrid Model in the Expended Prediction of Monthly Runoff

作　　者：朱聪[1] 王文圣[2] 程开宇[1] 杨娟[1] 

机构地区：[1]中国水电顾问集团华东勘测设计研究院,浙江杭州310014 [2]四川大学水利水电学院,四川成都610065

出　　处：《水力发电》2012年第1期14-16,共3页Water Power

基　　金：国家自然科学基金项目(50739002);科技基础性工作专项(2009IM020100);教育部新世纪优秀人才支持计划资助项目(2009SCU-NCET-09-0573)

摘　　要：基于投影寻踪分析与随机分析提出了一种新的耦合预测模型,运用投影寻踪技术将年内12个月径流由遗传算法优化得投影值,获取投影值与年径流的相关关系;建立年径流预测模型,由预测的年径流推算对应的投影值z^;寻找与z^最近邻的h个模式,由最近邻回归进行年内月径流展望预测。将耦合模型应用于宝珠寺和三峡水电站入库月径流展望预测,结果表明该耦合模型可行且预测效果较好。A kind of new hybrid prediction model is advanced based on projection pursuit analr＇sis and stochastic analysis. By using this model, the projection values are firstly calculated through optimizing monthly runoff in every year by genetic algorithm and the regressive function of annual runoff and projection values is set up, and then a corresponding projection value z is calculated based on the annual runoff predicted with stochastic model, and finally some modes which are nearest from 2 are selected and the monthly runoff of the next year will be predicted. The model is used in the prediction of monthly inflow of Baozhusi Reservoir and Three Gorges Reservoir, and the results show that the model is feasible and the prediction is satisfied.

关 键 词：投影寻踪 最近邻抽样回归 月径流预测 

分 类 号：P338.2[天文地球—水文科学]