相似洪水分析与智能推荐模型研究  

Research on Similar Flood Analysis and Intelligent Recommendation Modeling

作　　者：田逸飞 杨海从 纪国良 王乐 严方家 马振亮 张潇 

机构地区：[1]长江水利委员会水文局,湖北 武汉 [2]长江水利委员会水旱灾害防御创新团队,湖北 武汉 [3]流域水模拟与预报调度智能技术创新中心,湖北 武汉 [4]汉江水利水电(集团)有限责任公司,湖北 武汉 [5]中国长江三峡集团有限公司,湖北 武汉 [6]长江大学资源与环境学院油气地球化学与环境湖北省重点实验室,湖北 武汉

出　　处：《水资源研究》2023年第5期472-485,共14页Journal of Water Resources Research

摘　　要：本文提出了基于流域历史资料的相似洪水智能推荐模型,根据降雨数值预报过程对历史典型洪水进行相似洪水搜索达到洪水预报的效果。首先,选取暴雨洪水特征指标并采用主成分分析方法对特征指标降维,然后采用K-means聚类算法对历史典型场次暴雨洪水进行分类,构建雨洪聚类推荐模型,最终实现相似洪水智能推荐。在嘉陵江流域的应用显示,本研究构建的智能推荐模型可以匹配到相似的历史洪水过程,为洪水预报业务提供参考。This paper proposes a similar flood intelligent recommendation model based on historical data of the ba-sin, and achieves the effect of flood forecasting by searching for similar floods according to the numerical precipitation forecasting process of historical typical floods. First, the characteristic indicators of rainstorm floods are selected and dimensionality reduction is performed by principal component analysis method. Then, K-means clustering algorithm is used to classify the historical typical cases of rainstorm floods to construct the rainstorm flood clustering recommendation model, and finally the similar flood intelligent recommendation is realized. The application in the Jialing River basin shows that this method can match similar historical flood processes through the constructed intelligent recommendation model, and provide reference for the operational application of flood forecasting.

关 键 词：洪水预报 洪水分类 主成分分析 聚类分析 嘉陵江流域 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]