基于主成分分析和模糊聚类的水文分区  

Hydrologic regionalization based on principal component analysis and fuzzy clustering

作　　者：丁亚明[1] 赵艳平[1] 张志红[1] 胡学钢[2] 

机构地区：[1]安徽水利水电职业技术学院电子信息工程系,安徽合肥231603 [2]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《合肥工业大学学报（自然科学版）》2009年第6期796-801,共6页Journal of Hefei University of Technology：Natural Science

基　　金：安徽省自然科学基金资助项目(050420207)

摘　　要：水文分区有各种方法,文章提出了集模糊聚类与主成分分析方法的各优点组合的水文分区方法,首先采用主成分分析法获得水文特性主成分属性,然后运用模糊聚类算法NFC(Net Fuzzy Cluster)进行模糊聚类。利用主成分分析法对分区指标进行降维处理,简化了计算。应用NFC模糊聚类,在一定程度上解决了FCM算法局部极值问题且具有良好的聚类性能,实现了聚类的科学化与自动化。对安徽省淮河流域的124716个原始水文数据进行实验,结果表明,与传统分区方法相比,所提出的方法有效地改善了时间性能,提高了求解精度,所得结果为水文站网规划提供了理论依据。Based on the analysis of various methods of hydrologic regionalization, this paper presents a new hydrologic regionalization method, which integrates the advantages of both fuzzy clustering and principal component analysis（PCA）. Firstly, the principal components analysis is made to select relevant attributes of hydrological features, and then the Net Fuzzy Cluster（NFC） algorithm is used in fuzzy clustering. The regionalization indicators are processed by PCA to reduce the dimensions. By applying of the NFC algorithm, the method solves the problem of local minima in the FCM algorithm to some extent, brings about the perfect performance of fuzzy clustering, and realizes scientific and automatic clustering. Experiments were done on collected 124716 original hydrologic data of the Huaihe river in Anhui Province, and the results show that compared with traditional methods, the presented method improves the time performance and increases the solving precision effectively. The obtained results are helpful for planning of hydrologic networks.

关 键 词：主成分分析 NFC模糊聚类 水文分区 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] P336[自动化与计算机技术—控制科学与工程]