首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于改进FCM算法的加权马尔可夫链的年降水预测
引用本文:苗正伟,徐利岗.基于改进FCM算法的加权马尔可夫链的年降水预测[J].灌溉排水学报,2017,36(10).
作者姓名:苗正伟  徐利岗
作者单位:1. 河北水利电力学院 水利工程系,河北 沧州,061001;2. 宁夏水利科学研究院,银川,750021
基金项目:河北省教育厅青年基金项目
摘    要:应用改进的FCM(Fuzzy C-Means,模糊C均值聚类)算法对承德市1951—2015年的年降水序列进行模糊聚类,获得10个聚类中心和隶属度矩阵。根据最大隶属原则确定每年的降水状态,采用规范化的各阶自相关系数为权重,建立了加权马尔可夫链模型。以隶属度向量作为预测时的初始状态向量,通过该模型逐年预测了承德市2004—2015年的降水状态,结果与实际情况一致。引入模糊集中的级别特征值公式,并对该公式做出修正。基于马尔可夫链的预测结果,应用修正后的级别特征值公式预测了2004—2015年的降水量,所有预测结果的相对误差都在7%以内,初步表明基于模糊聚类和加权马尔可夫链的降水预测模型是合理可行的。

关 键 词:预测  降水  模糊聚类  FCM算法  加权马尔可夫链  承德

Predicting Annual Precipitation Using the Weighted Markov Chain Solved by the Improved FCM Algorithm
MIAO Zhengwei,XU Ligang.Predicting Annual Precipitation Using the Weighted Markov Chain Solved by the Improved FCM Algorithm[J].Journal of Irrigation and Drainage,2017,36(10).
Authors:MIAO Zhengwei  XU Ligang
Abstract:We predicted the annual precipitation from 1951 to 2015 in Chengde using the improved FCM algo-rithm.The precipitation series was firstly classified using the fuzzy clustering, from which 10 cluster centers and their associated membership matrix were obtained. The weighted Markov chain model was established based on the maximum membership principle and the calculated annual precipitation series, by using the standardized auto-correlation coefficients as the weights. We used the membership vector as the initial state vector and predicted the annual precipitation from 2004 to 2015 in Chengde;the calculated results agree well with the measurements. We introduced and modified the level characteristic value formula of the fuzzy set. Based on the predicted results from the Markov chain model, we predicted the annual precipitation from 2004 to 2015 using the modified level characteristics value formula. The relative error was less than 7%. The preliminary results show that theproposed model is reliable for predicting annual precipitation.
Keywords:prediction  annualprecipitation  fuzzy clustering  FCM algorithm  weighted Markov chain  Chengde
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号