基于聚类分析的模糊马尔科夫链在降雨量预测中的应用  

Rainfall Prediction Using Clustering-fuzzy-markov Chain Model

作　　者：宋帆 杨晓华[1] 武翡翡 刘童 SONG Fan;YANG Xiao-hua;WU Fei-fei;LIU Tong(State Key Laboratory of Water Environment Simulation,School of Environment,Beijing Normal University,Beijing 100875,China)

机构地区：[1]北京师范大学环境学院水环境模拟国家重点实验室,北京100875

出　　处：《节水灌溉》2018年第10期33-36,41,共5页Water Saving Irrigation

基　　金：国家重点研发计划项目(2017YFC0506603;2016YFC0401305);国家自然科学基金重点项目(41530635);国家自然科学基金面上项目(51379013;51679007)

摘　　要：针对降雨量序列的复杂性和随机性,基于马尔科夫链原理,采用聚类分析对降雨量序列进行分类,引入隶属度对样本状态向量进行测算。建立了聚类-模糊马尔科夫降雨量预测模型,并对结果进行了改进。采用全国各地共16个站点的2011-2013年48个降雨量数据作为待测样本进行计算,结果表明:48个预测样本的平均绝对误差为12.4%,误差低于10%的年份占56.25%。精度较高,将模型用于降雨量的预测是合理的,可以为水资源合理规划利用提供依据。For the complexity and randomness of rainfall sequence, based on Markov chain principle, clustering analysis was used to classify the rainfall sequence, and the degree of membership was used to measure the state vector of samples. The rainfall prediction model of clustering-fuzzy-markov chain model was established and the results are improved. A total of 48 years of rainfall data from 16 sites across the country were calculated as predicted samples. The results showed that the mean error of the 48 predicted samples was 12.4%. The year with the error less than 10% accounts for 56.25% of the total year, and the accuracy is higher. It is reasonable to apply the model to rainfall prediction. It can provide basis for rational planning and utilization of water resources.

关 键 词：降雨量预测 马尔科夫链 聚类分析 隶属度 

分 类 号：P333.9[天文地球—水文科学]