基于K-means的模糊马尔科夫模型的降水预测  

Precipitation prediction of fuzzy⁃Markov chain model based on K⁃means algorithm

作　　者：董亚 杜景林[1] 胡玉杰 DONG Ya;DU Jinglin;HU Yujie(School of Electronic&Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,江苏南京210044

出　　处：《现代电子技术》2021年第19期85-89,共5页Modern Electronics Technique

基　　金：国家自然科学基金项目(41575155);2018年江苏省研究生实践创新计划项目(1344051901038)。

摘　　要：针对中长期降水预测精度低的问题,基于马尔科夫原理,应用K-means算法对降水序列进行聚类,引入权重和模糊理论,建立K-means的模糊马尔科夫链模型,并对结果进行分析。采用南京地区1951—2014年的降水序列作为样本进行模型训练,对包括南京市在内的9个地区的2015—2017年的降水进行预测,结果表明,7个年平均降水量低于1 000 mm的地区中有6个地区的平均误差低于10%,另外27个预测样本中有14个误差低于10%,精度较高,可以将模型应用于降水量低于1 000 mm地区的降水预测,为当地水资源的合理使用提供依据。On the basis of the Markov principle,the precipitation sequence is clustered by K⁃means algorithm to improve the precision in medium⁃and long⁃term precipitation prediction.The weight and fuzzy theory are introduced to establish a K⁃means⁃fuzzy⁃Markov chain model.The prediction results of the model are analyzed.The precipitation sequence from 1951 to 2014 in Nanjing area is taken as the samples.The precipitation from 2015 to 2017 in nine areas including Nanjing is predicted.The results show that the average errors of 6 regions in the 7 regions with an annual mean rainfall<1000 mm are less than 10%,and the errors of 14 predicted samples in the 27 predicted samples are less than 10%.It can be seen that the model has high precision,and can be applied to the precipitation prediction in the areas with precipitation less than 1000 mm.Therefore,the proposed model can provide a basis for the rational use of local water resources.

关 键 词：降水预测 马尔科夫链 K-MEANS算法 降水序列 聚类 预测模型 模型训练 

分 类 号：TN911.1-34[电子电信—通信与信息系统] TP301.6[电子电信—信息与通信工程]