摘要
马尔可夫模型是一种水质定性预测的常用方法.本文将模糊集理论中级别特征值的概念引入马尔可夫模型,提出一种基于投影距离的方法(M2),使之能够用于水质的定量预测.利用1992到2004年黄河干流潼关、三门峡两断面BOD5、氨氮、溶解氧等三项水质指标的历史序列推断2005年的水质,发现除潼关的BOD5误差较大(23.3%)以外,其它水质指标的预测值误差均在1.5%到5.6%之间.和传统的处理级别特征值的方法(M1)相比,M2方法具有较高的精度.利用该模型计算两断面2005—2010年的水质变化,发现水质趋于改善.
The Markov model is a method widely used in the qualitative prediction of water quality.This paper introduced the concept of Grade Characteristic Value（GCV）from the fuzzy-set theory into the Markov model,and proposed a projective-distance-based method（M2）to enable Markov model to quantitatively predict water quality.With this method,data series of BOD5,ammonia nitrogen,and dissolved oxygen from 1992 to 2004 were used to deduce water quality in 2005 at two sections traversing the Yellow River mainstream（Tongguan and Sanmenxia）.The errors are all within 1.5%—5.6% except the BOD5 at Tongguan（23.3%）.Compared with a conventional method（M1）also based on GCV,M2 can produce results with higher precision.Prediction from 2005 to 2019 showed that water quality at both sections tends to improve.
出处
《应用基础与工程科学学报》
EI
CSCD
2011年第2期231-242,共12页
Journal of Basic Science and Engineering
基金
国家自然科学基金项目(50979003)
关键词
水质定量预测
马尔可夫链
级别特征值
quantitative prediction of water quality
Markov chain
grade characteristic value
