【作者】 吴双双；

【导师】 裴洪平；

【作者基本信息】 浙江大学 ， 环境科学， 2006， 硕士

【摘要】 本文将人工神经网络原理运用于水质评价问题中,建立了多指标水质综合评价的Hopfield网络模型（HNN）,以探索西湖水体富营养化状况及随时空分布的规律。 本研究在杭州西湖5个子湖区共设8个采样点,收集了2000年12个月的西湖常规监测的水生生态数据。根据《中国水资源公报编制大纲》中的《湖泊、水库富营养化评分与分类方法》,选取叶绿素a（Chl-a）、总磷（TP）、总氮（TN）、化学耗氧量(COD_Mn)、透明度（SD）等五个水质因子,将评价标准分为贫营养、中营养、富营养三大类,细分为十类,构建包含50个神经元的Hopfield网络。将十类富营养化标准模式作为学习的标准模式,经网络“记忆”后,再输入实测水质参数作为测试样本,采用网络的模式联想功能,得出水样的富营养化类别。 结果表明,西湖水域整体上处于中营养及富营养水平;从空间上看,小南湖的水质较好,而北里湖的较差;从时间上看,一到三月份水质较好,六到十月份水质较差。研究表明Hopfield网络模型能有效合理地模拟富营养化进程中的非线性行为,并能准确评价西湖水质的富营养化状况,为及时有效地的提出控制措施提供科学依据。更多还原

【Abstract】 This paper applies neural network theory to assessment of water quality, establishing multi-indexes integrated water quality assessment model with Hopfield Neural Network（HNN）, in order to explore the spatio-temporal class distribution rules in West Lake.We have established 8 sample spots in West Lake, and collected the aquatic data（2000.1-12） of the lake by routine measurement. According to Grading and Classification Standard of Lake and Reservoir in Communique Weaving Precis of Water Resource of China. Chose Chl-a、 TP、 TN、 COD_Mn、SD as input factors. The eutrophication assessment standard is classified to ten sorts. Established a Hopfield Neural Network with 50 nerves, input ten standard models as training model, then remembered by the network. Input actual samples, the sort which sample belong to will be given by the association function of HNN. This paper analyzed 96 samples in 2000, the result showed that the whole lake belonged to eutrophication standard. In spatial, water quality of L. Xiaonan is the best while that of L.Beili is the worst relatively. In temporal, water quality is better during January to March than June to October.This indicates that Hopfield network model has the ability to analyze the non-linear information in eutrophication processes and evaluate the water quality applausively in West Lake, which can provide scientific basis for environmental management and control measurements.更多还原