BP神经网络在千岛湖水体富营养化变化预测中的应用  

Back-Propagation Network Model for Predicting the Change of Eutrophication of Qiandao Lake

作　　者：刘恒[1] 严力蛟[1] 

机构地区：[1]浙江大学生命科学学院生态规划与景观设计研究所,杭州310058

出　　处：《科技通报》2008年第3期411-416,共6页Bulletin of Science and Technology

基　　金：国家自然科学基金(69673044);杭州市环境保护局资助项目(9901)

摘　　要：将人工神经网络模型引入水质预测中,并据此建立了千岛湖水体富营养化预测的BP神经网络模型。该模型选取了Chla作为网络的输出变量,通过主因子分析,得到温度Tw、pH、Chla、SD、TN 5个水质因子作为网络的输入变量,构建了5个网络模型。本研究表明,以上周的Tw、pH、Chla、SD为输入变量,下周Chla为输出变量的网络方案能够对千岛湖的水质变化进行很好的短期预测,从而能够使管理部门根据此模型掌握千岛湖水质变化趋势,为其制定千岛湖水质管理方案提供理论依据。The study introduced the Artificial Neural Network Model into predicting the change of eutrophication of Qiandao Lake, and the BP network model for predicting the change of eutrophication of Qiandao Lake was established. The model chose the factor of Chla as the output of the network and Tw,pH,Chla,SD,TN as the input. Five models were established. The result showed that the second model was the best one. In this model, Tw,pH, Chla,SD of the last week were the input, and the Chla of the next week was the output. This model can predict the change of water quality in a short term. As a result, the administration department can master the change of water quality in future and make the proper measurement.

关 键 词：水体富营养化 BP神经网络模型 千岛湖 

分 类 号：X22[环境科学与工程—环境科学]