基于BP神经网络的藻类水华预测模型与研究.docVIP

基于BP神经网络的藻类水华预测模型研究 张克鑫，陆开宏 （ 宁波大学应用海洋生物技术教育部重点实验室宁波 315211）ZHANG Ke-xin,LU Kai-hong,ZHU Jin-yong,LIU Xia-song,XIE Li-feng （Ministry of Education Key Laboratory of Applied Marine Biotechnology， Ningbo University， 315211，Ningbo，Zhejiang，China） Abstract: 30 weeks of monitored data from March to October in 2009 in the School pond of Ningbo University,we constructed a predicting model to deal with the relation between the density of Oscillatoria and 6 environmental factors such as total nitrogen, total phosphorus, secchi depth,etc with the back propagation artifical neutral network method.We selected the best predicting model,and sensitivity analysis was performed to the model.The results showed that the forecasted value of the density of Oscillatoria according to the BP neural network predicting model had a better fit with actual value of the density of Oscillatoria,and the correlation coefficient achieved 0.984,it indicated the BP neural network predicting model can be used for short-term forecast of the algal blooms; and through carried on sensitivity analysis to the constructed BP neural network predicting model,it clarified the main driver factor of algal blooms in the School pond of Ningbo University,and the result showed that controlling PH value would be important to prevent and control the algal blooms in the School pond of Ningbo University. Key words: BP neural network; algal bloom; predicting model; the density of Oscillatoria 中图分类号：X824 文献标识码：A 文章编号： 1 研究对象及方法 1.1 宁波大学校内池塘概况 宁波大学校内池塘是一个小型藻华水体，该水体富营养化严重，年蓝藻水华，优势藻类主要为蓝藻门的微囊藻属（Microcystis sp.）、色球藻属（Chroococcus sp.）颤藻属（Oscillatoria sp.）等的一些种类。水华常于春夏之交开始出现，持续至深秋甚至初冬。Oscillatoria(cells/L) 0 21×107 2.81×107 4.28×107 图 1 池塘各环境变量及颤藻（Oscillatoria）生物量随采样时间的监测值 Fig.1 The monitoring value of the environmental factors and the density of Oscillatoria with time in the pond 3 模型的设计、选择及验证 3.1 模型数据集的选取及处理 在对宁波大学校内池塘小型藻华水体进行跟踪调查期间，共有30周的监测数据。在模型构建过程中，选择总氮（TN，mg/L）、总磷（TP，mg/L）、透明度（SD，m）、溶解氧（DO，mg/L）、PH、温度（WT，℃）这6个环境