基于BP神经网络方法的引滦入津工程黎河段水质预测研究  

Predicted water-quality evaluation of Lihe sector in Luanhe diversion to Tianjin water project based on BP neural network

作　　者：赵世新[1] 张晨[1] 李静怡[2] 高学平[1] 

机构地区：[1]天津大学建筑工程学院,天津300072 [2]中交第四航务工程勘察设计院有限公司,广州510230

出　　处：《安全与环境学报》2010年第2期93-96,共4页Journal of Safety and Environment

基　　金：国家自然科学基金项目(50909070)

摘　　要：根据引滦入津工程黎河段前毛庄断面水质监测数据,采用BP神经网络与非线性时间序列相结合的方法,建立BP网络非线性时间序列水质模型。应用该模型对氯化物和氨氮水质指标进行预测。结果表明,模型预测精度较好。通过预测结果验证了模型的可靠性。与机理性模型相比,提出了该模型的应用条件及优缺点。The present paper is aimed at introducing its authors＇ predicted water-quality study of Lihe River sector in the water project for luanbe River diversion to Tianjin based on the BP neural network evaluation approach. So far, little attention has been for the time being paid to the study of irrational water quality. Based on the field-study data of the water quality of the QIANMAOZHUAN section of Lihe River in the Luanhe River-Tianjin water diversion project, we have established the non-liner time series of BP neural network water quality model by joining the methods of non-linear time series and BP neural network evaluation method. The application of the model in predicting the concentration of water quality indexes （Chloride and Ammonia Nitrogen, for example） in the QIANMAOZHUAN section indicates that the model has higher precision. The relative errors of results are within 20%. According to the training and testing of the network, the results of prediction show that the model is reliable. Compared with the rational water quality model, we have also pre- sented the practical conditions and advantages and disadvantages of the non-linear time series of BP neural network model. The so-called non-linear time series of BP neural network method, so far as we can see, belongs to the black box-type method and the mapping of non- linear relationship. Therefore, the method is exPected to overcome the disadvantages of solving the complex relationships of the variables in the rational water quality model. The relative errors of results can be reduced and the simulation accuracy can be improved in the irrational water quality model. The establishment of the model helps to expand the irrational water quality model and provide a new research approach to water quality simulation. It can thus be able to provide technical support and reference to securing the water quality of Luan River-water-diverted- to-Tianjin project.

关 键 词：环境水利 非机理水质模型 水质预测 BP神经网络 非线性时间序列 

分 类 号：X522[环境科学与工程—环境工程]