RBF神经网络氧化沟系统出水氮磷预报模型  

RBF Neural Network Modeling of Effluent TN and TP in Oxidation Ditch System

作　　者：陈安[1] 孙耀华[2] 王勉[2] 徐新年[1] 唐大平 

机构地区：[1]黄石市环境保护研究所,湖北黄石435000 [2]河南省漯河市污水净化中心,河南漯河462000

出　　处：《环境科学与技术》2009年第10期124-128,136,共6页Environmental Science & Technology

摘　　要：为定量模拟污水处理系统进水及出水水质参数数学关系,为污水处理系统的智能反馈控制奠定理论基础,文章以河南漯河市污水净化中心氧化沟系统为考察对象,采用径向基函数(RBF)神经网络对其模拟分析,建立了氧化沟系统出水TN、TP预报的RBF网络模型。建模过程采用的主成分分析与聚类分析有效挖掘了样本信息,采用的数据预处理方法缩减了模型误差。模型性能及灵敏度检验表明,建成的模型对出水TN、TP预报准确率分别达到90%、70%,相关性检验系数分别达到0.95和0.89,可用于该系统出水TN、TP预报,为系统在线控制提供指导。研究同时表明,RBF神经网络由于克服了误差反向传播(BP)网络收敛慢、局部极值等缺点,在水处理系统模拟及其反馈控制中,具有巨大的应用潜力。To study the basic theory of intelligent feedback control of wastewater treatment system, the quantitative relations linking influent and effluent parameters in the carrousel oxidation ditch system in a wastewater treatment plant was simulated. Advanced radius basis function （RBF） neural network （NN） was used to develop an adaptive model for predicting effluent total nitrogen （TN） and total phosphorus （TP）. In the process of model building, the methods of principal component analysis and cluster analysis were effective in data mining, and the data pretreatment was helpful to reduce the errors of prediction. The neural network has a good performance in test and can adapt to various situations, with TN 94% accurate and TP 70% accurate in effluent parameters forecasting. Correlation coefficient test indicated that TN was 0.95 and TP 0.89. The neural network can provide guidance for system control on-line. The study showed that RBFNN was superior to BPNN, which has great potential in modeling and control feedback of water treatment system.

关 键 词：径向基函数 神经网络 氧化沟系统 总氮 总磷 预报 

分 类 号：X84[环境科学与工程—环境工程]