支持向量机和神经网络联合软测量SBR污水处理中COD的方法  

SVM and Neural Networks Joint Approach to the Soft Measurement of COD Values in SBR Wastewater Treatment Systems

作　　者：张杰[1] 张建秋[1] 冯辉[1] 雷中方[2] 胡波[1] 

机构地区：[1]复旦大学电子工程系,上海200433 [2]复旦大学环境科学与工程系,上海200433

出　　处：《传感技术学报》2009年第10期1519-1524,共6页Chinese Journal of Sensors and Actuators

基　　金：上海市科委科技攻关项目资助(07dz15013)

摘　　要：为了实时测量序批式活性污泥法(SBR)污水处理系统中的化学需氧量(COD),提出了一种支持向量机和神经网络联合软测量SBR污水处理中COD的方法。针对COD软测量建模中有限种类辅助变量造成的矛盾数据问题和神经网络学习的局部最小问题,该方法通过引入支持向量机对COD值进行预估计,再根据COD的变化规律使用两种神经网络模型分别估计污水COD值。实验表明:本文方法的软测量结果优于单一神经网络的软测量结果。In order to measure the chemical oxygen demand （COD） of sequencing batch reactor （S]3R） sewage treatment systems in real-time, a method jointing a support vector machine and neural networks to softmeasuring COD values of SBR wastewater treatment is presented. To solve the conflicting data problems caused by the limited types of auxiliary variables and local minimum problems of the neural network models for the COD soft measurement, the proposed method first estimates the COD values by a support vector machine. The two neural network models in accordance with the changes of the COD are respectively used to estimate the COD values. The experiment results show that the soft-measurement results obtained by the proposed method in this paper are superior to those achieved by a single neural network.

关 键 词：软测量 数据融合 实验及仿真 序批式活性污泥法(SBR) COD 支持向量机 神经网络 

分 类 号：X703[环境科学与工程—环境工程] TP183[自动化与计算机技术—控制理论与控制工程]