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.
出处
《传感技术学报》
CAS
CSCD
北大核心
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)
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
soft measurement
data fusion
experiment and simulation
SBR(sequencing batch reactor)
COD(chemical oxygen demand)
SVM(support vector machine)
neural network