基于多尺度小波LSSVR的污水处理过程软测量  

Soft sensor modeling of sewage disposal process based on multi-scale wavelet least square support vector regression

作　　者：王鲜芳[1,2] 朱晓霞[3] 吴瑞红[1] 郑延斌[1,2] 

机构地区：[1]河南师范大学计算机与信息技术学院,河南新乡453007 [2]河南省高校计算智能与数据挖掘工程技术研究中心,河南新乡453007 [3]河北科技大学理学院,河北石家庄050018

出　　处：《计算机与应用化学》2012年第7期797-800,共4页Computers and Applied Chemistry

基　　金：国家自然科学基金资助项目(61173071);河南省高校创新人才支持计划项目(2012HASTIT011);河南省科技攻关计划项目(112102210412;102102210176);;河南师范大学博士启动基金项目(1039)

摘　　要：针对污水处理中某些生物参数难以在线测量的情况,本文提出了一种基于小波核的多尺度最小二乘小波支持向量机软测量建模方法:。首先,选取墨西哥草帽小波函数作为最小二乘支持向量机的核函数,进而设计出多尺度小波最小二乘支持向量回归机(MW-LSSVR)。然后利用该支持向量机和出水水质参数特性建立混合软测量模型,实现对出水BOD浓度、COD浓度在线预测。通过在实际污水处理过程的应用,结果:表明本建模方法:具有较高的预测精度和较快的模型学习速度,能对BOD的做出准确的预测,一定程度上可以替代某些昂贵的在线测量仪表,给污水处理厂工作人员提供了控制操作依据,具有一定的实际应用价值。To solve the problem that some parameters are difficult to be measured on-line in the process of waste water disposal, a soft measurement modeling method is presented base on multi-scale wavelet least square support vector machine in this Paper. Mexican-hat wavelet function is used as the support vector kernel function, and further the Multi-scale Wavelet Least square Support Vector Regression （MW-LSSVR） algorithm is presented. Build an advanced model with above SVR and characteristics between BOD＆COD, predicting BOD＆COD of drainage that had been treated. Through using this method in practical sewage disposal process, the result shows that this modeling method has higher precision and faster learning speed of BOD model, can make accurate predictions, can replace online measuring instrument in some expensive, provide control operation basis to the sewage treatment plant workers, and has a certain practical value.

关 键 词：小波 最小二乘支持向量机 多尺度学习 污水处理 

分 类 号：TQ015.9[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]