Prediction and Analyses of Residual Chlorine Based on Support Vector Regression in Urban Water Distribution System
TIAN Yi - mei, WU Mi - fang, Wang Yang
DOI:10.11835/j.issn.1674-4764.2006.02.021
Received December 30, 2005,Revised December 30, 2005, Accepted , Available online July 01, 2015
Volume ,2006,Pages 74-78
- Abstract
Support vector regression(SVR) algorithm is an application of structural risk minimization principle in function regression.In this paper,a residual chlorine prediction model based on SVR is established by using the data of manual sampling residual chlorine of water distribution system in a certain city in the north of China.SVR model is compared with the artificial neural network and multivariate linear regression.The result shows that SVR model has better generalization ability for small samples,the predicted average relative error of all monitoring points is 1.80%~8.73%,and can achieve unique and globally optimal solutions.It is practical and can solve the problem for small samples of residual chlorine when the fit precision of model is good but the predicted effect is worse.