首页 | 本学科首页   官方微博 | 高级检索  
     

基于SVM的管网状态估计模型
引用本文:俞亭超,张土乔,吕谋.基于SVM的管网状态估计模型[J].哈尔滨工业大学学报,2005,37(9):1205-1208.
作者姓名:俞亭超  张土乔  吕谋
作者单位:浙江大学,土木工程学系,浙江,杭州,310027;青岛建筑工程学院,环境工程系,山东,青岛,266000
基金项目:国家自然科学基金资助项目(50078048).
摘    要:为建立未知节点压力和可知监测信息之间的管网状态估计模型,应用支持向量机算法,建立基于支持向量机的管网状态估计模型和测压点压力宏观模型.经仿真分析,SVM模型90%以上预测数据的绝对误差控制在0.01MPa;与BP神经网络模型相比,同样的样本集均方误差情况下,其测试集均方误差一般比BP神经网络精度高.杭州市管网的实例计算中,85%以上测压点预测数据的相对误差都在5%以内,结果相当理想.

关 键 词:支持向量机  给水管网  状态估计
文章编号:0367-6234(2005)09-1205-04
修稿时间:2003年12月1日

State estimation model of water distribution network based on SVM
YU Ting-chao,ZHANG Tu-qiao,Lu Mou.State estimation model of water distribution network based on SVM[J].Journal of Harbin Institute of Technology,2005,37(9):1205-1208.
Authors:YU Ting-chao  ZHANG Tu-qiao  Lu Mou
Abstract:The state estimation model of water distribution network, which establishes the relation between unknown junction pressure of pipes and known monitor information, can been seen as a nonlinear function Estimator implementing the nonlinear mapping of input and output data. Moreover, support vector machine (SVM) is a good nonlinear function estimator. SVM is used to establish a state estimation model and a macroscopic model of monitor pressures. According to simulating analysis, the absolute error of 90% forecasting data is smaller than 0. 01 MPa. Compared with BP neural network, under the same mean square error of samples, the SVM approach has higher precision of the mean square error of testing sets. The application examples in Hangzhou city show that the relative error of 85% forecasting junction pressure is smaller than 5%, which is considerably perfect.
Keywords:support vector machine  water distribution network  state estimation
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号