供热管网的二级BP神经网络泄漏故障诊断  

Two-stage BP neural network leakage fault diagnosis of heating networks

作　　者：雷翠红[1] 邹平华[1] 

机构地区：[1]哈尔滨工业大学市政环境工程学院,哈尔滨150090

出　　处：《哈尔滨工业大学学报》2011年第2期75-79,共5页Journal of Harbin Institute of Technology

基　　金：国家自然科学基金资助项目(50378029)

摘　　要：为诊断供热管网泄漏情况,利用图论理论描述基于空间管网的泄漏工况水力计算模型,得出节点泄漏和管段泄漏工况下管网各点的压力变化情况.在此基础上采用人工神经网络方法建立基于二级BP神经网络的供热管网泄漏诊断系统,该方法可根据管网中压力监测点的压力变化进行泄漏位置和泄漏量的诊断,并通过实例验证了方法的有效性.结果表明:一级神经网络对泄漏管段的预测结果准确率达100%,二级神经网络对泄漏位置和泄漏量的预测平均相对误差均为0.03%,检测结果令人满意.To investigate the leak detection strategy of heating network,a math model of failure hydraulic regime on spatial network is described by graph theory,and by which the pressure changes of all nodes of the network can be obtained for both node leak and pipe leak regime.On this basis,the leakage diagnosis system based on two-stage BP neural network is established,which can predict the leakage location and rate by collecting the real time data of pressure changes of the monitoring points,and its usefulness is proved by an example.The results show that the prediction of the first stage BP neural network for leakage pipe can reach to 100%,and the average relative error of the second stage prediction for leakage location and rate are both 0.03%,which are satisfactory.

关 键 词：供热管网 故障 诊断 人工神经网络 

分 类 号：TU995[建筑科学—供热、供燃气、通风及空调工程]