【作者】 王丽娟；

【导师】 张宏伟；

【作者基本信息】 天津大学 ， 环境工程， 2010， 博士

【摘要】 本文以城市供水管网漏损控制为核心研究内容,通过管网漏损模拟实验从微观上对漏损管网进行水力分析,建立了管网漏损模型与漏损检测与定位模型,并对既有漏失模型进行了验证分析,为有效地减少漏损提供了理论依据;通过对实际运行管网漏损数据的挖掘,从宏观上剖析了管网漏损的成因与规律,建立了管网漏损预测模型,为管网漏损的早期预防、维护与更新以及资金的合理分配与使用提供决策支持。首先,在实验室条件下模拟管网渗漏和节点爆管,通过对单点渗漏、两点渗漏、节点爆管、节点爆管与渗漏共同引起的压降及漏损前后漏损管段和大用户流量变化对管网漏损的初步定位进行了研究并用EPANET对管网渗漏及节点爆管压降进行了模拟。其次,结合实验管网监测点压力值和选取管段的流量测定值对供水管网漏损检测与定位模型进行了研究,建立了基于贝叶斯理论和费希尔定律的漏损辨识与定位模型、基于遗传算法的漏量估计与漏损定位模型、基于聚类与判别分析的漏损定位模型,为漏损故障实时诊断奠定了基础,减少了维修的盲目性。另外,对影响管网漏损的水力因素及漏损模型进行了研究。通过漏损试验分析及对漏损管段上所有影响管网漏损的水力因素进行主成分分析及逐步递归分析,确立了影响漏失量的显著因素并建立了管网漏失模型;对漏失量和管道压力的指数模型进行分析并拟合出压力指数区间;通过对模型漏损系数点估计与区间估计对既有典型管网漏失模型进行了验证研究。最后,对实际运行供水管网历史漏损数据进行了统计分析,建立了RBF径向基神经网络(敷设时间<5年)、多元统计回归(敷设时间>5年)管网漏损时间预测模型;综合自回归移动平均(ARIMA ( p ,d,q)(P,D,Q)~s)和elman反馈型神经网络两种漏子数时间序列预测模型;对漏损部位进行预测的有序多分类逻辑斯谛(logistic)非线性回归模型。更多还原

【Abstract】 In this paper, leakage control in municipal water supply network was mainly studied. The models of water loss as well as the models of leakage detection and location were established by hydraulic analysis under simulation lab-scale environment which can provide theory to reduce water loss effectively,the existing models were verified simultaneously;Based on date mining, the macro factors and law of leakage was disserted and the prediction model was presented which can provide technical support for taking initiative control of leakage,policy decision to prevention,maintenance,renewal and rational use of renewal fund.First of all, leakage on pipes and bursts at nodes were simulated and the pressure loss as well as flow changes of pipe and consumer caused by single、two leakages and bursts was studied which was the basis of preliminary location.The pressure loss caused by leakages and bursts was simulated by EPANET The leakage at the same time.Secondly, approaches to identify the location and severity based on the pressure and flow rate at some monitored points were developed based on Genetic Algorithm and BP neural network , Bayesian probabilistic framework as well as R. A. Fisher theory, Cluster Observations Discrimination Analysis theory which can lay a theory foundation to real-time troubleshooting and reduce blindness to maintenance of water distribution network.Further more, the remarkable factors were found after principal components analysis and stepwise regression to hydraulic factors influencing water loss on pipe and the models were built; Index section of pressure was drawn up based on analysis to relationship between the amount of leakage and pressure in different leakage area; The existing models were verified by means of estimate the leakage value and section value of coefficients.Lastly, The leakage and burst number was studied statistically.Following this, the leakage time prediction models of RBF neural network (laid time<5year) and multi-linear regression model (laid time>5year) were found, The time alignment prediction model of ARIMA ( p ,d,q)(P,D,Q)sand ELMAN feedback neural network as well as Ordinal Logistic Regression model to leakage point were established too.更多还原