贝叶斯方法在水力管网阻力特性辨识估计中的应用研究  

Resistance Characteristic Identification and Estimation of Water Networks Based on Bayesian Method

作　　者：刘永鑫 金虹 罗鹏 展长虹 LIU Yongxin;JIN Hong;LUO Peng;ZHAN Changhong(School of Architecture,Harbin Institute of Technolog;Heilongjiang Cold Region Architectural Science Key Laboratory,Harbin 150090,China)

机构地区：[1]哈尔滨工业大学建筑学院黑龙江省寒地建筑科学重点实验室,哈尔滨150090

出　　处：《建筑科学》2018年第8期89-93,105,共6页Building Science

基　　金：国家自然科学基金项目(51678177)

摘　　要：为更准确地构建管网水力模型,获取管网运行条件下阻力特性是关键步骤之一。本文针对以往不能充分利用管网设计和运行数据资料、多水力工况管段阻力特性辨识计算耗时较大等问题,提出基于贝叶斯统计分析的管网阻力特性辨识估计方法。将管段阻力特性系数作为统计分析中的模型参数,在获取管网拓扑结构以及基本设计资料的前提下,依据管网设计资料确定符合均匀分布的管段阻力特性系数先验概率分布;结合运行工况下的管网压力和流量观测数据,利用基于广义逆矩阵理论的辨识方法计算对应各水力工况的辨识结果,作为后验样本数据;将管网设计信息与运行观测信息综合体现于后验概率分布中,从而确定各管段阻力特性系数的估计结果。本文采用数值模拟手段验证方法的应用效果,结果表明:分析所得可信区间可以保证以95%以上的可信水平包含各管段阻力特性系数真实值。该辨识估计方法可为构建准确的管网水力模型奠定基础。To build an accurate hydraulic model of water networks,obtaining resistance characteristics under operational conditions is one of the key steps. Designed data and operational data of water networks cannot be used at the same time with previous methods due to the time-consuming calculation process. Parameter identification of pipe resistance characteristics based on Bayesian statistical analysis is proposed. The pipe resistance characteristic coefficients are seen as model parameters in Bayesian statistical analysis. When topology and basic design data of a water network are given,the prior probability distribution of pipe resistance characteristic coefficients will be determined via uniform distribution. Combining the operational data of node pressures and pipe flows, the identification results of pipe resistance characteristic coefficients can be obtained with the method based on MoorePenrose pseudo-inverse solution. The solution also indicates posterior probability in our study. The posterior probability distribution is determined by both design material and operational data. Then, pipe resistance characteristic coefficients can be obtained according to the probability estimation results. In this paper,the numerical simulation method is used to validate the method. The results show that the probability of the credibility interval including the true value of pipe resistance characteristic coefficients is higher than 95%（ the credible level）. This identification estimation method can be the foundation for building an accurate hydraulic model of water networks.

关 键 词：水力管网 均匀分布 贝叶斯方法 阻力特性 辨识估计 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程]