基于伴随方程和MCMC方法的室内污染源反演模型研究  

Investigation of the Inversion Modeling for Indoor Contaminant Source Based on the Adjoint Equation and MCMC Method

作　　者：郭少冬[1,2] 杨锐[1] 苏国锋[1] 张辉[1] 

机构地区：[1]清华大学工程物理系,公共安全研究中心,北京100084 [2]北京应用物理与计算数学研究所,北京100094

出　　处：《应用基础与工程科学学报》2010年第4期695-704,共10页Journal of Basic Science and Engineering

基　　金：国家自然科学基金(50804027)

摘　　要：研究了室内污染物扩散后的源反演方法.通过数值求解浓度场的伴随方程,并结合传感器测量信息,构造似然函数;利用基于贝叶斯推断理论的MarkovChainMonte Carlo(MCMC)抽样方法,对污染源的位置、强度的后验概率进行计算,反演结果与污染源的真实参数吻合.该方法与传统的室内污染物反演方法相比,极大地降低了计算量.此外,还讨论了传感器性能对结果的影响,研究表明传感器误差概率分布越平坦,污染源反演信息的不确定度越大;而过低的测量灵敏度,则会导致反演结果呈现多个局部极值点的特性.Source inversion and identification method of contaminant dispersion in buildings was studied.An adjoint transportation equation was introduced and solved numerically,and a likelihood function was computed using predicted concentration database.Markov Chain Monte Carlo（MCMC） sampling based on Bayesian inference was used to inverse the parameters, including the source location and intensity.The agreement of the predicted results with the actual source location and intensity indicated that the method could be used to estimate indoor source parameters effectively.The studies also showed that the proposed procedure was more efficient when using the adjoint transportation equation with MCMC methods.The effects of the error probability distribution and detection limits of the sensors on the sensitivity and reliability of the method were then discussed.Analyses show that flatter and broader probability distribution of the sensor error leads to inversion results with larger uncertainty,while the lower sensor sensitivity leads to insufficient information and ill-posed characteristics of the inversion results with several clusters of possible source locations.

关 键 词：室内污染源反演 贝叶斯推断 伴随方程 计算流体力学 

分 类 号：X830.2[环境科学与工程—环境工程]