摘要
基于贝叶斯理论,结合蒙特卡罗马尔可夫链算法,对SWMM的水质水动力模块进行了二次开发,反推了管网入渗的节点位置、入渗流量时间序列等特征参数;分析了随机游走、均匀分布和正态分布分别作为建议分布对反演精度的影响,并对贝叶斯理论下蒙特卡罗马尔科夫链算法的似然函数进行了改进。结果表明:3种建议分布均可在贝叶斯理论下达到收敛,均匀分布与正态分布对算法初值的依赖较小,算法的全局遍历性较好,随机游走则会因局部最优降低反演精度;3种建议分布下的反演过程均出现误差累积现象,采用改进后的蒙特卡罗马尔科夫链算法可有效避免误差的累积,提高了时间序列变量的反演精度。
Based on Bayesian theory and combined with Monte Carlo Markov chain algorithm,a secondary development of water quality hydrodynamic module of SWMM was carried out,and characteristic parameters such as node positions and infiltration flow time series of the pipeline network were deduced.Analyze the impact of random walk,uniform distribution,and normal distribution as suggested distributions on inversion accuracy,and improve the likelihood function of the Monte Carlo Markov chain algorithm under Bayesian theory.The results indicate that all three suggested distributions can converge under Bayesian theory.Uniform distribution and normal distribution have less dependence on the initial value of the algorithm,and the algorithm has good global traversal.Random walk will reduce the inversion accuracy due to local optima.The inversion process under the three suggested distributions all showed error accumulation.The modified Monte Carlo Markov chain algorithm can effectively avoid the accumulation of errors and improve the inversion accuracy of the time series variable.
作者
常留红
薛雄
郭洋
高宏宇
邬传峰
CHANG Liuhong;XUE Xiong;GUO Yang;GAO Hongyu;WU Chuanfeng(School of Hydraulic and Environmental Engineering,Changsha University of Science&Technology,Changsha 410114,China;Key Laboratory of Water and Sand Science and Water Disaster Prevention and Control in Hunan Province,Changsha University of Science&Technology,Changsha 410114,China;Key Laboratory of Port,Waterway and Sedimentation Engineering,MOT,Nanjing Hydraulic Research Institute,Nanjing 210029,China)
出处
《水资源保护》
EI
CAS
CSCD
北大核心
2024年第5期141-147,共7页
Water Resources Protection
基金
国家重点研发计划项目(2021YFC3200403)
南京水利科学研究院开放基金项目(YK222001-8)。
关键词
污水管网
入渗流量
溯源反演
贝叶斯理论
SWMM
sewage pipeline network
water leakage flow
traceability inversion
Bayesian theory
SWMM
