基于贝叶斯方法的突发水污染事件溯源研究

Bayesian Inference for Source Determination of Sudden Water Pollution Events

为快速对突发性水污染事件进行溯源,求得污染物的排放位置、排放量及排放时间,提出一种基于贝叶斯(Bayesian)和马尔科夫蒙特卡洛方法(MCMC)的污染源信息反演算法。基于Bayesian-MCMC方法可在已知污染源先验信息的基础上,构造似然函数,求得污染源的后验概率密度函数,进而将溯源问题转化为对后验概率密度函数的抽样问题。在抽样方法上本文选用M-H采样方法及GIBBS采样方法并加以改进、对比。结果表明:该方法能够较准确地对突发点源岸边污染物瞬时排放事件进行溯源,其计算结果接近于真实值,能够有效地解决点源岸边污染物瞬时排放的溯源问题。且改进的M-H采样方法可以有效加快迭代时的收敛速度,使待反演参数的抽样值更快地趋近于目标值。

In order to quickly trace the sudden water pollution incident and obtain the emission location,emission and emission time of pollutants,this paper proposes a pollution source information inversion algorithm based on Bayesian and Markov Monte Carlo method.Based on the Bayesian-MCMC method,the likelihood function can be constructed based on the previously known information of the pollution source,and the posterior probability density function of the pollution source can be obtained,and then the traceability problem can be transformed into the sampling problem of the posterior probability density function.In the sampling method,the M-H sampling method and the Gibbs sampling method are selected,improved and compared.The results show that the method can trace the transient emission events of sudden point source more accurately,and the calculation result is close to the true value,which can effectively solve the traceability problem of instantaneous emissions of point source rim pollutants.And the improved M-H sampling method can effectively speed up the convergence speed in the iteration,so that the sampled value of the parameter to be inverted is closer to the target value.