应用贝叶斯理论的河流污染源重建探讨

Reconstruction of the polluted source based on Bayesian inference

为深入了解河流中有毒有害物质事故性污染的污染源、污染物质扩散等情况,通过MATLAB实现了河流中污染物质一维扩散模型的差分解法。该扩散模型可用于模拟1个或多个瞬时点源、持续稳定点源和持续变化点源情况下,河流中污染物质浓度随时间的变化,具有一定准确度。根据贝叶斯理论,结合扩散模型和监测数据估计污染源的位置和强度。并采用马尔可夫链蒙特卡罗方法进行后验推断,以解决贝叶斯模型计算复杂的问题。已知污染源情况下,可采用本文给出的扩散模型模拟河流下游污染物质浓度随时间的变化;污染源未知而拥有一定监测数据情况下,则可以采用本文提出的河流污染源重建技术估计污染源的位置、强度等。

The given paper intends to introduce the author＇s approach to reconstructing the pollution source in the river. As a matter of fact, it is often the case for the accidental release of hazardous material into a river to result in serious condemnation of water in a large area of the river. Such hazardous accidents often lead to the urgent need to simulate the consequence of the pollutant in the lower reach before the pollutants reach the lower reach areas. However, in most situations, it is passible for us to get somewhat monitoring data without knowing the actual pollution source. In such cases, it is necessary for us to find a method to estimate the pollution source with the help of the monitoring data. The method herein proposed should combine the data with the dispersion model by Bayesian inference so as to get to know the distribution of the parameters of the pollution source. In doing so, first of all, we need to use the difference algorithm of the so-called one-dimensional dispersion model with MATLAB so as to simulate the consequence of the river pollutants, ceaselessly changing with one or more instantaneous point-sources, the consequential stable point sources or/and some consequential dynamic point sources. Then Bayesian inference can be used to join the dispersion model and the data together to estimate the accurate location and the intensity of the pollution source. The application of Markov Chain Monte Carlo algorithm to calculating the posterior distribution has made us overcome the problem of complex computation. Such algorithms as Bayesian inference and MCMC can also be realized by MATLAB. Therefore, with the help of our method, if we have the data of pollution source, we can simulate the situation of pollutant consequence along with the time. In addition, we can also collect some more data from the lower-reach monitoring work so as to estimate the parameters of pollution source with the help of the reconstruction model. Thus, it can be seen that the method of reconstruction proves to be a method both flexible and robust, which is suitable for study in operational emergency response together with any dispersion models.