基于Bayesian-MCMC方法的水体污染识别反问题

Event Source Identification of Water Pollution Based on Bayesian-MCMC

针对具有不适定性的环境水力学反问题,基于贝叶斯推理和二维水质模型建立水体污染识别反演模型,运用马尔科夫链蒙特卡罗法抽样获得污染源源强、污染源位置和污染泄漏时间等模型参数的后验概率分布和统计结果.实例研究结果表明,基于马尔科夫链蒙特卡罗抽样算法的贝叶斯推理可以较好地用来实现水体污染识别,具有识别精度高,误差小的特点,其可靠性和稳定性高于混合遗传模式搜索优化算法.

For the ill-posed environment hydraulic inverse problem, a methodical model was constructed based on Bayesian inference and two-dimensional water quality model. Markov chain Monte Carlo simulation was applied to get posterior probability distribution of the source＇s position, intensity and event init time. The result of case study shows that the method based on Bayesian inference with Markov chain Monte Carlo simulation is fit for in- verse problem such as contamination event source identification featuring high accuracy and little error. Compared with the identification results of hybrid genetic algorithm and pattern search, the presented approach indicated high stability and robust on the same inverse problem.