State Key Laboratory of Hydroscience and Engineering, Department of Hydraulic Engineering, Tsinghua University, Beijing, 100084, China.
State Key Laboratory of Hydroscience and Engineering, Department of Hydraulic Engineering, Tsinghua University, Beijing, 100084, China.
Environ Pollut. 2018 Oct;241:759-774. doi: 10.1016/j.envpol.2018.05.093. Epub 2018 Jun 13.
Pollution risk from the discharge of industrial waste or accidental spills during transportation poses a considerable threat to the security of rivers. The ability to quickly identify the pollution source is extremely important to enable emergency disposal of pollutants. This study proposes a new approach for point source identification of sudden water pollution in rivers, which aims to determine where (source location), when (release time) and how much pollutant (released mass) was introduced into the river. Based on the backward probability method (BPM) and the linear regression model (LR), the proposed LR-BPM converts the ill-posed problem of source identification into an optimization model, which is solved using a Differential Evolution Algorithm (DEA). The decoupled parameters of released mass are not dependent on prior information, which improves the identification efficiency. A hypothetical case study with a different number of pollution sources was conducted to test the proposed approach, and the largest relative errors for identified location, release time, and released mass in all tests were not greater than 10%. Uncertainty in the LR-BPM is mainly due to a problem with model equifinality, but averaging the results of repeated tests greatly reduces errors. Furthermore, increasing the gauging sections further improves identification results. A real-world case study examines the applicability of the LR-BPM in practice, where it is demonstrated to be more accurate and time-saving than two existing approaches, Bayesian-MCMC and basic DEA.
工业废水排放或运输过程中的意外泄漏造成的污染风险对河流的安全构成了相当大的威胁。快速识别污染源对于污染物的紧急处理至关重要。本研究提出了一种新的河流突发性水污染点源识别方法，旨在确定污染物在何处（源位置）、何时（释放时间）以及多少（释放质量）进入河流。基于后向概率法（BPM）和线性回归模型（LR），所提出的 LR-BPM 将源识别的不适定问题转化为优化模型，使用差分进化算法（DEA）进行求解。解耦的释放质量参数不依赖于先验信息，提高了识别效率。通过对不同数量污染源的假设案例研究，验证了所提出方法的有效性，所有测试中识别出的位置、释放时间和释放质量的最大相对误差均不超过 10%。LR-BPM 的不确定性主要归因于模型等价性问题，但重复测试结果的平均值可大大降低误差。此外，增加测站数量进一步提高了识别结果的准确性。实际案例研究检验了 LR-BPM 在实践中的适用性，结果表明它比贝叶斯-MCMC 和基本 DEA 两种现有方法更准确、更节省时间。
