Study On Dynamic Prediction Method Of Sudden Pollution Accident In River Based On Model Parameters Uncertainty

Water quality has great impacts on human survival and development. However, in recent years, sudden river water pollution accidents occurred frequently, resulting in a huge loss of national economy, causing a series of environmental and social problems. Therefore, it has become an urgent problem to provide accurate and reliable information of pollutants to the relevant emergency decision makers after the occurrence of sudden river water pollution accident.This paper established the dynamic optimization method to forecast water quality after water pollution emergencies, in which uncertainty method, dynamic optimization technology, GLUE algorithm and the improved GLUE algorithm based on MCMC algorithm are used. The method provides the emergency decision makers with uncertainty-predicting information of pollutants after the occurrence of sudden water pollution accident in the river, so as to facilitate them to take targeted emergency measures to deal with pollution accidents.The main research work and innovative aspects of the thesis are as follows:(1) In order to reduce uncertainty and improve the poor real-time performance in the process of water quality prediction, the uncertainty sources in river water pollution accidents are analyzed and the original water quality prediction framework based on uncertainty analysis is summarized. Based on this, the dynamic optimization method is introduced to improve the aforementioned water quality prediction framework, which makes full use of the measured data to correct predicted results dynamically after the accidents. In this way, prediction results can retain high accuracy and reliability as pollution accidents develop.(2) Based on one-dimensional water quality prediction model, this paper realizes a dynamic optimization framework for water quality prediction based on GLUE algorithm and Bayesian theory. The procedure to apply the method is proposed after verification of the method by real examples.(3) Considering that the dynamic optimization model for water quality prediction based on GLUE algorithm is not able to recognize of the actual distribution of model parameters, MHGLUE algorithm is proposed which is the improved GLUE algorithm based on MCMC algorithm. It can not only recognize the nearly actual distribution of model parameters but also conduct real-time uncertainty forecast based on this.(4) To further compare the performances of the proposed two kinds of methods on water quality prediction, an experiment is designed to simulate pollutant diffusion based on the wind wave flume. The two methods are then used to conduct uncertainty forecast based on experimental data. Through comparison and analysis, this paper casts light on merits and drawbacks of the two methods, and summarizes the applicable occasions respectively.