An Approach Based On SWMM And Bayesian-inference To Identify Illicit Discharge Sources In Sewer Systems

In the urban drainage system,accidents caused by the illicit discharge of industry waste water to the municipal sewage system frequently occur.They often have severe impacts on the activated sludge process in the sewage treatment plant,and even cause deterioration of effluent quality due to activated sludge poisoning.Hence it is important to identify illicit discharge location and discharging history.Existing source identification methods based on physical approaches have low efficiency and take a long feedback time,which is insufficient in the identification of the drain accidents.Meanwhile,limited source inference methods based on mathematic model have difficulties in the application in the drainage pipe system,in which source inference problem is highly uncertain due to the complex topology and hydraulic conditions.Hence,in this research,the illicitly discharged industrial pollutants in the drainage system was taken as the research object.A mathematical source identification model based on Bayesian inference algorithm,SWMM model and MATLAB programming in drainage networks was constructed.The three unknown discharge characteristic parameters including discharge location,discharge amount,and discharge time period,were statistically retrieved,and the probability distributions of the variables were obtained.The influence of inference parameters such as the value of the walking step size in the statistical inversion algorithm、the spatial arrangement of the on-line monitoring points and the time interval of the sampling monitoring,etc.,were studied in detail.The model used the advanced Markov Monte Carlo.The MCMC sampling algorithm greatly improves the efficiency and provides more accurate results within a limited feedback time.Hence it could improve the management level of waste water and the emergent treatment efficiency in the discharge accidents.The main research contents and conclusions are as follows:(1)The integration of SWMM,Bayesian method and MATLAB was carried out.This integration can output the time series of SWMM simulation results through MATLAB and match the corresponding monitoring data sequence to improve the analysis accuracy.Hence it could provide a dynamic feedback that is closer to the real condition of the pipe network.Results show that the coupled source identification model can provide a useful range of the unknown discharge source parameters and enhance the efficiency of a physical-based model.(2)Bayesian MCMC sampling algorithm was applied to study the relationship between walking step size and inversion efficiency.Results show that if the step size is too small,the sampling will be limited to the local solution while a large step size can move in a larger search space and avoid being confined to a local solution.Hence a larger step size can realize the sampling of the whole posterior space.However,a step size that is too large would increase the number of samples and hence an extended inversion time.(3)Influence of monitoring point spatial arrangement and monitoring time interval on the effectiveness and accuracy of inversion algorithm was investigated.Results show that when the observation data of the monitoring point is too small,the posterior probability density of the pollution source parameters cannot be accurately inverted.This is because the observation data does not reflect the concentration sequence of the monitoring point when observation data increase and monitoring time interval is relatively small.A more complete pollutant concentration curve means a longer monitoring time.Hence it can provide a more accurate Inversion results.(4)A real drainage network was taken as a study case using the source-identification model developed.Three scenarios that reflects the availability of the source information are investigated including a single-variable,double-variable,and three-variable that represents one,two and three discharge characteristics of the discharge accident needs to be identified.Two discharging mode was studied including an instantaneous and a continuous discharge.Inferenced source information’s were compared with the theoretical solution to evaluate the accuracy of the statistical traceability model.Results show that the model provides accurate inversion results in cases that involves only one unknown inversion parameter.However in cases where two or three inversion parameters needs to be inferred,the model can provide a reasonable range of source information’s.But more inversion parameters,more discrete of the posterior probability density。Since the purpose of the Bayesian inference method is to find out the possible values and give the probabilities of all the values,the basic idea is to narrow down the prior range and focus on the points where the probability distribution is large.Therefore,the result of SWMM-Bayesian inference model is reasonable and practical.The research results of the paper can provide technical support for water pollution control and drainage management.