Industrial Park Air Pollution Source Detection And Localization With Sparse Data

In recent years,chemical industrial parks have received increasing attention.Nowadays,it is the dominant form of chemical industrial clusters worldwide.Whereas,the compactly arranged enterprises in chemical industrial parks have raised some problems as well.For example,the emission sources in chemical industrial parks are concentrated in limited space which induces difficulties in monitoring and tracing the source of the abnormal emissions.In order to meet the requirements of air pollution control,it is particularly important to investigate these problems.The available datasets for source tracing in chemical industrial parks are relatively sparse compared to the intensive targets.The processes of atmospheric pollutant diffusion are complicated and involve a number of nonlinear and dynamic processes which require sufficient data to identify these processes.However,the monitoring methods for such complex processes are usually limited in chemical industrial parks,as sensors are sparsely distributed along the dimensions of space,time,and pollutant species.State-of-art theories and methods are usually developed based on sufficient data and ideal conditions therefore,there are currently few solutions to these challenges.Monitoring and tracing real-time anomalies of atmospheric pollutants in industrial parks is extremely difficult.To address these problems,this thesis focuses on the theory and technology used to monitor and trace abnormal emissions in chemical industrial parks with deficient sensors.The ultimate objective of this thesis is to establish a comprehensive,accurate,efficient and feasible monitoring and tracing system for chemical industrial parks.Due to the available data is sparse,the source tracing may be impossible in some conditions.The first step is to clarify whether the sources are traceable.To address this,the concept of"traceability" for atmospheric emission source tracing problems is introduced in this thesis followed by the criterion for determining the traceability of a source tracing problem.Based on this traceability criterion,an equivalent source method has been proposed to solve the untraceable problems caused by deficient sensors in chemical parks.In densely populated industrial areas,a large number of emission sources are concentrated in a limited space.Both normal and abnormal emissions overlap and share the limited space.It is challenging to monitor and identify abnormal emissions with a relatively sparse sensor network.The equivalent source method has some difficulties in accurately identifying the abnormal sources.Based on the proposed equivalent source method,an orthogonal equivalent source method using the QR decomposition algorithm has been proposed.This method uncovers a semi-independent relationship between the orthogonal equivalent sources and the real sources,and converts the tracing process into an iterative logical judgment process with sparse monitoring data,enabling the accurate identification and localization of abnormal emission sources with sparse sensors.In addition,continuously monitoring and tracking the dynamic status of emission sources in chemical industrial parks is of great significance for real-time source tracing and emission monitoring.However,the continuous monitoring and tracing of emission sources based on existing dynamic models is rarely feasible due to the curse of dimensionality caused by the increase of matrix dimensions.This thesis proposed a method of using the impulse response model as an equivalent model to substitute the dynamic diffusion model for the dynamic source tracing problem.Computation time of atmospheric dispersion model is important to realize real-time source tracing.Nevertheless,the complex models usually show superiority in accuracy and defects in computation cost i.e.,AERMOD model.In this thesis,a metamodelling-based method for source tracing has been proposed.By using the metamodel instead of the AERMOD model,the time consumption for source tracing is dramatically reduced.This improvement enables the application of the proposed source tracing methods in real-world scenarios.To investigate the performance of the methods proposed in this thesis,real world experiments have been performed in a chemical industrial park in Yangtze river delta region in China.The results of the methods are consistent with the validation data which validated the theory and performance of the technologies proposed in this thesis.