Prediction Of Air Pollution In Chemical Industry Park Based On Information Fusion

With the rapid development of industrialization and urbanization,a large number of harmful substances produced by human activities have led to serious urban environmental pollution events,especially the serious air pollution in petrochemical parks.There are many enterprises with high energy consumption,high emission and high pollution in the chemical industry park,which have a great impact on the air quality in this area.Fortunately,the enterprises in the existing petrochemical industry park have installed a relatively complete pollutant emission monitoring system,which can give early warning when the emission data exceeds the standard.Unfortunately,the acquired large amount of original monitoring data cannot be directly related to the air pollution in the park.It is difficult to play a role in the follow-up emergency management process because of the serious information overload caused by the poor data analysis and processing ability.How to identify the characteristic state of heavy air pollution,and reasonably predict the trend of heavy air pollution in the surrounding areas according to the pollutant emission data of enterprises in the chemical industry park,build a set of high-precision and strong-usability prediction system of air pollution in the chemical industry park which is independent of the monitoring station.It is of great significance for the government to make the real-time control of enterprise emission in the park and make emergency decision for urban environmental crisis in advance.Based on the information fusion technology,this study constructs the chemical industry park air pollution prediction system framework,and initially establishes the method and technical system of feature processing and multi model fusion.Firstly,the scenario analysis of air pollution in the chemical industry park is carried out to analyze and obtain feature factors affecting pollution events,build the disaster chain of air pollution in the chemical industry park,analyze the different air pollution events in the park,simulate the accident scenario,and provide the data source for feature extraction;then according to the calculation instructions of the ambient air quality index AQI,for the different types of air pollution in the park,the different feature processing methods are used to fuse the feature layers of air pollution in the chemical industry park to improve the prediction accuracy of feature fusion.Finally,the decision-making layer fusion of air pollution in the chemical industry park is carried out.On the basis of single model feature fusion,a prediction method based on multi model fusion Stacking ensemble learning is proposed to enhance the reliability of decision-making layer fusion.The specific research work is as follows:(1)Research on the air pollution prediction system framework of chemical industry park based on information fusion.This paper analyzes the feature factors that affect the environmental pollution of the chemical industry park,and puts forward an air pollution prediction system framework based on information fusion.The method and technical system of feature processing and multi model fusion based on information fusion in the framework of air pollution prediction system of chemical industry park are established,which provides technical support for air pollution prediction of chemical industry park.This paper introduces the main structure,function,key technology,application process and advantages of the architecture compared with the traditional air pollution prediction system.(2)Scenario analysis of air pollution in chemical industry park.First of all,the disaster chain of environmental pollution in the chemical industry park is constructed according to the complexity of the occurrence and evolution mechanism of the events in the emergency field,as well as the situation dependence of the events.Then,it analyzes the regular air pollution events caused by the discharge of enterprises in chemical industry park and sudden(fire,leakage and explosion)pollution events,simulates the accident scenarios,prediction of the impact of pollution sources on the environment,the extent and scope of the impact,to provide data sources for feature extraction and pollution response solutions.(3)Air pollution prediction of chemical industry park based on feature fusion.Through feature selection,a method based on feature fusion is proposed to predict the impact of air pollution in chemical industry park,and it is used to deal with the feature variables of air pollution in chemical industry park.First of all,according to the situation of regular air pollution which is affected by multiple features,through the correlation analysis of influencing factors,the main influencing feature variable is temperature,so the fusion weighting method of temperature is established.When neural network is used to predict air pollution,the short-term prediction accuracy of regular air pollution feature fusion can be effectively improved by the weighted correction of cumulative effect.Secondly,aiming at the sudden air pollution situation dominated by single feature,the diffusion results of single feature pollutants are calculated by Gaussian plume model,and then the pollution situation is predicted based on the feature fusion of ant colony neural network.The results show that extracting the correlation between features and weighting the main feature variables can effectively improve the quality of feature layer information fusion,and provide a feature source for decision-making layer fusion.(4)Air pollution prediction of chemical industry park based on ensemble learning strategy.First,aiming at the differences of data observation and training principles of different algorithms and the limitations of single model,based on ensemble learning strategy Stacking is proposed to deal with the air pollution prediction framework of chemical industrial parks in multi source,large volume and multi feature data for decision-making layer fusion.At the same time,according to the trend of the heavy pollution events in the surrounding residential areas caused by the pollutant discharge from the enterprises in the chemical industry park,the reliability of Stacking strategy decision-making fusion is verified.Furthermore,in order to obtain the best prediction effect,the Stacking selection strategy is further explored.The results show that the prediction results of the Stacking strategy are significantly improved compared with the prediction results of single model.The primary model selects the strong learner and the secondary one is the linear model has the best effect.Finally,the control scheme based on DCS is given,which provides the data interpretation of the mitigation measures and air quality control for the government.All in all,based on the information fusion technology,this paper mainly carries out the research on the air pollution prediction of the chemical industry park from the feature layer fusion and the decision-making layer fusion.The high-precision and easy-to-use air pollution prediction system,which is independent of the monitoring station,accurately and reasonably predicted the development trend of the heavy pollution events in the chemical industry park,and realized the reasonable,timely and dynamic control of limited monitoring force.The decision-making body of environmental pollution events can grasp the dynamic change law of air pollution in the park in real time through the prediction model,and check the hidden dangers of emergencies.It is an important means to deal with the critical situation of urban environment,which has important scientific value and practical significance for protecting people's health and promoting social harmony and stability.