Research And Implementation Of Pollutant Concentration Prediction Model In Specific Regions

With the development of China's modernization,the problem of air pollution has become more and more serious,which has aroused widespread public concern.While accurately predicting the concentration change law of pollutants will help the state to formulate relevant policies and improve air quality.Compared with traditional meteorological prediction methods,the pollutant prediction method based on deep-learning does not need to involve complex meteorological knowledge and is highly versatile.Therefore,it is necessary to study and implement related technologies.In view of the above problems,this thesis uses Python,MySQL,Keras,mathematical statistics,neural network and other technologies to study and implement single-step and multi-step prediction for the concentration of pollutants in specific regions.Also,the corresponding model has been created.In this thesis,the prediction method and accuracy of single-step prediction model is improved.At the same time,a variety of prediction methods for multi-step prediction model are designed and implemented.Also,the test and comparison of models are completed.This thesis first introduces the resear-ch background and subject content,and then briefly introduces the main related technologies used in the topic.In order to complete the research task and display the research results,a pollutant concentration prediction model system was realized through demand analysis and summary design.The system includes a data acquisition module,a data storage module,a model design module,and a data display module.Meanwhile,this thesis gives a detailed introduction to the design details,implementation logic and interaction between modules.Secondly,this thesis introduces the model structure and algorithm in detail.For the single-step prediction model,this thesis realizes the prediction of pollutant data through multi-point raw data based on previous research,researches and compares a total of five algorithm models.There are two models based on mathematical statistics and three models based on neural networks.For the multi-step prediction model,two prediction ideas are proposed in this thesis.The related algorithm models are established,which are machine learning based model and vector matching based model.The thesis introduces the structure of each model and the principle of the algorithm in detail,and gives the conclusion and principle analysis.Finally,this thesis describes the evaluation of the model.In this thesis,the prediction models are tested through several comparison experiments,and the quantitative indicators and prediction images are displayed.The test data effectively demonstrates the conclusion of the proj ect.In addition,this thesis concludes with a summary of the results and shortcomings of the proj ect and points out further research directions.