| This research is a part of the project funded by the Importation and Development of High-Caliber Talents Project of Beijing Municipal Institutions,"Research on Real-time Monitoring System for Urban Ambient Air Parameters based on Wireless Sensor Network"(CIT&TCD20130320).Beijing is suffering from serious air quality problems at present, particulate matters based pollutants are the key to increase urban air pollution. In order to master pollutants concentrations and their relationships, exploring the reasonable evaluation and prediction methods, analyzing the variation characteristics of pollutants, have important implications for revealing laws of air quality and their control in Beijing. By taking main pollutants in Beijing as research objects, regional environment and micro-environment as research scales, different seasons and time intervals between March2013to February2014as research periods, data mining analysis of air pollutants and their sampling experiments were researched systematically. The main works completed are as follow:(1) Current main pollutants concentrations in ambient air and their relationship in Beijing were analyzed. By constructing complete time series data based on muti-point monitoring stations and using mathematical statistics method, frequency distributions, correlations and hourly variation laws of main pollutants concentrations in six districts of Beijing were analyzed. Pollutants level and variation laws, daily and hourly variation characteristics of main pollutants in different seasons of Beijing were then obtained.(2) Comprehensive evaluation methods based on classification and clustering algorithms were developed and improved. Using improved clustering algorithm, BP and RBF neural networks on the basis of AQI evaluation method and data mining theory, comprehensive evaluation results of air quality in Beijing urban were obtained. By analyzing the differences of the results, the comprehensive evaluation index system based on human health was explored.(3) Prediction model was improved and a new method to predict urban air pollutants concentrations was proposed. Inputs of mechanism model based on meteorological factors were modified firstly, then optimized prediction model was selected based on multiple linear regression, BP neural network and support vector machine. Finally, a non-mechanism model based on fuzzy time series and support vector machine was proposed, it solved the problem of unstable prediction results caused by the incomplete influencing factors considered.(4) Temporal-spatial distribution characteristics of particulate matters in Beijing area were proved. According to particulate matters concentrations variation with time scale and their regional distribution, an effective and high dimensional interpolation algorithm was used to research the distribution of average concentrations of particulate matters in Beijing area. Temporal-spatial distribution of particulate matters and their partial pollution characteristics were simulated and the interpolation results were validated. On this basis, the possible influencing factors of particulate matters and their sources were discussed.(5) Indoor and outdoor and vertical variation laws of particulate matters in micro-environment were researched and their physical and chemical experiment was developed. Through discussing concentrations variation with height of particulates with different sizes and their relationship between indoor and outdoor, pollution level of particulate matters and their permeability were firstly analyzed. Then particulate matters were classified by topographic observation of electron microscope and their chemical elements was analyzed by EDX. Finally, analysis methods of enrichment factors and cluster were used to identify contribution degree of pollution source of particulate matters and their possible chemical composition. |
PDF Full Text Request