【作者】 蒋志方；

【导师】 孟祥旭；

【作者基本信息】 山东大学 ， 计算机软件与理论， 2011， 博士

【摘要】 由于工业化、城市化进程的快速发展,城市的生产、消费规模不断扩大,导致城市能源、交通规模持续扩大。我国是发展中国家,能源结构不尽合理,以燃煤作为电力、热力等的主要燃料。因此,以可吸入颗粒物、一氧化碳、二氧化硫、氮氧化物等为主要污染物的大气环境污染问题日趋严重,对资源、环境与人民生活和社会经济带来巨大的冲击,严重威胁着可持续发展的基础。开展空气质量监测、预测与数据分析与可视化的研究可以全面掌握城市空气污染源的排放数据和各种空气污染物在不同空间区域内的浓度数据,可以对影响城市空气质量的因素有所了解和把握,掌握城市空气质量在时间和空间维的变化发展趋势,对城市规划与建设、污染控制、环境管理、公共事业发展均有重要的理论意义与实用价值。“十五”规划和“十一五”规划期间,我国各省市都开展了大规模的基于互联网、物联网的基本覆盖城市区域的环境空气质量监测系统和重点空气污染源监测监控系统的建设。对影响城市环境空气质量进行实时监测监控,获取了大量环境空气质量和重点空气污染源排放的基础数据和实时动态监测数据。如何充分利用这些海量数据,来对城市环境空气质量的各种指标的进行科学的评价、对环境空气质量变化趋势进行分析预测,研究城市环境空气质量与各种空气排放源之间的时空相关性等方面的工作有待进一步开展。在国家863计划项目《大规模动态场景中海量数据的远程可视化技术及系统实现(2008)》、山东省科技攻关项目《城市空气污染应急仿真与可视化研究及系统实现(2008)》和山东省科委《山东省可持续发展科技示范项目-济南市环境空气质量及空气污染源监控、预警技术及监控网络的研究开发(2002))以及济南市科技攻关项目《济南市烟气污染源在线监控系统研究开发与示范(2003)》等科研项目的支持下,本文以山东省济南市的环境监测系统为基础,围绕城市环境空气质量监测监控与预测模型的构建和空气质量数据的可视化等课题,重点开展了基于神经网络的城市环境空气质量预测模型研究、城市环境空气质量预报数据可视化方法的研究、城市空气污染源排放扩散模拟仿真方法的研究、以及城市空气污染源排放监测数据与空气质量的相关性分析等方面的研究。本文研究了以区域内空气污染源排放监测数据为输入,区域内的环境空气监测站的污染物浓度数据预测值为输出的人工神经网络预测模型的构建方法。提出了一个基于粗糙集理论的BP神经网络预测模型,对指定区域内的环境空气污染源排放原始监测数据进行属性约简、构造人工神经网络隐层的神经元节点、确定节点间的连接权值和网络初始拓扑结构,通过BP算法迭代对数据进行训练,求出网络的各种参数,完成预测模型的构建。提出了一个基于资源分配网络的神经网络预测模型,利用RAN神经网络的距离准则和误差准则,进行隐层节点的动态生成和参数调节,生成能满足误差要求的最小神经网络结构,避免了网络中隐含节点个数和初始网络参数难以选取的缺点,这两个模型,都比经典的BP神经网络有更快的网络的训练速度和更高的预测精度。本文对城市区域内的环境空气质量预报数据的可视化方法的进行了研究。运用径向基函数对给定区域环境空气质量预报数据中污染物浓度值在时间和空间维度上进行插值的方法,生成了在某一指定空间层次上的空气污染物浓度等值面；采用Marching Cubes算法在城市空间区域维度上进行污染物浓度数据的三维等值面绘制；研究了如何实现城市区域空气质量预报数据中污染物浓度在空间的整体弥撒效果,首先定义了在空间网格中的污染物粒子,并以对空间区域的所有格网中的污染物浓度值数据进行数据预处理、转换,利用光线对污染物的散射效果,实现对空气污染数据动态模拟,绘制出了类似雾霾的视觉效果；研究了基于高斯点源扩散模式的城市空气污染源排放监测数据与城市空气质量监测数据的相关性及可视化方法,在给定的气象数据基础上,计算出一组选定污染源的污染物排放监测数据对城市各环境质量监测站点的污染物浓度值监测数据贡献率,建立了相关性分析矩阵,对两者的相关性进行了分析,并研究了相关性的表征方法。更多还原

【Abstract】 As the rapid development of industrialization and urbanization, the city’s production and consumption have been expanding, it led the urban energy, transportation to grow continue sly. China is a developing country; the energy structure is not reasonable, coal is as the main fuel of the electricity, heat. Thus the atmospheric environment pollution. problems that the repairable particulate matter, carbon monoxide, sulfur dioxide, nitrogen oxides in the air as the main pollutants have become more serious and bring a great social and economic impact on resources, environment, people’s lives and socio-economic and that will also threat meanwhile the foundation of the sustainable development.To develop the air quality monitoring system, forecasting, data analysis and visualization of urban air quality and air pollution sources, will let us master the emissions data of the air pollution sources and the concentration data of a variety of air pollutants at different spatial area.That also can make us understand the factors which impact the urban air quality and grasp the air quality change’s trend in time and space dimension. It has very important theoretical significance and practical value in urban planning and construction, pollution control, environmental management and public undertakings.Many provinces and municipalities in China have carried out the large-scale, urban areas covered ambient air quality monitoring systems and the key sources of air pollution monitoring and control systems by in "Tenth Five-Year Plan" Planning and the "Eleventh Five-Year Plan" period. By using these systems, we can obtain a large number of basis data and the real-time monitoring data of the ambient air qualities and key air pollution sources. How to take advantage of these huge amounts of data, to study and evaluate the urban environment air quality indicators, to analyze and forecast the ambient air quality trends, to study the spatial and temporal correlation between air quality and air emissions of various air pollution sources scientifically will be the important work recently.With the supported of the National High Technology Research and Development program of China, that the title is "The Massive Amounts of Data in Dynamic Scene Visualization Technology and A Remote System to Achieve" and the Scientific and Technological Project of Shandong province that the title is "Urban Air Pollution Emergency Response Simulation and Visualization and System Implementation" and the project of Shandong Science and Technology that the title is "The Sustainable Development in Shandong Province Science and Technology Demonstration Projects-Jinan Ambient Air Quality and Air Pollution Monitoring, Early Warning and Monitoring Network Technology Research and Development", and the project scientific and technological of Jinan City that the title is "Jinan Smoke Pollution On-line Monitoring Systems Research, Development and Demonstration" and other research projects, this paper focus on the urban environment based on the neural network prediction model of air quality, urban air quality forecasting data visualization methods, and the reverent of the urban air pollution emission monitoring data and the air quality monitoring data based on Jinan’s environmental air quality monitoring system and Jinan’s air pollution emission monitoring and control systems to build some predictive models based on the neural network,to implement the visualization of the data and to get the correlation of the air quality monitoring data and the air pollution emission monitoring data.This paper studies the construction method of artificial neural network prediction model with the regional air pollution emissions monitoring data as input, ambient air monitoring stations within the region of the predictive value of pollutant concentration data as output. It proposed a BP neural network prediction model based on a rough set theory, which reduct attributes in a selected area of the original ambient air pollution emission monitoring data, construct of the hidden layer neuron node, of the neural network, determine the initial connection weights between neural nodes and make the initial network topology, obtain the network parameters through training by iterative BP algorithm and complete the construction of predictive models. The paper also proposed a neural network prediction model based on the RAN (Resource Allocation Network), it can meet the minimum requirements of neural network error structure using neural network RAN distance criterion and error criterion, and can generate dynamically the hidden nodes of the neural network and adjust the parameters dynamically. These two models have faster network training speed and higher prediction accuracy than the classical BP neural network. How to visualize the city’s ambient air quality forecasting data has been studied in this paper. We generate the air pollutants concentration isosurface in a specified level of the space using the interpolation method of RBF(Radial Basis Function) for a given air quality forecast regional pollutant’s concentrations in the data values in time and space dimensions. Using Marching Cubes algorithm we render the three-dimensional isosurface of pollutants concentration data on the regional dimension in the whole city space. Through the study of regional air pollutant concentration data in spatial area, we establish a pollution concentration data grid cube and how to render the overall diffusion effect for the certain pollutants concentrations data in the specified space area of the urban also be studied in this paper. Firstly, an abstract particle associated with the certain pollutant concentration data of every grid cube in the space is defined, Then pretreated all pollutant concentration data in every grid to form a normalized data set to be rendered, use the abstract particle to construct the corresponding relationship between pollutant concentration of a grid and the particle radius, After that we use the Rayleigh and Mie scattering models to render the all particles in the space and get the Visual effects as haze. The rendering results show that the rendering of the translucent material in spatial region has better visual effect. The paper studies the correlation of the urban air pollution emissions monitoring data and the urban air quality monitoring data and how to visualize it based on Gaussian point source dispersion model. Under the certain meteorological conditions we calculate the contribution rate by selecting some sources of pollutant emissions monitoring data to the air quality monitoring sites of the urban. Established a correlation matrix, analyzed the correlation of two data sets. The paper also studied the visualization technology to visualize the correlation relation of emissions monitoring data of the pollutant sources and that of the air quality monitoring sites.更多还原