基于遗传神经网络的城市扩张模拟和城市热岛研究

发布时间：2018-04-20 01:15

本文选题：BP神经网络 + 遗传算法　；参考：《电子科技大学》2013年硕士论文

【摘要】：广州市地处珠江三角洲经济发达密集群地带，城市化发展速度极快，由于高强度的政治经济活动造成的城市化扩张现象十分明显，而由此带来了一系列环境问题也逐渐引起了各界学者的高度关注。但是由于城市化进程是一个影响因子众多、变化过程复杂的城市复合系统的变化过程，目前相关的理论研究依然存在较大的局限。为此，基于HJ-1A/1B遥感影像的数据基础，本文将以广州市为例学习和研究城市扩张驱动因子和非线性人工智能算法在城市扩张模拟中的应用。最终通过建立广州市城市扩张动态变化模型，结合分析城市扩张影响下的城市热岛效应，来综合评价广州市生态环境的变化趋势。本文主要研究成果如下：（1）基于城市土地结构的理论知识，本文通过采用缓冲区分析、主成分分析和聚类分析等方法建立城市土地利用类型和经济、政策、人口等因素的相关关系，并依此评价驱动因子在城市变化过程中发挥作用，研究广州市城市化进程的特点。（2）基于面向对象方法，本文提取和分析广州市2009年至2011年的土地利用覆盖特征，并完成城市扩张环境因素的定量化描述，最终利用MATLAB实现基于BP神经网络和遗传算法优化的元胞自动机模型在广州市的土地覆盖变化中的模拟预测。通过对比模拟结果可知：BP神经网络较传统的Geo-Urban元胞自动机模型，，能更好地模拟分布较为集中的耕地和林地等区域，精度可达到70%以上，而对于面积较为零碎的建筑用地区域模拟效果较差；遗传神经网络优化算法则能够较BP神经网络总体提高约5%的模拟精度，部分土地类型的模拟精度能提高至20%；并且遗传神经优化算法还能够充分考虑影响土地变化的各种扰动因素，优化选择驱动因子和缩短迭代次数，对于城市土地扩张的动态模拟研究具有可行性。（3）对于城市扩张影响下的城市热岛效应研究，本文主要运用JMS普适性单通道算法来定量反演广州市的城市热环境，重点研究广州市不透水面、土地覆盖和植被指数与城市热环境的相关性。研究结果显示：广州市建筑用地的扩张导致连续3年城市热效应显著加剧；并且城市平均地表温度与不透水面面积呈现正相关，与城市的植被指数和裸土指数呈现负相关，为此加强城市绿地建设是缓解城市热岛效应，实现城市可持续发展的一项重要手段。
[Abstract]:Guangzhou is located in the Pearl River Delta, where the development of urbanization is very fast. The phenomenon of urbanization expansion caused by the high intensity of political and economic activities is very obvious. As a result, a series of environmental problems have gradually attracted the attention of scholars from all walks of life. However, due to the fact that urbanization is a changing process of urban complex system with many influencing factors and complex changing process, there are still some limitations in the relevant theoretical research at present. Therefore, based on the data base of HJ-1A/1B remote sensing image, this paper will take Guangzhou as an example to study and study the application of urban expansion driving factor and nonlinear artificial intelligence algorithm in urban expansion simulation. Finally, by establishing the dynamic change model of urban expansion in Guangzhou and analyzing the urban heat island effect under the influence of urban expansion, the change trend of ecological environment in Guangzhou is evaluated synthetically. The main research results of this paper are as follows: Based on the theoretical knowledge of urban land structure, this paper uses buffer zone analysis, principal component analysis and cluster analysis to establish the correlation between urban land use types and economic, policy, population and other factors. According to this evaluation, the driving factors play a role in the process of urban change, and study the characteristics of the urbanization process in Guangzhou. Based on the object-oriented method, this paper extracts and analyzes the land use and cover characteristics of Guangzhou from 2009 to 2011, and completes the quantitative description of the environmental factors of urban expansion. Finally, the cellular automata model based on BP neural network and genetic algorithm optimization is realized by using MATLAB to predict the land cover change in Guangzhou city. Compared with the traditional Geo-Urban cellular automata model, the comparison of the simulation results shows that the proportion BP neural network can better simulate the distribution of cultivated land and woodland, and the precision can reach more than 70%. But the simulation effect of the area of the building land is not good, and the optimization algorithm of genetic neural network can improve the simulation accuracy by about 5% compared with the BP neural network as a whole. The simulation accuracy of some land types can be improved to 20 parts, and the genetic neural optimization algorithm can also take into account all kinds of disturbance factors that affect land changes, optimize the selection of driving factors and shorten the number of iterations. It is feasible to study the dynamic simulation of urban land expansion. 3) for the study of urban heat island effect under the influence of urban expansion, this paper mainly uses JMS universal single-channel algorithm to quantitatively inverse the urban thermal environment of Guangzhou, with emphasis on the study of the impermeable water surface in Guangzhou. The correlation between land cover and vegetation index and urban thermal environment. The results show that the urban thermal effect is significantly increased due to the expansion of urban land for buildings in Guangzhou for three consecutive years, and the average urban surface temperature is positively correlated with the area of impervious surface, and negatively correlated with the vegetation index and bare soil index of the city. Therefore, strengthening the construction of urban green space is an important means to alleviate the urban heat island effect and realize the sustainable development of the city.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU984.2;X16

【参考文献】