- 上一篇：初中数学素质教育以翻转课堂法为例
- 下一篇：MATLAB人工神经网络股票股价模型研究
摘要近年来，随着经济发展、工业建设、城市化进程的加速，大气污染问题日益严重，其中大气细颗粒物（PM2.5）的关注度不断提高。本文利用杭州市基准气候站内PM2.5观测资料，对杭州市2011至2014年PM2.5质量浓度进行特征变化分析。48449
由于PM2.5的时空分布具有地域性，因此利用杭州地区的气溶胶光学厚度(AOD)与PM2.5实测数据，借助时间序列分析方法，发现二者之间具有一定正相关。结合实际建立基于时间序列的一元线性回归模型，得到基于AOD的杭州地区PM2.5反演模型。利用验证数据集对模型精度进行评价，得到其反演精度为25.66%,效果较为理想。将PM2.5反演模型应用于MODIS AOD产品，得到2011至2014年PM2.5年平均时空分布图。PM2.5反演模型可为公共健康服务，还可为制定大气污染防治、节能减排等政府决策提供科学支撑。
In recent years, with economic development, industrial construction, and the acceleration of urbanization process, the air pollution problem is increasingly serious, in which the attention of fine particles PM2.5 enhances unceasingly. Using the PM2.5 data measured by national basic meteorological station in Hangzhou city of China, the mass concentration of atmospheric PM2.5 were analyzed by year, season, month, daily variation characteristics from 2011 to 2014 in Hangzhou.
The spatial and temporal distribution of PM2.5 is regional to a degree. Thus, after analyzing the long time series measured data of Hangzhou, a conclusion is drawn that there is a certain positive correlation between Aerosol Optical Depth and atmospheric fine particles PM2.5. Combined with the actual situation of PM2.5 mass concentration variation, a model of linear regression is established based on time series. Then, the PM2.5 inversion model in Hangzhou area based on AOD is obtained. Using this model and validating data to do the calculation, the MRE between simulated and measured values was 25.66%, indicating that this retrieval model was better. The PM2.5 retrieval model is applied to MODIS satellite products, can be obtained temporal and spatial distribution results from 2011 to 2014. The PM2.5 retrieval model not only for public health services, but also for the development of air pollution control, energy conservation and other government decision to provide scientific support.
毕业论文关键词：PM2.5； AOD； BP神经网络； 一元线性回归； 时间序列； 反演模型
Keyword: Fine particulate matter; Aerosol optical depth; BP neural network model; Linear regression model; Series analysis; Retrieval model
目 录
前言 4
1. 数据与方法 5
2. 研究区概况 5
3. PM2.5质量浓度变化特征分析 6
3.1 PM2.5年均浓度分析 6
3.2 PM2.5季节平均浓度分析 7
3.3 PM2.5日均浓度分析 9
4. PM2.5反演模型 10
4.1 神经网络模型 10
4.1.1 数据分析与预处理 11
4.1.2 BP神经网络模型 12
4.1.3 模型结果与对比分析 14
4.2 线性回归模型 15
4.2.1 一元线性回归模型 15
4.2.2 参数确定 16
4.2.3 模型检验及分析 17
4.3 基于时间序列的线性回归模型 17