毕业论文 基于神经网络的空气质量检测.doc

目录
摘要 1
关键词 2
ABSTRACT 2
KEY WORDS 2
引言 2
1 BP神经网络概述 3
1.1 基本原理 3
1.2 BP算法学习过程 4
2 空气质量检测模型的建立 6
2.1样本数据 6
2.1.1收集和整理分组 6
2.1.2输入/输出变量的确定及其数据的预处理 7
2.2神经网络拓扑结构的确定 7
2.2.1隐层数 7
2.2.2隐层节点数 7
2.3神经网络的训练 8
2.4神经网络模型参数的确定 10
2.4.1隐层的数目 10
2.4.2隐层神经元数的选择 10
2.4.3 学习率 和动量因子 13
2.4.4 初始权值的选择 13
2.4.5 收敛误差界值Emin 13
2.4.6输入数据的预处理 13
3 MATLAB实现和结果分析 14
3.1 MATLAB神经网络工具箱的应用 14
3.2 基于MATLAB的BP算法的实现过程 14
3.3训练神经网络 15
4结语 23
致谢 23
参考文献 23









基于神经网络的空气质量检测
05级自动化 臧鹏娟
指导教师 史丽红
摘要：空气质量的好坏反映了空气污染程度，它是依据空气中污染物浓度的高低来判断的。污染物浓度由于受风向、风速、气温、湿度、污染源排放情况等多种因素的影响，使得空气质量问题具有很大的不确定性和一定的复杂性。神经网络作为一种描述和刻画非线性的强有力工具,具有较强的自学习、自组织、自适应能力等特点,特别适合于对具有多因素性、不确定性、随机性、非线性和随时间变化特性的对象进行研究。本文基于神经网络的BP算法，利用MATLAB神经网络工具箱建立了空气质量模型。文中，采用MATLAB 的rand()函数在各级评价标准内按随机均匀分布方式内插生成训练样本和检验样本，利用premnmx()函数对数据进行预处理，调用激活函数对网络权值进行训练，并同其他评价方法比较，取得了良好的评价结果。同时表明此方法具有一定的客观性和积极性。
关键词：BP神经网络；空气质量； MATLAB神经网络工具箱

The detection of air quality based on neural network
Student majoring in automation Zang Pengjuan
Tutor Shi Lihong
Abstract：The quality of air quality reflects the extent of air pollution, which is based on the concentration of pollutants in the air to determine the level of the air. Concentration of pollutants due to wind direction, wind speed, air temperature, humidity, pollutant emissions and other factors, makes the issue of air quality is a great uncertainty and a certain degree of complexity. Neural network description and characterization as a powerful tool for non-linear phenomenon, with strong self-learning, self-organization, the characteristics of adaptive capacity, especially suitable for multi-factor, uncertainty, randomness, non-linear and time-varying characteristics of the object of research. This design bases on the BP neural network algorithm, using MATLAB neural network toolbox to establish air quality model. In this text, using the MATLAB’s rand () function at all levels within the eva luation criteria uniformly distributes random interpolation methods to generate training samples and the samples tested. Then the paper uses premnmx () function on the data pre-processing, and transfers activation function of network weights training and compares with other eva luation methods, and achieved good results which indicate the objectivity and enthusiasm of the design.
Key words：BP neural network; Air quality; MATLAB neural network toolbox