温室湿度环境的主成分分析人工神经网络建模研究  

Greenhouse Air Humidity Modeling Based on Principal Component Analysis and Artificial Neural Network

作　　者：何芬[1] 马承伟[1] 

机构地区：[1]中国农业大学农业部设施农业生物环境工程重点开放实验室,北京100083

出　　处：《上海交通大学学报（农业科学版）》2008年第5期428-431,共4页Journal of Shanghai Jiaotong University(Agricultural Science)

基　　金：北京市教育委员会共建项目建设计划资助(XK100190650)

摘　　要：实测温室内影响空气湿度的环境因子组成数据样本,对数据样本进行主成分分析,提取出影响温室湿度的4个主要成分,讨论提取的主成分与原始过程数据样本间的关系。以采用主成分分析后的数据样本作为神经网络模型的输入变量,模型模拟值和实测值之间的相关系数R^2为0.8842。以±0.1作为模拟相对误差,命中率达到85%。用训练后的网络模型对20组未参加建模的样本数据进行模拟.均方根误差为1.6745,优于回归方程法的4.4349。基于神经网络模型,运用敏感性分析法对影响湿度的各因素进行重要性分析和排序,得出各影响因素的重要程度依次为室内温度、室外湿度、室外温度、保温帘展开度、室外风速、室外太阳辐射照度、天窗开窗角度和侧窗开窗角度。The environment factors influencing air humidity were measured as data samples in greenhouse. Through the principal compo- nent analysis of data samples, 4 main factors were extracted. The relationship between the main factors and the original data was discussed. These selected principal component values were taken as the input variables of neural network model to simulate the humidity, and the determination coefficient was 0.8842. With simulating relative errors ranging ±0.1, the hit rate reached 85%. The trained neural network model was applied to simulate the indoor humidity with twenty groups of data samples which had not been used to establish the neural network model, and the root mean square error was 1.6745, better than the root mean square error of regression function, which was 4.4349. Based on the neural network model, the influencing factors importance of inside air humidity was analyzed. The inside temperature is the most important factor, and the next is outside humidity and temperature, open ration of sunshade curtain, wind speed, solar radiation, and open angle of top vent and side vent.

关 键 词：温室 湿度模拟 主成分分析 人工神经网络 

分 类 号：S625[农业科学—园艺学] TP183[自动化与计算机技术—控制理论与控制工程]