基于主成分分析法的BP神经网络在气敏传感器阵列检测方法中的应用  

Application of PCA-Based BP Neural Network in Gas Sensor Array Detection

作　　者：徐佳伟[1] 

机构地区：[1]国网上海市电力公司市南集团,上海20000

出　　处：《电力与能源》2015年第2期154-157,共4页Power & Energy

摘　　要：模式识别方法在变压器的油色谱数据处理中起着至关重要的作用。采用主成分分析法与BP神经网络相结合的方式对传感器检测的油中气体进行处理,解决了传统的BP神经网络模式识别收敛速度慢且精度不高的问题。之后将基本的BP神经网络分析法与BP神经网络主成分分析法作了对比。结果表明,该方法能有效解决气体传感器的交叉敏感问题,提高了网络的训练速度和气体的检测精度。Pattern recognition method plays a vital role in transformer oil chromatography data processing. In this research, principal component analysis （PCA） is combined with BP neural network to process oil gas in sensor detection, which solves such problems of traditional BP neural network pattern recognition as slow convergence speed and low accuracy. The comparison between the basic analytical method with PCA of BP neural network shows that the proposed method can effectively solve the cross-sensitivity problem of gas sensor and improve the network training speed and gas detection precision.

关 键 词：气体传感器阵列 交叉敏感 主成分分析 神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]