摘要
以某电厂的监控信息系统(SIS)中的历史数据库作为分析平台,对常规光学火检信号所包含的丰富信息进行深层次挖掘。首先对火检信号数据进行必要的预处理,然后提取出火焰亮度平均值、火焰亮度方差、火焰亮度峰峰值和均匀度等4个特征量,分8种典型燃烧工况对火检强度信号作了大量统计分析,结果表明这些特征量能够反映不同工况下的火焰燃烧状态。将不同燃烧工况下的火焰信号特征值作为神经网络输入,利用该网络的自动聚类功能,分别将它们聚集到不同的区域内。经过验证,这种方法能对燃烧状态稳定性作出有效判断。
The historical database in Supervisory Information System （SIS） of some power plant is used as platform to mine more information of the normal optical flame signals. The flame signals of SIS are preprocessed. Four characteristic components including mean values, variance, peak-to-peak value and uniformity of light intensity are obtained to reflect the difference of flame state under eight typical combustion conditions. Through data analysis, it is concluded that these four characteristic components are evident to reflect the combustion state under different conditions, Then these four characteristic components under different combustion conditions are used as input signals ofneural network. Different output maps of the network are produced corresponding to stable and unstable ,flame signals with its self-organizing function. Verification shows that this method can quantitatively judge the stability of combustion conditions.
出处
《华北电力大学学报（自然科学版）》
CAS
北大核心
2007年第1期37-40,共4页
Journal of North China Electric Power University：Natural Science Edition
关键词
燃烧火焰
特征量
监控信息系统(SIS)
统计分析
自组织神经网络
combustion flame
characteristic components
Supervisory Information System（SIS）
statistic analysis
self-organizing neural network
