基于主元分析的火电厂生产过程故障检测研究

由于工况变化频繁,使用单一主元模型难以准确描述火电厂生产过程的统计特性,因此应用传统主元分析(PCA)故障检测方法会带来大量的误检。提出了一种适用于火电厂生产过程的改进PCA故障检测方法:首先用K均值聚类分析方法对过程数据进行分类得到各稳态工况下的数据;然后根据分类数据建立主元模型组来描述整个过程;最后在故障检测中对检测样本进行模糊划分,动态计算出与当前工况相适应的主元模型并进行检测。使用现场数据对比研究了传统方法和改进方法在锅炉过程故障检测中的应用情况。结果表明改进方法能适应工况变化,减少误检并提高检测灵敏度。

Fault detection study based on principal component analysis in power plant operation

Due to the misdiagnosis of the traditional fault detection method based on PC A (Principal Component Analysis), an improved PCA-based fault detection method for power plant operation is proposed. The K-mean cluster analysis is used to classify the process data and obtain the data sets under various steady operating condition. The PCM group is established by using the classified data sets to describe the entire process. The detection samples are processed with fuzzy partition in fault detecting process. The principal models suiting for current operating conditions are found out with dynamic calculation and are evaluated. The field data is used to compare the two methods. The results indicate that the improved method is effective.