【摘要】 主元分析方法作为一种经典的建模方法已被成功应用到电站锅炉的过程监测中,然而建模时样本数据如何选取的问题并没有得到很好的解决。通过推导主元模型的故障可检测性指标和建模数据样本的方差之间的解析关系,发现主元模型的故障可检测性随建模数据样本标准差的变大而提高。采用2段600MW电站锅炉相同工况下的历史数据分别建立对应的主元模型,通过比较2段数据的方差和模型的故障可检测性,验证了上述结论的正确性。因此建立电站锅炉模型时,应采用稳定工况下波动性较强的过程数据样本建模,以提高模型的精度。更多还原

【Abstract】 Principal component analysis(PCA) is a classical modeling method and has been successfully applied in the process monitoring system for power plant boilers.However,the problem how to select sample data in PCA model building has not been solved.The relationship between fault detectability of PCA model and standard deviation of sample data was derived.The formula indicates that PCA model has better fault detectability using the sample data with bigger standard deviation.Two PCA models were built with on-site data from a 600 MW power plant boiler to validate the relationship.The results show that stronger fluctuant sample data should be selected to build better PCA model.更多还原