Research Of Fault Detedtion And Diagnosis Method For Power Plant Equipment Based On Multivariate Statistical Analysis

Unit equipment condition detection and diagnosis plays an important role in ensuring the safety and economy of power plant unit operation.This paper mainly research power plant equipment intelligent diagnosis method based on multivariate statistical analysis,detailed work are listed as below:1)This paper theoretically analyzes the problem of "smearing effect" in fault isolation based on traditional contribution plot and reconstruction contribution plot in multivariate statistical methods,and proposes a least angle regression reconstruction principal component analysis.This method transforms the combinatorial optimization into a regression on the problem of fault isolation,and uses least angle regression of fusion monitoring indicators to achieve a sparse search for potential fault combinations and finally combination Fisher-Jenks classification to realizes fault isolation.2)Based on the dynamic characteristics and the non-Gaussian characteristics of power plant,an independent principal component analysis method based on dynamic feature extraction is proposed,which use ICA to realize the fault monitoring modeling of the static nonGaussian residual part after the dynamic change of data is extracted.Meanwhile,after fault detection,the fault isolation method based on Lars reconstruction is further improved to realize the fault isolation under dynamic conditions.Finally,the typical TE process and high-pressure heater system are used to verify the method.3)In allusion to problems of few sample,incomplete data and unbalanced distribution in the actual situation,a fault detection method based on principal component analysis of geodesic flow kernel for domain adaptation and unsupervised non-linear system fault detection is proposed.The method applied the data from similar fields to other models by transfer learning,which can significantly improve the model accuracy.4)Based on the previous research methods,a system platform for intelligent monitoring and intelligent diagnosis of the thermal power plant is established.And the detailed system development scheme is given from the aspects of system principle,architecture design and function realization.