Research On Fault Diagnosis Of High Pressure System Based On KPCA-IDEPSO-PNN

As an important part of the regenerative system of power plant,the high-pressure heater system can significantly reduce the fuel consumption of boiler.However,the equipment of the high-pressure heater system bears high temperature and pressure in its work,and is prone to various faults.Introducing fault diagnosis technology into the fault diagnosis of high-pressure heating system is of great significance to reduce energy consumption and ensure production safety in thermal power plants.In this paper,the high-pressure heater system of a 1000 MW unit in a thermal power plant is taken as the research object.Aiming at the problems of strong nonlinear characteristics of fault data and low fault diagnosis accuracy of single method,the fault diagnosis of high-pressure heater system based on kernel principal component analysis(KPCA)to extract nonlinear characteristics and differential particle swarm optimization(IDEPSO)to optimize probabilistic neural network(PNN)is studied.The main research contents are as follows:(1)The fault feature extraction method of high-pressure heater system based on KPCA is studied.In view of the strong correlation between fault sample data of high-pressure heater system,multivariate statistical analysis method is used to extract fault features of high-pressure heater system.In view of the deficiency of PCA method in nonlinear feature extraction,KPCA method is introduced to design a fault monitoring experiment of high-pressure heater system based on KPCA and PCA method.The experimental results prove the superiority of KPCA method in nonlinear feature extraction of high-pressure heater system.(2)A fault diagnosis method based on IDEPSO-PNN is studied.The learning speed of PNN method is fast,which is very suitable for real-time data processing.However,its smoothing factor parameters have an important impact on classification performance.PSO is introduced to optimize PNN parameters.Aiming at the problem that PSO is easy to fall into local optimum,the premature particles are mutated,crossed and selected by using the difference idea for reference,and the inertia weight coefficient and learning factor parameters of PSO are improved.A fault diagnosis method based on IDEPSO to optimize PNN is proposed.The experimental results show that the diagnosis accuracy of the new method is greatly improved.(3)Research on fault diagnosis method of high pressure system based on KPCA-IDEPSO-PNN.In order to solve the problem of strong nonlinear characteristics and noise interference in the fault sample data of the high-plus system,the KPCA method is introduced to extract the nonlinear features of the fault samples,and the IDEPSO algorithm is used to optimize the KPCA kernel parameters and PNN smoothing factor parameters.The fault diagnosis method based on KPCA-IDEPSO-PNN effectively solves the problems of strong nonlinear characteristics of fault sample data of high pressure system and low fault diagnosis accuracy of single method.Finally,a fault diagnosis model based on PSO-PNN,IDEPSO-PNN,PCA-IDEPSO-PNN and KPCA-IDEPSO-PNN is established.The experimental results show that the accuracy of the fault monitoring and diagnosis method proposed in this paper is obviously improved,which verifies the effectiveness of the proposed method.