Study On Data Driven Fault Diagnosis Technology Of Power Plant Key Auxiliary Eauipment

On-line monitoring and fault diagnosis of power plant auxiliary equipment are important to ensure reliable and stable operation of the generator.Thermal power plant system has nonlinear and strong coupling among different variables properties,so it is difficult to establish accurate analytical models of the equipment to descript the system.Focus on data mining and machine learning,with a primary fan and coal pulverizing system as the main research object,on-online monitoring and fault diagnosis of power plant auxiliary equipment was researched.From the operating characteristics of auxiliary equipment,the overall design scheme of fault diagnosis based on real-time data mining was detailed introduced.In the aspect of fault warning,a fault warning method based on similarity measure and support vector machine(SVM)state estimation was proposed.The effectiveness of the proposed method was verified based on actual data.Taking advantage of solving the problem of high dimension,nonlinear and small scale of(SVM)in pattern,a new method based on Bayesian posterior probability estimation for LS-SVM multi-label fault classification was proposed to deal with the uncertainty between the fault type and fault characteristics in the multi classification problem of the coal pulverizing system.The effectiveness and rationality of the algorithm are verified by an example of the power plant coal pulverizing system.In addition,this paper presents a fault feature extraction method based on existing data and model validation,which is used to enrich the sample knowledge base.In the parameter optimization of LS-SVM,a particle swarm optimization algorithm based on adaptive parameter adjustment was proposed,which can effectively avoid the existing problem,easily falling into local optimum and slow convergence speed.