Research On Wind Turbine Condition Monitoring And Fault Diagnosis Based On SCADA

With the rapid development of global social economy,the problem of energy crisis and environmental protection have become the primary factor that restricting social development.Many countries have paid more attention on wind power due to its green and reproduction and wind power has developed quickly.However,along with the enlargement of the scale of wind farm,the incidence of the accident of wind turbines appears an upward trend significantly.Taking account of the difficulty of replacing the components and the high maintenance cost,it is very important to research the condition monitoring and fault diagnosis technology on wind turbines to ensure it operating safety and decrease the fault rate.By introducing the basic structure and working principle of wind turbine and its key components,this thesis analyzes the fault reasons,the data signal and data preprocessing method of SCADA(supervisory control and data acquisition)system are listed,which lay foundation for the condition monitioring and fault diagnosis.The wind turbine has complicated fault causes and interaction among the components.Aiming at this problem,a method of condition monitoring of wind turbine based on ANFIS(adaptive neural fuzzy inference system)is proposed.The data are derived from the actual SCADA system of 1.5 MW wind turbines from a wind farm located in HeBei Provice.The thesis digs out the relationship between the 20 groups of input and output,the normal operation data is used as the training data,20 sets of ANFIS models are established to get the probability distribution of the forecast error,and then determine the running status of the wind turnbine.Based on the accuracy of these models,11 characteristic parameters which can best represent the fault information of wind turbines are extracted,which can provide theoretical support for fault diagnosis.In order to obtain the faults type of wind turbine,a new fault diagnosis method based on KPCA(kernel principal component analysis)and FLS-SVM(fuzzy least square support vector machines)is proposed.KPCA is used to extract the characteristic parameters with the most information of fault types so that ruduce the dimensionality of the sample data.Then FLS-SVM use the ruduced dimension data classifies several faults and the normal state of the key components to achieve the purpose of fault diagnosis.The KPCA and FLS-SVM are used to establish the gearbox fault diagnosis model,pitch system fault diagnosis model and generator fault diagnosis model.The accuracy of the model have been demonstrated by comparing the diagnostic correct rate with the OneToOne Least Squares-Support Vector Machines and ANFIS.