Research On Neural Network Based Fault Diagnosis Technology Of Large Wind Turbine Yaw System

In the operation of wind turbines,the failure of the yaw system has caused great hidden dangers to the safe operation of the wind turbines.It will be shut down for a long time by the difficulty of maintenance,which seriously affects the safety and economy of the wind turbine.As the structure of wind turbines becomes more complex and management becomes more refined,the various types of data generated by wind farm are increasing day by day.These data contain a large amount of operation information,and most of the existing SCADA systems can only monitor the power production process in real time.With information storage,it is impossible to effectively predict the failure type of wind turbines,which causes a huge waste of massive SCADA data resources.Compared with the additional installation of sensors,the use of the wind turbine’s own SCADA data for fault analysis means no need to add sensors,can reduce the sensor installation process and the lower cost.In this dissertation,a combination of theoretical analysis and actual engineering data are used to explore and model the big data environment generated by the actual wind farm SCADA system,a Diagnostic model of yaw system is established,and implement a wind turbine yaw system fault diagnosis software system.First,according to the operating principle of the wind turbine and its yaw system,the degree of fault system is used to use the yaw angle error,and combined with the SCADA system parameters,based on MATLAB,the nuclear density-mean method is used in normal status and fault.Status data is filtered.After normalization,the ReliefF algorithm is used to sort the SCADA parameters by weight.Seven SCADA parameters that can better reflect the operating conditions of the yaw system are selected,and the corresponding six fault feature indicators are extracted.Then,the neural network fault diagnosis model is developed based on SCADA data,and the extracted fault characteristic index is used as the input of the diagnosis model to diagnose the normal,mild,moderate and severe faults of the yaw system,and the position of the faulty components.The research results show that the error accuracy after the neural network diagnosis model training meets the diagnosis requirements and can accurately diagnose the yaw system faults.Finally,based on the Lab VIEW,it is designed to develop a set of data display,analysis,and saving of a wind turbine system fault diagnosis software.Based on the SCADA data of the actual wind farm,the software has functionally verified.The results show that the wind turbine system fault diagnosis software developed in this dissertation has the characteristics of the diagnostic conclusions,interface-friendly,clear and intuitive display icons,and has certain engineering applications.