Research On Fault Diagnosis Method Of Wind Turbine Variable Pitch System Driven By Data/Design

The complexity of structure,the harshness of operating conditions and the diversity of relevant factors,as well as the rapid increase in the total installed capacity of wind power and the continuous increase in the running time of wind turbines,the failure rate has been increased due to the above reasons.Pitch system is one of the core control components of the unit,whose fault becomes the primary reason for the current shutdown of the entire unit.Therefore,it is important to diagnose and predict the fault of the wind turbine pitch system.Firstly,this paper introduces the basic structure and working principle of wind turbine.Then,the basic structure of the pitch system is introduced on the basis of the structure of the wind turbine.The fault principle,fault mode and fault cause are analyzed for two major electrical and mechanical faults of wind turbine pitch system.The reason for each fault was studied,which lays the foundation for the fault diagnosis and prediction.Secondly,due to the complicated reasons of the wind turbine fault and the characteristics of the interaction between the components,this paper base on the SCADA system of the actual operating 1.5MW grid-connected actual wind turbines in a wind farm,combining the experience summary and analysis of pitch system subsystem and structural failure from experts and field operations personnel,the FMEA qualitative analysis method is used to analyze the possible faults of the various components of the pitch system and the impact on the upper system of the entire pitch system.At the same time,FMEA fault risk assessment method is used to evaluate the risk level of all kinds of faults of the pitch system.Finally,through the comprehensive consideration of expert experience,the relevant parameters applicable to engineering practice and fault research are selected.Thirdly,based on the actual SCADA data of the wind turbines,this paper studies two specific pitch angle fault diagnosis of the wind turbine pitch system,adopting data cleaning method to filter out redundant data.After that,the curve fitting and deviation calculation methods are used to perform faults.Through all these methods the wind turbine pitch system fault can be diagnosed.Finally,based on dynamic particle swarm optimization support vector regression(DAPSO-SVR)theory,a pitch angle fault prediction model is established and tested by simulations.It is compared with SVR and BP neural networks by the root mean square error and square correlation coefficient.The comparison verified the accuracy and effectiveness of the DAPSO-SVR fault diagnosis model.Then the sliding window residual statistics method is used to analyze the trend of residual error,as well as evaluate the operating status of the pitch system and predict the fault.