Research On Fault Diagnosis And Fault Warning Of Wind Turbine Based On Data Drive

Due to the randomness of wind,frequent pitching movements of wind turbines lead to frequent faults in the pitching system.Therefore,research on fault diagnosis and early warning for wind turbine pitching systems has become increasingly important.This paper aims to develop a platform for fault diagnosis and early warning for wind turbine pitching systems in industrial production by utilizing the advantages of industrial internet technology in cloud computing and big data,combined with fault diagnosis and early warning technology.The paper includes three main research contents.Firstly,focusing on the fault of blade zero offset of wind turbines,GH-Bladed wind turbine simulation software is used to simulate blade zero offset under different working conditions.The influence of zero offset on the power generation performance of the operating unit is studied.By judging the forward or reverse deviation of blade zero position through the change of root bending moment,a blade zero position deviation determination model is constructed by fitting the ratio of 1P harmonic amplitude and 3P harmonic amplitude of the axial acceleration of wind turbine.Secondly,based on neural network technology,actual operating SCADA historical data is analyzed to complete fault feature extraction and data analysis.An improved neural network model is trained to construct an AT-LSTM fault diagnosis and early warning model for pitching faults.Finally,industrial internet technology is studied to explore a platform-based practice plan for fault diagnosis and early warning models for wind turbine pitching systems,and to achieve the platformization of the blade zero position deviation determination model and the AT-LSTM fault diagnosis and early warning model.The main contribution of this research is the construction of a blade zero position deviation determination model and an improved neural network-based fault diagnosis and early warning model for pitching faults,which are platformized using industrial internet technology.The platformized fault diagnosis and early warning models can fully utilize the digital advantages of industrial internet platforms to achieve fault diagnosis and early warning,reducing losses and economic costs caused by pitching faults.