State Analysis And Prediction Model Of Wind Turbine Gearbox Driven By Data

In view of the hidden trouble caused by the inherent instability of wind power to the safe operation of power grid,this paper builds a data-driven state analysis and prediction model for the gearbox of wind turbine.80%of the shutdown faults of domestic wind turbines are due to gearbox problems,which will lead to the decline of power generation.Therefore,effective fault prediction is required for gearbox so that wind farms can take timely response measures to prevent further deterioration of the faults.And 34%of the gearbox failures are caused by lubrication failure,so it is necessary to select the lubricating oil matching with the gearbox.There are many types of lubricants,so accurate,reliable and efficient oil product type identification technology and gearbox fault prediction technology can reduce and avoid the development of wind turbine equipment to more severe damage,and then improve the utilization efficiency and power generation of the unit.Based on the above problems,based on the SCADA historical data of wind turbine gearbox and the infrared spectrum data of lubricating oil as information carrier,this paper establishes the wind turbine power generation prediction model,the gearbox fault prediction model and the lubricating oil additive type identification model respectively through machine learning algorithm simulation and optimization.The effectiveness of the method is demonstrated by comparing with the measured data.The main work contents are as follows:(1)Aiming at the problem of power generation prediction of wind power system,the LSTM power generation prediction model was built based on Python platform based on the SCADA historical data of wind turbine gearbox.In the model,the first-order difference and normalization methods were used for data preprocessing to reduce the influence of sample sampling deviation on the analysis results.The Bat algorithm(BA)was introduced into the LSTM hyperparameter optimization,and a BA-LSTM based model parameter adaptive optimization algorithm(BA-LSTM)was constructed to improve the accuracy of the power generation prediction model.(2)For gearbox fault the wind generator outage capacity degradation problem,in this paper,the method of principal component analysis(PCA)and promote tree(Xgboost),the combination of extreme gradient PCA-Xgboost gearbox oil temperature early warning model,and combining with the principle of statistical process control(SPC),through the change of the residual oil temperature,discovered the potential failure of gear box,this method can effectively avoid the lack of the influence of the gear box failure sample,and through the example analysis proves the validity of the model.(3)In order to solve the problem that 34%of the fault shutdown of wind turbine gearbox is caused by lubrication failure,it is necessary to select the matching lubricating oil for different gearbox,and then accurately identify the type of lubricating oil.Based on the theory of infrared spectral analysis technology,lubricating oil species identification combination forecast model was constructed based on the Python platform,and to improve the model recognition efficiency and accuracy,the base model combined with the optimization algorithm of input band filter,remove the band part of high correlation,achieve denoising,comparing the output results of different processing methods,The optimal results of the model were obtained,and the band interval used in the model was recorded.The model was compared with the characteristic peaks known by traditional methods,and a simulation model was provided for the scientific research on lubricating oil performance to guide the research direction.