Research On Identification Of Abnormal State And Early Warning Method Of Defect In Wind Turbine

In recent years,with the increasing of the installed capacity of wind turbine,the safety and reliability in the use of wind turbine become very important.The identification of abnormal state and the early warning of defects in the wind turbine are the two main contents of its safe and reliable operation.This thesis focuses on the methodological study of these two contents,which has the important application value to ensure the safety and to improve the economy of the wind turbine.Firstly,the state parameters are classified based on the SCADA system.The cross-correlation analysis of the state parameters and the auto-correlation analysis of the same state parameters is carried out by the Pearson correlation coefficient,which select the data with strong correlation and the length of time with significant correlation respectively.The conclusion of correlation analysis provides the basis for the identification of abnormal state and the early warning of defect in the wind turbine.Secondly,the model is built for the identification of abnormal state in the wind turbine.From the point of view of optimizing the prediction model,the combination prediction model is constructed by using a single prediction model,and then combining the linear combination model with the least square sum of prediction error.The superiority of the combined forecasting model is verified by error precision analysis,and the abnormal state identification model of wind turbine is built by this model.The residual value is analyzed by normal distribution,and the threshold of abnormal state of wind turbine is set,which can be used to identify the abnormal state of wind turbine.However,the residual threshold method based on statistics has errors,and the hidden Markov model has higher accuracy,which serves for the early warning of defects of the following.Finally,the early warning of defects in the wind turbine is studied.Whenthe number of samples is not enough,the residual threshold method based on statistics has the characteristics of large error.From the point of view of state transition,Hidden Markov Model is used to build the model of early warning of defects.Firstly,the Markov process of wind turbine in the process of state transition is analyzed.The model of early warning of defects in the wind turbine is built based on the Hidden Markov Mode,where we compare the probability value P value under the HMM of each fan in each period,and the time period in which the maximum value appears corresponding to the time period when the state changes.The validity of the model is verified by an example in the end.