Research On Abnormal State Recognition Method Of Wind Turbine Based On SCADA Data

With the economic development and scientific and technological progress,the global wind power industry continues to show strong growth.Wind turbines are increasingly being deployed in remote onshore and offshore areas.Due to the harsh environment of wind farms and other factors,wind turbine components are prone to failure.Failure of wind turbine components can lead to expensive repairs and longer repair times.Therefore,abnormal state monitoring and fault diagnosis of wind turbines are very important.Wind turbines are equipped with a complete supervisory control and data acquisition(SCADA)system.The SCADA system records a large amount of wind turbine operating status data.These SCADA data include the normal operation of the equipment,shutdown,abnormalities and failures.Therefore,SCADA data is effectively extracted and used to monitor the state of the wind turbine and diagnose faults.The normal operation and maintenance of this field have important practical significance.Based on the SCADA data of wind turbines,this paper monitors and recognizes the abnormal state of wind turbines.The specific research contents are as follows:(1)The correlation between wind speed and state parameters of other 16 wind turbines is studied.First of all,the SCADA data of the wind turbine is subjected to data screening and standard normalization,and the corresponding judgment conditions are selected to remove the abnormal data and the standby data.Finally,the data of the 17 state parameters are processed by the data preprocessing method of regional probability weights Processing analysis was carried out,and the correlation between wind speed and other 16 state parameters was qualitatively studied.(2)Method for selecting state parameters of wind turbine based on Relief algorithm.According to the principle of the Relief algorithm and the process of calculating the weights of characteristic parameters;the data for the past year is evenly divided into two sample sets,and the weight values of the 17 state parameters are calculated and iterated in turn;After 17 state parameters are sorted in sequence,8feature parameters are selected.Finally,the results are compared with the data preprocessing method of regional probability weights.(3)On the basis of studying the correlation between SCADA data,a health status index is introduced to judge the abnormal fault condition of the wind turbine,and an on-site verification test is performed on this health status index through examples.The results show that this method not only has the potential ability to monitor the initial wind turbine gearbox failure,but also has the ability to track its further deterioration.Finally,it is verified that the state parameters selected based on the Relief algorithm can more fully reflect the validity and accuracy of the health state index.