The Research Of Gearbox Failure Warning Based On Wind Turbine SCADA Data

In recent years,China’s wind power industry has developed rapidly,and clean energy such as wind energy has received strong support from the state.As a clean,efficient and resource-rich renewable energy source,wind energy has been widely recognized and promoted worldwide.China’s wind turbines assembly machine has increased greatly,but the state monitoring device matched with it has lagged behind.The high maintenance cost caused by the turbine’s failure has caused a large loss of economic benefits for the wind farm.The gear box is a frequent failure component of the wind turbine,which has the characteristics of difficult maintenance and high maintenance cost.Wind Turbine Supervisory Control and Data Acquisition(SCADA)contains a large amount of temperature variable information.Traditional sensor monitoring has the disadvantages of high cost and easy damage.Wind turbine condition monitoring and early warning based on SCADA data will be need by future development.Applying SCADA data for wind turbine gearbox failure warning,and find the turbines’ failure as early as possible,thus reducing and avoiding the development of equipment to more severe damage.Based on the SCADA operating data,this paper establishes the wind turbine gearbox fault early warning model,and supervising the wind turbine operating state by applying this model.including the following aspects.This paper contains the following aspects.1 Introducing the structure of the wind turbine SCADA system and analyzing its function.The wind turbine gearbox fault is classified and the cause of the fault is analyzed.2 An early warning method for wind turbine gearbox temperature based on BootStrap confidence calculation is proposed.The operating conditions and state information of large size wind turbines are complex.The problem of missing alarms and false alarms is existed when a single threshold is used as an evaluation standard for fault early warning.This method implements the spatial division of two-dimensional working conditions through choosing appropriate working condition partitioning parameters and dividing interval scales,which fully considers the common false alarm problem of wind turbine fault warning and the influence of ambient temperature on gearbox temperature.In order to avoid the false alarm problem that is common in engineering applications for fault diagnosis and early warning methods,BootStrap is used to construct the over-limit confidence calculation method,which overcomes the problem of single threshold false alarm in traditional state monitoring.3 A wind turbine gearbox temperature warning method based on XGBoost modeling and Change-Point residual processing is proposed.The low forecasting precision and computational efficiency,the high resource occupancy are the common problems in wind turbines gearbox fault warning based on SCADA data.The eXtreme Gradient Boosting(XGBoost)algorithm combines multithreading,data compression,and fragmentation,which make its higher versatility when processing large data volume samples.The XGBoost algorithm is used to establish the temperature model of the normal working state of the gearbox.In order to improve the reliability of gearbox fault warning and solve the problem of time uncertainty in fault monitoring,a new Change-Point algorithm is constructed.The XGBoost model’s predicteed residual sequence is analysed by using Change-point technology.which can be also used to locate the status deterioration time of gearbox and trace back the cause of state deterioration.