Early Fault Prediction of Wind Turbine Drivetrain System

Hong-shan Zhao, Xiao-tian Zhang, and Wei Guo


Drivetrain system, Fault prediction, Temperature, Wind Turbine


Condition monitoring for wind turbines could reduce maintenance costs and improve the operational reliability of wind farm. This paper will propose a new method to monitor the condition of drivetrain system by real-time temperature data. Firstly, the temperature prediction models for the normal behaviour of main bearing, gearbox and high speed shaft bearing are built up by Autoregressive-Moving Average Model (ARMA) and then the corresponding temperature can be predicted. To some degree, the prediction residuals give an indication of actual running state of components. The residual and its distribution characteristics would change when the abnormality exists in some components. Based on this principle, this paper presents a new fault prediction method of the drivetrain system. In the proposed method, the residuals of ARMA are taken as input. An alert is given to remind the operators to check the devices in order to prevent fault and serious accident when the discrimination function value exceeds the pre-set threshold. By the simulating analysis under the normal state and the fault state respectively, it was verified that the proposed fault prediction method can find out the abnormal state of the drivetrain system of wind turbine and achieve the goal of condition monitoring in real time.

Important Links:

Go Back