Fault Detection of a Wind Turbine's Gearbox, based on Power Curve Modeling and an on-line Statistical Change Detection Algorithm

Abstract

An early model-based fault detection was developed, based on the wind turbine's power curve to detect the degradation (faults) in gearbox efficiency, resulted from the existing mechanical losses (torque losses) through the low-speed shaft and the high-speed shaft, then to assist in implementing predictive maintenance strategy. The detection was performed on two levels; the first level represents a slight and progressive degradation in the gearbox efficiency. The other one represents a radical (abrupt) degradation in the efficiency. Artificial SCADA data for different measurements (wind speed and active power) in both fault-free and faulty operating modes were generated using a FAST-NREL simulator. The wind turbine power curves' parameters were estimated, then power residuals were generated from each power point. Finally, an on-line CUSUM statistical change detection algorithm was used to evaluate and detect small changes in power residuals generated from the model. The presented fault detection system successfully detected faults in both detection levels under realistic wind turbulence and with a fault magnitude of 2% efficiency degradation for the progressive degradation level.