Modelling and estimation of gear train backlash present in wind turbine driven DFIG system

Abstract

The backlash, which is always present in the gears of drive-trains to enable smooth operation, may increase with time due to the wear and tear during mechanical power transmission. Accurate knowledge of the amount of backlash non-linearity is required in such systems for monitoring control applications as well as developing maintenance strategies. In this study, the modelling and online estimation of the backlash existing in the gear train of wind turbine driven doubly fed induction generator (DFIG) systems is presented. The estimation has been performed using the unscented Kalman filter considering the backlash as a parameter and augmenting it as a state into the dynamics. Local measurements of the DFIG system have been used to estimate the backlash. The performance and efficacy of the method have been investigated in different scenarios viz. system parameters, noise levels of the measurements and operating points of the DFIG system. The estimation has also been performed in an environment of the time-varying wind, modelled by Van der Hoven's spectral model. The estimation has been enabled by the bad-data detection and examined in the presence of unknown mechanical parameters of the shaft.