Generalized stochastic constraint TARMA models for in-operation identification of wind turbine non-stationary dynamics

The identification of the operational Wind Turbine (WT) dynamics is a challenging problem, due to both the complex dynamics and the non-stationarity of the vibration response. In this study the novel class of Generalized Stochastic Constraint (GSC) Time-dependent ARMA models is used for the identification of the non-stationary characteristics of acceleration vibration signals acquired in the tower top of an operating WT, in the fore-aft and lateral directions. The results demonstrate the improved performance of GSC-TARMA models compared to their conventional Smoothness Priors (SP) and Functional Series (FS) counterparts. The obtained models confirm the presence of cyclo-stationary and broader non-stationary behavior in WT vibration. The model based frozen time-varying modes of vibration are analyzed, and the modal components of the vibration accelerations are pictorially presented on the plane defined by the fore-aft and lateral axes.