Wake Modelling of Horizontal-Axis Wind Turbines Using Sparse Identification of Non-linear Dynamics (SINDy)

Performing accurate modelling of measurement data from a non-linear dynamic system is a challenging task prone to many idealized assumptions and sensitivity to noise resulting in non-reliable and inaccurate models. Several data-driven approaches, such as genetic programming, artificial neural network, and others, are capable of producing dynamic models with extended relations. However, such models often carry long and non-intuitive Equations, it thus becomes difficult to derive a linear representation of the non-linear dynamics. The paper proposes an innovative application of the SINDy algorithm for deriving a dynamical model from the measurement data obtained using an optical method called Particle Image Velocimetry (PIV). The methodology has been used to identify and model the non-linear dynamics involved in wind turbine wake. The application of the model analogy will have a great implication in wind farm modelling, where economic constraints related to available space and the turbine installation is a prime concern. The study helps in determining the optimal distance of placement of turbines to minimize the effect of wake interference and enhancement of performance.