Data-driven Identification for Frequency Regulation Model of Wind Farms

Abstract

The increasing penetration of the renewable energy sources (RES) brings new challenges to the frequency security of power systems. Therefore, qualifying the frequency regulation capacity of RES is indispensable. However, the diversity of the control strategies deployed for wind turbines makes it hard to explicitly formulate the frequency regulation model. In this paper, Sparse Identification of Nonlinear Dynamics (SINDy), a data-driven method is employed to identify the aggregated frequency regulation model of wind farms, which can be represented as an equivalent governor transfer function. Compared with physical modelling approach which needs to know all the control parameters in advance, this data-driven method can learn the model characteristics of dynamic systems from the historical data using a universal framework. Numerical simulations indicate that the proposed general frequency regulation model identification framework is applicable to wind farms with high accuracy.