Hierarchical nonlinear model predictive control for frequency support of wind farm

Abstract

Nonlinear model predictive control (NMPC) has been applied to frequency support of a wind farm due to its capability of handling nonlinear dynamic constraints of wind turbines (WTs). However, the centralized NMPC (C-NMPC) suffers from huge computational burden and requires high performance computation facilities due to solution of the embedded optimization problem with a large number of variables. To resolve the problem, the C-NMPC is significantly improved and a new NMPC with a hierarchical structure (H-NMPC) is presented for frequency support of a wind farm. The NMPC optimization problem is decomposed into a coupled sub-problem and several decoupled sub-problems by exploiting its partially separable structure. WT controllers locally solve the decoupled sub-problems in a parallel way and then send sensitivities and active set information to the central controller. Afterwards, the central controller located at the wind farm immediately solves the coupled sub-problem and computes the Newton direction for next iteration. Compared with the centralized computing of C-NMPC, the proposed H-NMPC significantly reduces the computational burden of the central controller and increases the efficiency of NMPC without loss of optimality. The effectiveness, efficiency and robustness of H-NMPC are intensively validated by numerical simulation.