Robust Decentralized PI Controller Design for Permanent Magnet Synchronous Generator

This paper considers the problem of controller design for a grid-connected Wind Energy Con- version System (WECS) employing a variable-speed Permanent Magnet Synchronous Generator (PMSG). The system architecture comprises of two back-to-back converters connecting the PMSG to the grid. The variable-speed control of the PMSG using the machine-side converter employs a Multi-Input Multi-Output (MIMO) inner-loop current control. Although there are sophisticated control strategies to address MIMO design requirements, their practical deployment is often hindered by implementation complexity and computational demands for higher-order controllers. In contrast, conventional Proportional-Integral (PI) controllers remain attractive for embedded applications because of their simplicity and easy hardware realizability. However, traditional PI control approaches rely on current-loop decoupling through feedforward compensation and thereby design in the single-input single-output framework, leading to suboptimal performances. To overcome this, a MIMO control design framework is proposed that explicitly considers the coupling in the system model. The controller design employs a two-stage approach: (i) designing the decentralized inner-loop current controllers in a MIMO framework and then (ii) the outer-loop speed controller design. The structured PI controller requires the controller to be designed in a static output feedback framework. Simple weight functions in $H_{\infty }$ robust control framework are chosen for the current-loop controller design to simplify the output feedback design criterion. While achieving robustness to disturbance rejection, the simultaneous regional pole placement criterion is used to ensure well-damped transient responses. Experiments on a laboratory-scale PMSG-based WECS validate the effectiveness of the proposed control design method that outperforms conventionally designed PI controllers in both transient and steady-state performances.