Robust Quasi-Predictive Control of $LCL$-Filtered Grid Converters

This paper proposes a new control strategy for grid-connected LCL-filtered voltage source converters (VSCs). It is realized by cascading a proportional-resonant (PR) controller, which regulates the grid-side current, and a finite-set model predictive controller (FS-MPC), which is responsible for controlling the filter's capacitor voltage and active damping of the resonance. The overall control circumvents the drawbacks of using only the FS-MPC to control the converter, such as steady state tracking error, weighting factor tuning complexity, and need to use long prediction horizons for optimal performance, but it keeps its advantageous properties. Namely, due to FS-MPC's direct manipulation of the VSC switches, the pulsewidth-modulation delay does not exist, while a high sampling rate leads to only insignificant computational delay. As a consequence, the system achieves far less magnitude and phase roll-off in the high-frequency region that allows considerable increase of dynamic performance compared to conventional control approaches. Moreover, the inner FS-MPC-based regulator exhibits a flat frequency response, which indicates that there is no need for designing a dedicated AD. The overall control design procedure is then largely simplified as only the proportional gain of the PR controller needs to be tuned. These properties are proved using a describing function method where a linear approximation of the FS-MPC regulated VSC and an inner LC filter is derived in the frequency domain and integrated together with the model of the PR controller and grid side inductor. The controller has been analyzed analytically and validated through experimental results, where design correctness and robustness to grid-side inductance variations have been tested.