Robust Predictive Control of Grid-Connected Converters: Sensor Noise Suppression With Parallel-Cascade Extended State Observer

Abstract

Model predictive control (MPC) is a constrained optimization control method with superior performance than linear methods for multi-variable and multi-objective control of power converters. Nonetheless, its performance is limited by model uncertainties and measurement noise. This study tackles this challenge by proposing a new hybrid parallel-cascade extended state observer (PC-ESO) with two key advantages, viz.: (i) higher disturbance rejection than the conventional linear ESO and cascade ESO (CESO) at low bandwidth, and (ii) better noise suppression than the conventional ESO. PC-ESO's time-domain structure, and comprehensive frequency-domain analysis are presented. Furthermore, PC-ESO is applied to improve the transient disturbance rejection of CESO through a novel structurally-adaptive ESO (SAESO) algorithm. The proposed SAESO provides both high-frequency noise-suppression and better disturbance rejection than CESO and cascade-parallel ESO. Finally, the proposed methods are experimentally validated by the model-free predictive control of a grid-connected power converter.