Model Predictive Control for Grid-Connected Current-Source Converter With Enhanced Robustness and Grid-Current Feedback Only

Abstract

The current-source converter (CSC) features the second-order characteristic, and multivariable feedback is necessary to suppress the filter resonance. Model predictive control (MPC) has shown advantages in terms of the multiple-input–multiple-output (MIMO) system. This article contributes to the MPC scheme applied to CSC from two aspects. First, based on the typical one-step prediction, existing MPC with both grid-current and capacitor voltage feedback has been applied to mitigate the resonance issue but lacks the investigation and improvement of system robustness to parameter mismatch, especially in such a second-order system. Thus, the dual-disturbance observer is proposed, where both the additional capacitor voltage reference and the lumped input disturbance caused by parameter mismatch are estimated online. As a result, the grid-current reference tracking performance can be improved under mismatched parameters. Second, to reveal the resonance principle with different state feedback types, the tracking process of the grid current is intuitively analyzed in the time domain. The MPC scheme is further assessed by increasing the prediction horizon length, and the results show that the CSC can work well with only gird-current feedback. In this case, both the weighting factor and corresponding reference estimation for the capacitor voltage can be avoided, thus significantly simplifying the design stage and reducing the effect caused by the parameter mismatch. Experimental results verify the effectiveness of the proposed scheme and perspectives.