Model-Free Predictive Power Control of Active Front Ends with Fast Power Variation Update Rate

Abstract

Active front end (AFE) plays an essential role in grid-connected renewable energies. Model predictive power control (MPPC) has become a promising alternative for controlling AFE. However, due to MPPC’s fully model-based concept, model deviations, caused by variations of the filter parameters, will (seriously) deteriorate the system control performances. Model-free predictive power control (MFPPC) is an interesting alternative to improve MPPC’s robustness to parameter deviations. MFPPC requires no parameters in the power prediction stage, resulting in a completely model-free control. However, MFPPC demands a constant update of stored power variations associated with each AFE’s voltage vector. Classical MFPPC has low power variation update rate, since only one state can be updated per sampling period, resulting in degraded control performance. To solve this problem, the proposed MFPPC utilizes two previous power variations and the relevant feeding voltage vectors, to reconstruct all power variations under different voltage vectors. Eight states (corresponding to all AFE’s voltage vectors) can be updated per sampling period, without step-lag. With this new technique, the proposed method has fast power variation update rate. Finally, the proposed MFPPC has been compared with MPPC under nominal and mismatched parameters. Simulation results validate the effectiveness and robustness of the proposed MFPPC.