A Reinforced Model Predictive Control to Eliminate Power Pulsation Residual Components and Achieve High-Purity Grid Currents Under Unbalanced Grids

Unbalanced grid voltages negatively affect the steady-state performance of power converters. To address this problem, we propose a reinforced low-complexity model predictive direct power control (RLC-MPDPC) strategy; this approach overcomes the limitations of existing cost functions in the instantaneous power framework and extended reactive power theory, particularly under severely unbalanced grid voltage conditions. The extended complex power vector is utilized, with its negative conjugate defined as the control variable. Compensation through a feedforward term, derived using Kramer's rule, minimizes the optimization model. Our three-vector-based RLC-MPDPC strategy eliminates residual pulsation components, stabilizing the transient responses of active and reactive powers. In addition, a novel modulation ensures optimal voltage and duration, minimizing power ripples and grid current's total harmonic distortion. The RLC-MPDPC, compared with other multivector-based model predictive control (MPC) methods, provides superior steady-state performance and quicker dynamic response. Performance improvements are demonstrated through comparative experiments.