Improved three-vector based dead-beat model predictive direct power control strategy for grid-connected inverters

Abstract

Since only one inverter voltage vector is applied during each duty cycle, traditional model predictive direct power control (MPDPC) for grid-connected inverters (GCIs) results in serious harmonics in current and power. Moreover, a high sampling frequency is needed to ensure satisfactory steady-state performance, which is contradictory to its long execution time due to the iterative prediction calculations. To solve these problems, a novel dead-beat MPDPC strategy is proposed, using two active inverter voltage vectors and one zero inverter voltage vector during each duty cycle. Adoption of three inverter vectors ensures a constant switching frequency. Thus, smooth steady-state performance of both current and power can be obtained. Unlike the traditional three-vector based MPDPC strategy, the proposed three vectors are selected based on the power errors rather than the sector where the grid voltage vector is located, which ensures that the duration times of the selected vectors are positive all the time. Iterative calculations of the cost function in traditional predictive control are also removed, which makes the proposed strategy easy to implement on digital signal processors (DSPs) for industrial applications. Results of experiments based on a 1 kW inverter setup validate the feasibility of the proposed three-vector based dead-beat MPDPC strategy.