Limited States Multi-Objective Direct Model Predictive Control of a Grid-Tied 3L-Active Neutral-Point Clamped Converter with an LCL Filter

Abstract

This digest presents a computationally efficient multi-objective direct model predictive control of grid-tied 3L-active neutral-point clamped converter (ANPC) with an output LCL filter. The ANPC topology is widely used as an interface for efficient power conversion in medium-high power applications. Direct MPC allows fast and decoupled control of state-variables through a user defined cost function. However, the state-variable predictions require iteration of 27 switching vectors which increases the computational burden, making the deployment of the controller infeasible to low-cost mass-marketed DSPs. The method presented in this study restricts the number of switching states being checked through a limitation criteria. The criteria reduces the number of states being checked from 27 to just 7 at maximum, making the computational burden identical to that of a six switch 2L-VSC. The presented approach is verified by deploying the controller on a TMS320F28379D DSP and the plant on PLECS RTBox 3. In addition, the presented method is compared against an existing state-of-the-art state vector limitation technique. The results show an improvement in grid current THD throughout the power range at the expense of marginal increase in switching frequency of the converter.