Real-Time Implementation of Online Adaptive Space Vector Modulation for a Three-Level NPC Inverter

This paper presents the concept and real-time implementation of a new modulation method for a three-level neutral-point-clamped (NPC) inverter, called online adaptive space vector modulation (oaSVM). It is based on SVM but explores any triangle containing the reference voltage and chooses three voltage vectors that improve the trade-off between inverter switching losses and load current THD. The resulting mixed-integer optimization problem has a combinatorial complexity and is solved using a branch and bound method which is implemented for real-time execution on an FPGA. The loss-thdtrade-off of oaSVM and SVM are compared in MATLAB-simulations. Hardware-in-the-loop (HIL) simulations on a system-on-chip (SoC) consisting of a CPU and FPGA show execution times that make this implementation feasible for real-time applications.