Benchmarking of Modulation Strategies for Three-Phase Three-Level VSIs

Modern power electronics demand increasingly compact and highly efficient converters. To achieve this goal and significantly reduce the physical size of the output filters, these systems must operate at higher switching frequencies. This requirement drives the industrial adoption of three phase three level inverters, specifically the I-Type and the T-Type Neutral Point Clamped topologies. However, controlling these systems introduces a physical challenge, which is the balancing of the neutral point voltage. Uncontrolled voltage drift distorts the output waveforms and exposes the components to destructive electrical stress. This thesis presents a comprehensive benchmark of a selection of the state of the art carrier based modulation strategies to evaluate their different features for the following Three Level Neutral Point Clamped inverter topologies: a Si-based I-Type and a GaN-based T-Type. The flow of the research begins with a literature review to select the most relevant modulation techniques, ranging from high efficiency discontinuous methods to unconditional balancing algorithms. Following the theoretical selection, each modulation strategy is mathematically analyzed and validated via PLECS simulations. The experimental phase was made possible by Imperix, which hosted the research and provided the hardware test bench and rapid prototyping controllers. The control logic is deployed on this setup to evaluate industrial operating scenarios: motor drives, low-speed startups, grid-connected power injections, and reactive power compensation. The experimental results compare the selected modulations across the chosen operating points using a set of critical figures of merit. The empirical data prove that while some strategies minimize switching losses but fail to provide active neutral point balancing, other methods guarantee autonomous and active balancing but are limited due to the increased number of commutations. Ultimately, the thesis defines the operational boundaries of each selected strategy, showing that a universally optimal modulation for Three Level inverters does not exist. The empirical results prove that current methods force a rigid compromise between efficiency and neutral point stability.