Low-voltage DC (LVDC) systems offer a promising means for improving distribution system efficiency and reliability. The DC-AC conversion stage, however, is one of the main challenges for LVDC networks. A low-voltage modular multilevel converter (MMC) for LVDC distribution systems is proposed in this paper. Analysis is presented to show that its efficiency can exceed that of a conventional 2-level converter. The low voltage rating of each MMC submodule enables MOSFETs to be used in place of IGBTs to reduce power losses. The application of synchronous rectification (SR) further reduces conduction losses. It is shown that device switching frequency reduces as the number of MMC levels is increased. MMC power losses, for different numbers of levels, are compared with those of a conventional 2-level converter. Simulation and experimental results are presented to confirm the mathematical analysis.

No abstract

Efficency, power quality and EMI are three crucial performance drivers in LVDC applications such as electrical supply, EV charging or DC aerospace. Recent developments in SiC MOSFETs and MMC for LVDC promise two significant improvements in LVDC inverter performance. However, the designer is left with many combinations of technology and inverter level to choose from. This paper aims to clarify this choice by identifying one optimum Si design and one optimum SiC design, using detailed loss calculations. An IGBT inverter is included as a baseline. Loss calculations estimate the effects of Si MOSFET switching loss and all parasitic interconnection loss. The validity of the loss estimations are verified using careful experiments on a Si MOSFET cell. Close agreement indicates that the modelling approach is valid for extension to many cells in series, and to the parallel connection of many devices. Despite the lower EMI inherent in MMC inverters, Si MOSFETs risk worse EMI, due to poor reverse recovery characteristic. Slowed device gate switching experimentally demonstrates the reduction in switching noise, promising very low EMI. This initial study has therefore identified two promising candidate SiC and Si MOSFET inverters which will be fully constructed in future work, in order to aid designers in choosing the optimum semiconductor technology and topology for LVDC inverters

A highly efficient DC-AC converter is key to the success of low-voltage DC (LVDC) distribution networks. Calculated power losses in a conventional IGBT 2-level converter, a SiC MOSFET 2-level converter, a Si MOSFET modular multilevel converter (MMC) and a GaN HEMT MMC are compared. Calculations suggest that the parallel-connected Si MOSFET MMC may be the most efficient topology for this LVDC application. In this paper, the current unbalance limits for the parallel-connected MOSFETs and the optimal number of parallel-connected MOSFETs for MMC are investigated. Experimental results are presented for current sharing in parallel-connected MOSFETs and for the verification of power loss in a single Si MMC module.

LVDC distribution networks have the potential to release larger capacity without having to upgrade the existing cables. One of the main challenges of LVDC networks is the extra customer-end DC-AC conversion stage. This paper proposes and evaluates a 5-level Si MOSFET-based MMC as a promising alternative to the conventional 2-level IGBT-based converter. This is due to the comparatively higher efficiency, power quality and reliability, and reduced EM emissions. A comprehensive analysis of a Si MOSFET 5-level MMC converter design is performed to investigate the suitability of the topology for LVDC applications. Detailed theoretical analysis of the 5-level MMC is presented, with simulated and experimental results to demonstrate circuit performance. To suppress the AC circulating current, especially the dominant 2 nd harmonics, this paper presents a double line-frequency PI with orthogonal imaginary axis control method. Comparison of simulation and experimental results with those for double line-frequency PR control shows that the proposed PI controller has better performance. In addition, it is simpler to implement and more immune to sampling/discretisation errors.