Multi-busbar sub-module modular multilevel converter

Abstract

Modular multilevel converter (MMC) plays increasingly significant roles in large scale power electronics system including high voltage direct current (HVDC) system, static synchronous compensator (STATCOM), large scale energy storage, motor control, and so on, thanks to its advantages including modular configurations, reduced dv/dt, low total harmonic distortions, and low power losses. The classic sub-module (SM) topologies (e.g. half or full bridge types) all have in common their single connection arrangement between each SM in their series connection within a stack; i.e. a single busbar. This single busbar arrangement does come however with some drawbacks in terms of performance, reliability and flexibility. The lack of redundant switching states limits the potential optimization for the whole MMC.

To solve the above mentioned issue, this thesis presents the control and performance of a new topology of SMs for MMCs, which uses multiple parallel connections between SMs and is referred to as multi-busbar sub-module (MBSM). Stacks made entirely of MBSMs can see improved functionalities such as pre-charging capability, capacitor paralleling, lower power losses, improved reliability, and a rational bypass mechanism in the event of SM failure. The soft-parallel mechanism is proposed to maintain voltage balancing without the requirement of additional spike current inductors. Despite the fact that the number of semiconductors in MBSM MMC has been doubled, semiconductor losses have been reduced to 80% of those in its counterpart. Simulation results have verified the characteristics of a FB MBSM MMC in an HVDC scenario.

Several advanced control schemes for the control of the MBSM MMC are also investigated, including an algorithm to automatically generate independent variables state space models from linear electrical circuits, a model predictive control-based start-up controller to simplify the SM pre-charge procedure and at the same time improve the transient performance, and a reinforcement learningbased low-level controller to achieve low switching frequency operation of the MBSM MMC. The control schemes are validated by detailed theoretical analysis and simulation results.

Besides, some MBSM applications in the operation scenarios of STATCOM are studied. Two topologies of delta-configured, partially rated energy storage (PRES) MBSM STATCOM and their corresponding low-level controllers are presented to improve the active power output capability. The soft parallel of MBSM is more effective in reactive power mode than active power mode due to the location of ES, which sees their current circulation limited to their own SM capacitor. The proposed controller for the MBSM STATCOM dynamically switches between two operation modes to reduce the converter losses over the extended range of active power. Simulation results confirm the earlier point, in that PRES-MBSMSTATCOM performs better at pure reactive power set-points and marginally better at high active power. This is explained by the fact that MBSM operates more frequently in soft-paralleling mode when the ES releases less power, i.e. reactive power set-points. Then the MBSM concept is further extended to a structure with more busbars, named multi-H-bridge SM, aiming at solving the current sharing issue of paralleled discrete SiC MOSFETs in large current applications. When compared to conventional FBSM constructed directly paralleled SiC MOSFETs, simulation results show that the current sharing performance against on-state resistance mismatch is improved and the switching loss is reduced. The same converter rating can be achieved with fewer MHSMs compared with Si IGBT SMs.

Finally, the designing process of a benchtop-scale, low-voltage, open-source, and affordable hardware prototype of a MMC, the μMMC, is presented with a case study of a three-phase inverter-mode MMC. The proposed μMMC is configured as full bridge SMs type in the experiment, yet the flexible structure makes it capable to be configured as other SM types, including MBSMs. The cost for a single μMMC could be around 50 pounds. The control framework and concrete implementation are presented in detail. With the application of the μMMC, the STM32Cube Hardware Abstraction Layer, and the MATLAB/Simulink hardware support packages, it is possible to shorten the transition process from simulation to hardware realization to several hours. The experiment setup and results of a three-phase inverter mode MMC validate the proposed μMMC’s effectiveness, scalability, and convenience.