Search

In this thesis, model predictive control (MPC) techniques are investigated with their applications to modular multilevel converters (MMCs). Since normally a large number of submodule (SM) capacitor voltages and gate signals need to be handled in an MMC, the MPC schemes studied in this thesis are employed for determining only the voltage levels of converter arms, while gate signals are subsequently generated by the conventional sorting method. Emphasis is given to inner-loop current control in terms of phase current and circulating current, aiming at performance enhancement and computation reduction. A variable rounding level control (VRLC) approach is developed in this thesis, which is based on a modification of the conventional nearest level control (NLC) scheme: instead of the conventional nearest integer function, a proper rounding function is selected for each arm of the MMC employing the MPC method. As a result, the simplicity of the NLC is maintained while the current regulating ability is improved. The VRLC technique can also be generalized from an MPC perspective. Different current controllers can be considered to generate the arm voltage references as input of the VRLC block, thus refining the control sets of the MPC. Based on the decoupled current models, the accumulated effect of SM capacitor voltage ripples is investigated, revealing that the VRLC strategy may not achieve a proper performance if the accumulated ripple is nontrivial compared to the voltage per level. Two indexes are also proposed for quantifying the current controllability of the VRLC. Benefiting from this analysis, A SM-grouping solution is put forward to apply such MPC techniques to an MMC with a large number of SMs, leading to an equivalent operation of an MMC with much reduced number of SMs, which significantly increases the current regulating capability with reduced complexity. As an example, the SM-grouping VRLC proposal is analyzed and its system design principles are described. This thesis also develops another MPC technique which directly optimizes the cost function using quadratic programming technique. Both a rigorous and a simplified procedure are provided to solve the optimization problem. Compared with the conventional finite control set (FCS)-MPC method which evaluates all voltage level combinations, the proposed scheme presents apparent advantage in terms of calculation cost while achieving similar performance.

This paper proposes a new control method for modular multilevel converter (MMC), which is based on a mod ification of the conventional nearest level control (NLC) scheme implemented by a model predictive control (MPC) technique. Applying the proposed method, the simplicity of the NLC is maintained while the current control ability in terms of both phase current and circulating current is enhanced by the MPC technique. A proper rounding function is used for each arm of the MMC instead of the conventional nearest integer function. The rounding function is determined by evaluating a cost function within the frame of MPC. To validate the proposed technique, a three-phase MMC prototype is employed. The conventional NLC-based control approach of MMC is also evaluated for comparative purposes. Results show that the proposed method exhibits excellent current-regulating ability. Compared with the conventional control scheme, the overall performance is improved without significantly increasing the computational burden or adding relevant complexity to the control system design.

The finite control set-model predictive control (FCS MPC) has been adopted as an excellent choice for the applications of multilevel converters during the last two decades for its salient performance. However, in the case of modular multilevel converters (MMCs), a high amount of calculation is always involved in the implementation, making the FCS-MPC less suitable especially for an MMC with a high number of submodules. To cope with the issue, this article proposes an MPC technique for the MMC with a very low calculation cost. In each sampling period, the arm voltage references of each phase are determined analytically by solving a constrained quadratic programming problem formulated from the cost function. Both a rigorous and simplified procedure is provided to solve the optimization problem. Then, the four nearest candi dates around the arm voltage references are evaluated, leading to a proper selection of arm voltage levels. Several experimental tests on an MMC prototype are carried out to validate the effectiveness of the proposed method. Results show that compared with the con ventional FCS-MPC method which evaluates all voltage-level com binations, the proposed scheme presents an apparent advantage in terms of calculation cost while achieving similar performance.