Averaged models and control of modular multilevel converters

该研究项目涉及用于中压驱动或高压直流输电应用的模块化多电平变流器的控制设计。更具体地说，要解决臂之间能量对称的问题，同时输出电流必须满足负载的要求。这导致系统驱动不足，这意味着输入的尺寸太低，无法控制所有变量。然而，考虑到系统量在较慢的时间尺度上的平均行为提供了引入更多输入变量的可能性。这一步骤通常是含蓄地进行的，并且不需要为平均系统推导动态模型，它将得到填补现有科学鸿沟的单扰动方法的支持。作为副作用，希望表明已知的实现与待开发的平均框架兼容。期望通过模型方案的正式应用来保持原始系统的特征特性，以便在控制器设计中考虑这些特性。此外，平均形式主义将引入比所需更多的输入变量，这提供了在它们之间共享控制行动的可能性，以满足诸如考虑限制等附加目标。由于该建模方案对其他(欠驱动)基于ARM的拓扑也具有吸引力，因此希望适当地推广这些发现。建模和控制设计的结果将使用现有的试验台进行实验验证。

The research project deals with the control design for modular multilevel converters which are used for instance in medium-voltage drives or for high-voltage dc transmission applications. More specifically, the problem of symmetrizing the energies between the arms is to be tackled while the output currents must meet the requirements of the load. This renders the system underactuated which means that the dimension of the input is too low in order to control all variables. However, considering the averaged behavior of the system quantities on a slower timescale offers the possibility to introduce more input variables. This step, often made implicitly and without deriving a dynamic model for the averaged system, is to be backed by singular perturbation methods closing the existing scientific gap. As a side effect, it is desired to show that the known implementations are compatible with the averaging framework to be developed. It is expected that characteristic properties of the original system are maintained by a formal application of the modelling scheme such that these properties can be taken care of in the controller design. Furthermore, the averaging formalism will introduce more input variables than required which offers the possibility to share the control action among them in order to meet additional goals such as the consideration of limitations. Since the modelling scheme will be attractive for other (underactuated) arm-based topologies as well, it is desired to generalize the findings appropriately. The results of the modelling and control design will be verified experimentally using an already existing testbench.