Model Predictive Control in Multilevel Inverters Part I: Basic Strategy and Performance Improvement

Abstract

Multilevel inverters (MLIs) have lately become important due to their extended application to electrical transmission and distribution systems. At the same time, the control and modulation of MLIs are especially challenging due to the high number of switching states, many of them redundant in terms of output voltage generation, and their nonlinear characteristics. In order to ease their implementation in real environment, model predictive control (MPC) is often considered, where the main control targets are: 1) to generate a the desired output current and 2) to keep the internal converter capacitor voltages at their reference value. However, a major issue with the implementation of MPC in MLIs is that the number of calculations to be done online increases dramatically with the number of levels, making it almost impossible to apply MPC in some practical cases. For these reasons, one of the main research trend in MPC for MLIs is to provide an algorithm which can reduce the computational burden necessary to operate the control. The article proposes a review of such control techniques. Starting from the basic MPC implementation and using a flying capacitor converter as an example the article review the basic strategies to avoid calculating the weighting factor in the cost function, simplifying the implementation. Also, methods to reduce the number of calculations necessary to implement MPC are shown and applied to cascaded H-bridge converters. These techniques allow to keep an high load current quality while reducing more than 95% in the number of calculations necessary to implement the control. Finally, other operation improvements of MPC are also included, such as fixed switching frequency operation and multistep MPC, reaching an important performance improvement compared to the basic MPC strategy.