A Two–Stage Model Predictive Control Strategy for Three–level NPC Converters based on Multistep Finite Control Set and Optimal Pulse Patterns

Abstract

Power transference between energy sources and the power system must meet a long and rigorous list of technical and operational requirements. To this end, LCL filter grid–connected three–level neutral point clamped converters have become a popular configuration for medium–voltage high– power applications, such as integrating distributed energy resources. The presence of the LCL filter increases the complexity of the control system when using traditional linear control schemes, so model predictive control (MPC) strategies arise as a convenient alternative as they have proved to render excellent reference tracking together with a fast dynamic response. Nevertheless, the computational complexity of MPC strategies becomes an issue as commercial control platforms must execute the real–time calculations in some tens of microseconds. Among the possible MPC strategies available, multistep–finite–control–set (MFCS) MPC has already been proven to be a vi able option regarding computational burden. However, MFCS controllers do not generate periodic nor symmetric voltages between the phases of the system, resulting in continuous Fourier spec tra in the system’s states. This a fundamental drawback for grid–connected applications, as grid codes pose limits to the admissible amplitudes of the grid currents’ harmonic components, where the analysis over interharmonics depends on the specific application, making MFCS controllers unsuitable for this type of application. On the other hand, optimal pulse patterns (OPP) guaran tee a low and fixed switching frequency of the semiconductors while minimizing the distortion of the state variables. However, the state–of–the–art formulations for OPPs assume that the system can be modeled as a purely inductive load. Such a model would lead to suboptimal performance and would not allow adding constraints that guarantee meeting with grid codes regarding the grid currents. This work proposes integrating two optimization stages dealing with different control objectives into a single and generalized control framework. This framework was then used to develop a control strategy (OPP–MFCS) that combines the benefits of OPPs and MFCS controllers. To this end, a novel formulation for OPPs considering the model of the LCL filter was proposed, allowing the direct minimization of the total demand distortion (TDD) of the grid currents while guaranteeing that the admissible limits for the harmonic components are met. On the other hand, by formulating a novel variable–step–size prediction horizon MFCS controller, extending the time–span covered by the prediction horizon with no additional computational burden was possible. This improvement led to an enhanced controller performance regarding both transient and steady–state responses of the system. Extensive analyses of the contributions regarding OPPs, MFCS controllers, and the OPP– MFCS strategy were held. The OPP–MFCS strategy was tested through both simulation and hardware–in–the–loop experiments, showing its capability of meeting with the IEEE Std. 519– 2022 technical standard while the converter operates at a low and fixed switching frequency of the semiconductors, with periodic and symmetric voltages, fulfilling all control objectives defined.