Multivariable control of multilevel converters in microgrid

• Main Author: Jayan, Vijesh
• Other Authors: Amer M. Y. M. Ghias
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering::Power electronics
• Online Access: https://hdl.handle.net/10356/161428
• Institution: Nanyang Technological University

Multilevel converters have gained significant attention in hybrid microgrid applications. This is because they deliver quality output at low-sized passive filters and reduced voltage stress on the power devices. The multilevel converter in hybrid microgrid mandates a multi-objective controller to control its multiple variables, such as inductor currents, ac/dc bus voltages, active/reactive power, neutral-point/flying capacitor voltages etc., to their desired nominal references. Conventionally, a cascade structure of multiple linear controllers is required to control these variables. However, such an approach not only deteriorates the converter dynamics but also intensifies the complication of obtaining the gain parameters for the linear controllers. These issues are easily addressed using a finite control set model predictive control (FCS-MPC). Unlike linear controllers, the FCS-MPC applies optimum control action to the converter by minimizing a cost function that contains the weighted sum of the Euclidean distance between the predicted variable and its reference; thus, offering better dynamic performance and easy implementation of multivariable control. However, the FCS-MPC with numerous control variables gets computationally intensive and requires complex mathematical analysis or numerical models to design appropriate weighting factors. Besides these complexities, the converter in a hybrid microgrid utilizes linear controllers as a secondary control loop to generate references for the control variables in FCS-MPC. Since the gain parameters of these linear controllers are tuned for a specific operating point, the desired dynamic performance is not guaranteed when the converter transits to a different operating point. This thesis aims to develop enhanced model predictive control (MPC) algorithms for multilevel converters in a hybrid ac/dc microgrid. The contributions of this thesis are classified into three parts. The first part proposes an adaptive dynamic reference (ADR) model that generates an appropriate current reference for FCS-MPC to realize dc bus voltage regulation in microgrid. Unlike linear controllers, the design of ADR model is simple and independent of the converter’s operating point. As a result, the proposed model guarantees the desired dynamic performance when the converter transits to any operating point. Compared to existing dynamic reference (DR) model, the proposed ADR model eliminates the steady-state error and is tolerant to model parameter uncertainties, unmodelled dynamics, and sensor imperfections. The proposed ADR model is validated experimentally on a grid-connected two-level converter operating on FCS-MPC, and its performance is compared with the DR model and linear controller. The second part proposes single-objective model predictive control (SO-MPC) and single-objective modulated model predictive control (SO-M2PC) algorithms for bidirectional dc-dc flying capacitor (FC) converter integrating battery to the dc microgrid. The key feature of the proposed algorithms is that they achieve multivariable control using a single-objective cost function based on battery current error. The SO-MPC first identifies the optimum voltage level by minimizing the single-objective cost function and then utilizes a redundant state selection (RSS) algorithm to obtain the optimum state. While the SO-M2PC identifies the optimum sequence of states using simplified duty ratio expressions of power switches, which are derived from the single-objective cost function and FC charge balances. Compared to conventional FCS-MPC, the proposed algorithms are computationally less intensive and free from the weighting factors. The proposed algorithms are validated experimentally under varying photovoltaic (PV) system power and load in a dc microgrid, and their performances are compared with the conventional FCS-MPC. The final part proposes a new hybrid microgrid configuration, where a multi-output multilevel (MOM) converter interfaces different ac sources and loads through its output ports. Adopting such configuration eliminates the redundant power conversion stages in the ac side; thereby, reducing number of control variables and power conversion losses in the microgrid. However, most of the MOM converter fails to operate its output ports independently beyond certain voltage level. Hence, the theoretical operational limits of the MOM converter are required to be explored for different operation modes. The operational limits of a cascaded dual-output multilevel (CDOM) converter are developed in this thesis using an FCS-MPC algorithm and is validated experimentally under different operation modes. This thesis also proposes a cascaded model predictive control (CMPC) algorithm for a MOM converter in a hybrid microgrid. Unlike conventional FCS-MPC, the proposed CMPC algorithm attains multivariable control by performing a sequential execution of multiple single-objective MPC units. Adopting such a strategy not only eliminates the need for weighting factors but also reduces the computation burden of the controller. The proposed CMPC is validated experimentally on a flying capacitor dual-output (FCDO) converter in hybrid microgrid, and its performance is compared with conventional FCS-MPC.