Digital model predictive control for multi-port DC-DC converters
Date of Issue2017-08-01
School of Electrical and Electronic Engineering
This thesis presents several digital model predictive control (MPC) methods for different types of multi-port DC-DC converters, including single-inductor single-input multiple-output (SI-SIMO) DC-DC converter, single-inductor multiple-input multiple output (SI-MIMO) DC-DC converter and single-input dual-output (SIDO) flyback converter. The multi -port DC-DC converters researched in this thesis have a common characteristic, i.e., the adoption of single inductor/transformer is implemented in the converters. Therefore, all loads driven by these multi-port DC-DC converters share the same inductor/transformer. Compared to the conventional multiple-load system using several single-input single-output (SISO) converters, this single inductor/transformer topology contributes to the reduction of component number, circuit size and cost. Moreover, the single inductor/transformer will also relieve the electromagnetic interference (EMI) problem, which IS senous resulting from maSSIve inductors/transformers in the multiple-load system with lots of SISO converters. The only disadvantage of single inductor/transformer is that the mutual interference among the loads, i.e., the cross regulation. Since there is residual energy stored in the shared inductor/transformer after the completion of driving one load, the residual energy will be delivered when driving next load, which means that it is influenced by the previous load. Consequently, the loads are supplied with mismatching energy rather than their corresponding demands. To solve the cross regulation problem, control methods with fast and dynamic response ability are imperative, which can provide optimal control actions to supply loads with the corresponding power requirements based on real-time data. Therefore, the cross regulation can be significantly reduced. MPC is an effective and robust method with the aforementioned fast and dynamic response capability. This method requires a state-space model of the targeted system. Then it is able to generate the optimal control actions in advance based on the state-space model, cost function, references and real-time states of the target system. The optimal control actions will be applied to the target system and the whole process will be repeated in the manner of receding horizon. Therefore, this MPC strategy has the capability to regulate the target system to track the references closely. For the SI-SIMO DC-DC converter, model predictive voltage control (MPVC) method and model predictive current control (MPCC) method are developed to suppress the cross regulation. In this thesis, the design of the proposed MPVC method with augmented state-space model, cost function and constraints for the SI-SIMO DC-DC converters are discussed. The simulation and hardware platforms of single-inductor single-input dual-output (SI-SIDO) buck converter are developed and the proposed the MPVC method is implemented. Simulation of the influences of MPVC method parameters, such as predict horizon, control horizon and Lagrange multiplier, is conducted to guide the control parameters setting for hardware implementation. Steady-state operation and dynamic performance of the proposed MPVC method are also simulated. Moreover, several experimental cases are conducted to study the performance of the proposed method based on the simulation results and the SI-SIDO buck converter hardware prototype. Simulation and experimental results demonstrate that the proposed MPVC method guarantees low cross regulation for the SI-SIMO DC-DC converter and has the capability to respond to variations in load and reference rapidly. The development of the MPCC method is also presented in this thesis. Furthermore, a charge equalization method for the series-connected batteries in electric vehicles (EVs) is proposed based on the MPCC method. The imbalance of batteries in EVs resulting from manufacturing differences and cycles of charging and discharging will shorten the lifetime of battery cells. Therefore, charging equalization is the most important factor for the lifetime extension of batteries. A possible charging equalization method is an individual charging operation based on the state-of-charge (SOC) of each battery cell. In order to realize the individual charging, SI-SIMO DC-DC converter using MPCC method is employed. Similar to the MPVC method, the MPCC method can regulate the output currents to drive loads and reduce the cross regulation. Since the battery charging is split into the constant current (CC) charging and the constant voltage (CV) charging, the corresponding current reference assignment algorithms in the CC stage and the CV stage are also developed to provide the proper reference for the MPCC method. Simulation studies in different SOC situations are conducted to verify the performance of the proposed charge equalization method, and the results indicate that it is able to balance the series-connected batteries successfully. The MPC method is extended to control the SI-MIMO DC-DC converter in this thesis. This SI-MIMO DC-DC converter also has the cross regulation problem resulting from the adoption of single inductor. Different from the SIMO topology, the MIMO topology also needs the regulation of input power considering the multiple power sources. Hence, a power sharing and cross regulation suppression method is developed for the SI-MIMO DC-DC converter. The proposed method is based on the MPC method, power sharing method and time-multiplexing method. The power sharing controller is developed to regulate the power from the multiple power sources, and the cross regulation suppression controller is designed to control the output voltages independently with reduced cross regulation. In order to perform the validation of the proposed method, simulation and experimental platforms are built. Several simulation and experimental cases, including steady-state operation and dynamic performance, are conducted, and the results demonstrate that the proposed method is able to regulate the SI-MIMO DC-DC converter effectively and robustly. Finally, a dynamic model predictive voltage control (DMPVC) technique for the SIDO flyback converter is presented. Cross regulation is still the critical problem because two secondary coils share the same primary coil in the transformer of the SIDO flyback converter. The proposed the DMPVC technique has the ability to reduce cross regulation by generating the optimal control signals dynamically. The state-space models of flyback for DMPVC method are formulated, as well as the corresponding cost function which defines the optimal control actions. The verification of the proposed DMPVC technique is conducted based on a simulation platform built in Simulink. With the proposed DMPVC method, the SIDO flyback converter is able to work in buck-buck, buck-boost and boost-boost modes, and complete the mode transformation successfully.
DRNTU::Engineering::Electrical and electronic engineering