Multisampling Based Modeling and Control of Medium-Voltage High-Power Inverter with Global Optimal Method

The medium-voltage DC power system has broad application prospects in new energy, rail transportation, and integrated power system for ships. As the connection channel between a medium-voltage DC network and a low-voltage AC network, a medium-voltage high-capacity inverter power supply undertakes the task of supplying power for daily use and auxiliary equipment, which needs to meet the requirements of power quality, efficiency, power density, and reliability. The MW-level inverter power supply has a low switching frequency (500 Hz～2 kHz) due to the performance of existing mature medium-voltage high-capacity power electronics devices. The main challenge is to achieve high dynamic response capability at a low switching frequency. This paper focuses on the refinement model, characterization, control parameter optimization, and other related research work under the application of multisampling techniques for medium-voltage DC inverter power supplies. Firstly, a state-space averaging model is established for the main circuit topology of the combined inverter and verified in the time domain. On this basis, a closed-loop control strategy containing double-closed-loop PI, feedforward decoupling, and active damping is adopted for the functional and index requirements of the inverter power supply in steady-state, dynamic, and grid-connected switching. A closed-loop state-space model under delay-free conditions is established to lay the foundation for the subsequent work. Secondly, the basic principle and the PWM model based on multisampling are introduced. The pure delay link is a transcendental equation, which is reduced to a finite-dimensional polynomial using the Pade approximation instead of the delay link. The system-level closed-loop model of the inverter considering multisampling is developed based on the Pade approximation. The effect of multisampling on the dynamic characteristics of the system is investigated by a single-axis equivalent model. The analysis results show that increasing sampling times can reduce the digital control delay, increase the system’s closed-loop bandwidth, and reduce the step response overshoot. In this paper, the closed-loop bandwidth of a single sampling is 51.95% of the delay-free model’s bandwidth, while the closed-loop bandwidth of the system reaches 81.71% of the bandwidth of the delay-free closed-loop model when the sampling times N is 5. Accordingly, a sampling number selection method is proposed that considers the bandwidth of the closed-loop system and the achievability of digital control. A 20 kW prototype platform is built to verify the correctness of the model and analysis results. Finally, according to the closed-loop model of the approximated inverter, the stability of the system is analyzed using the eigenvalue method, the sensitivity analysis of the PI control parameters to the eigenvalues is carried out, and the set of control parameters to be optimized is identified by sensitivity. Then, a mathematical model for control parameter optimization is established, and the final control parameters are obtained using a global optimization algorithm. In this paper, the closed-loop bandwidth of the optimized system is increased by 47.78%, and the system's overshoot of the step response is reduced by 20.09%. A 20 kW prototype platform verifies the effectiveness of the global optimization algorithm. © 2023 Chinese Machine Press. All rights reserved.