AI-based design of buck converter

• Main Author: Huang, Xianmiao
• Other Authors: Jack Zhang Xin
• Format: Final Year Project
• Language: English
• Published: 2019
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Power electronics
• Online Access: http://hdl.handle.net/10356/77346
• Institution: Nanyang Technological University

Multi-objective optimization of buck converter was done in CCM working condition in many researches. However, few researches focused on the optimization in DCM. In this report, three-objective optimization designed based on the buck converter which worked in discontinuous conduction mode. The equations of PC, Pon, PCu, PFe were different from the equations in CCM or BCM. In the first steps of this project, I deduced those equations and prepared all the parameters which was useful in the part B, which was a part to generate Pareto front based on 3 dimensions, including power loss, cutoff frequency and volume. NSGA-II, a kind of AI technology, was applied to generate Pareto frontier in this project. NSGA-II is an elitist and fast GA (genetic algorithm) evolved from basic genetic algorithm. In this report, the process of using NSGA-II to generate Pareto frontier was showed in details, including how the crossover process worked and the analysis of crowding distance sorting. I chose three extreme cases as three models from all the points in Pareto front. Case 1 had the minimum of power loss, while cutoff frequency is minimum among all the solutions in case 2. And the volume of filter was the minimal figure in the last case. In the final section in my project, I did a simulated experiment by MATLAB to visualize the difference between the working results of three extreme cases. And this will provide help designers of buck converter to select the suitable pairs of L and C.