ABSCA - A Boost Converter Switching Controller using Machine Learning Algorithms

B. Abegaz
Loyola University Chicago,
United States

Keywords: power converter, algorithms, AI, machine learning


The control of a type of switching voltage regulators (boost converters) is performed on a novel system-on chip named ABSCA. The ABSCA system executes cluster-based machine learning algorithms namely Gaussian, Hierarchical and Self-Organizing Mapping, and optimizes the performance of the boost converter for various power conversion applications. The results of the implementation show that the hierarchical clustering based machine learning algorithm implemented on the ABSCA controller improved the performance of the boost converter system by 3.907% in terms of reducing the voltage settling time. The SOM algorithm improves the performance of the system by 63.24% in terms of voltage settling time, and by 99% in terms of minimizing voltage overshoots.