Design GA & PSO- Based High-Efficiency SEPIC DC-DC Converter for Context-Aware Duty Cycle Control
Shreyas Rajendra Hole, Agam Das Goswami
Abstract
The SEPIC or Single-ended primary-inductor converter (SEPIC) via DC-DC converter can be controlled via duty cycle scheduling mechanisms. The output voltage levels of SEPIC depend upon internal components ratings and the control switch’s duty cycle, which can be intelligently controlled for context-aware circuits. The proposed DSLCADCC method uses a hybrid of Particle Swarm Optimization (PSO) with a Genetic Algorithm (GA) Model, which assists in real-time duty cycle reconfigurations. This model is supported via a customized lightweight and low-power pre-trained Q-Learning Model that continuously monitors the performance of the SEPIC circuit and performs incremental tuning of the GA & PSO Models. The Q-Learning Model estimates circuit efficiency, total harmonic distortion (THD), and power consumption & operation delay of the bioinspired SEPIC Model for retuning its performance. The model can reduce overall power consumption by 18.5% and improve the efficiency of SEPIC circuits by 15.4% while maintaining high operating speed and similar THD performance compared with its non-reconfigurable counterparts. This high performance enhances the capability of the proposed model in terms of deployment for a wide variety of circuits that are built with multiple in-circuit components and require different voltages for component-level operations.