Inductorless cascaded low-power DC-DC converter: optimizing performance metrics through machine learning techniques

Abstract

Abstract This study presents a groundbreaking methodology for optimizing the operational efficiency of a three-stage boost DC-DC cascaded converter through the application of a Random Forest (RF) machine learning algorithm. A novel figure of merit is meticulously formulated to quantitatively evaluate the converter’s performance, focusing on critical metrics such as power conversion efficiency, output DC ripple levels, and response time. The Random Forest model is trained on a comprehensive dataset encompassing a wide range of resistive and capacitive design parameters, with the figure of merit serving as the output indicator. Rigorous simulations and analyses demonstrate that the integration of LM741 operational amplifiers and strategically employed CMOS power MOSFETs yields remarkable enhancements in circuit performance. Specifically, the application of the Random Forest algorithm results in a significant reduction of voltage ripple to 15 mV, a marginal decrease in rise time to 120 μ s, and an impressive overall efficiency improvement, achieving a peak efficacy rating of 93%. Furthermore, an extensive evaluation of the converter design’s compatibility with CMOS technology, alongside detailed IC layout considerations, underscores the meticulous and innovative nature of this research. A comparative analysis between the Random Forest model and alternative machine learning approaches elucidates the superior utility and applicability of the RF algorithm in enhancing circuit performance. These findings not only advance the state of the art in DC-DC converter design but also pave the way for future innovations in power electronics optimized through machine learning techniques.