Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting DC-Link Capacitor Current Ripple in AC-DC Rectifier Circuits Using Fine-Tuned Large Language Models

Mohamed Zeid, Subir Majumder, Hasan Ibrahim, Prasad Enjeti, Le Xie, Chao Tian

Abstract

Foundational Large Language Models (LLMs) such as GPT-3.5-turbo allow users to refine the model based on newer information, known as ``fine-tuning''. This paper leverages this ability to analyze AC-DC converter behaviors, focusing on the ripple current in DC-link capacitors. Capacitors degrade faster under high ripple currents, complicating life monitoring and necessitating preemptive replacements. Using minimal invasive noisy hardware measurements from a full bridge rectifier and 90W Power Factor Correction (PFC) boost converter, an LLM-based models to predict ripple content in DC-link currents was developed which demonstrated the LLMs' ability for near-accurate predictions. This study also highlights data requirements for precise nonlinear power electronic circuit parameter predictions to predict component degradation without any additional sensors. Furthermore, the proposed framework could be extended to any non-linear function mapping problem as well as estimating the capacitor Equivalent Series Resistance (ESR).

Predicting DC-Link Capacitor Current Ripple in AC-DC Rectifier Circuits Using Fine-Tuned Large Language Models

Abstract

Foundational Large Language Models (LLMs) such as GPT-3.5-turbo allow users to refine the model based on newer information, known as ``fine-tuning''. This paper leverages this ability to analyze AC-DC converter behaviors, focusing on the ripple current in DC-link capacitors. Capacitors degrade faster under high ripple currents, complicating life monitoring and necessitating preemptive replacements. Using minimal invasive noisy hardware measurements from a full bridge rectifier and 90W Power Factor Correction (PFC) boost converter, an LLM-based models to predict ripple content in DC-link currents was developed which demonstrated the LLMs' ability for near-accurate predictions. This study also highlights data requirements for precise nonlinear power electronic circuit parameter predictions to predict component degradation without any additional sensors. Furthermore, the proposed framework could be extended to any non-linear function mapping problem as well as estimating the capacitor Equivalent Series Resistance (ESR).
Paper Structure (7 sections, 2 equations, 12 figures, 2 tables)

This paper contains 7 sections, 2 equations, 12 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Capacitor Ripple Current Estimation
  3. Use of LLM "Fine-tuning" for Modeling the Behavior of Power Electronic Circuitry
  4. Analysis of AC-DC Rectifier Systems
  5. Full-Bridge ac-dc Rectifier Model (Fig.\ref{['fig=BR_schenmatic']})
  6. Single Phase boost PFC AC-DC Converter (Fig. \ref{['PFC']})
  7. Conclusion

Figures (12)

  • Figure 1: Single phase boost Power Factor Correction (PFC) AC to DC rectifier schematic.
  • Figure 2: Block diagram of the proposed LLM fine tuned model for the ac/dc rectifier PFC circuit
  • Figure 3: Single phase bridge rectifier schematic, $V_{in}$, $V_{o}$, and $I_{cap}$ are the variables of interest.
  • Figure 4: Example of inaccurate output in power electronic circuit analysis when prompted to GPT 3.5. The LLM assumes that the values of interest changes in square root proportion as the system load changes, which is incorrect.
  • Figure 5: Motivating example for comparing in-context learning and fine-tuning for unknown function mapping.
  • ...and 7 more figures