Ensemble Machine Learning and Advanced Preprocessing applied to the Characterization of Carbon-Based Electrochemical Interfaces Present in Energy Storage Technologies

Vega Villamonte C.G.; Bazan-Aguilar A. · Proceedings - 2025 IEEE Latin American Conference on Nanotechnology, LANANO 2025 (2025) · DOI: 10.1109/lanano68466.2025.11512525

Resumen

The precise characterization of electrochemical systems using Electrochemical Impedance Spectroscopy (EIS) is a critical challenge in the development of sustainable energy storage technologies. The manual classification of Equivalent Circuit Models (ECMs) constitutes a significant bottleneck, requiring specialized knowledge and considerable time. This study presents an advanced Machine Learning methodology that integrates hierarchical preprocessing, multimodal feature extraction, and ensemble algorithms for the automatic classification of ECMs in carbon-based electrochemical systems. The results demonstrated a median test accuracy of 49.09 % with a 95 % confidence interval of 39.27 - 54.55 %, surpassing previous methodologies in statistical robustness and avoiding data leakage. This work reports the first successful automatic classification of an electrochemical double-layer capacitor (EDLC) fabricated from biomass-based carbon materials ever reported in the literature, independent of the biomass source. In this study, a carbonized orange peel-based EDLC was used as a representative example, which was entirely fabricated by the authors in the laboratory, and ECM-1 was correctly predicted with 69.80 % confidence. The methodology sets a new standard for scientific rigor in ECM-EIS classification, establishing a statistically robust framework for preliminary ECM screening despite moderate absolute performance, and contributing significantly to the development of biomass-based energy storage systems for electric mobility and renewable microgrid applications.

Instituciones: Universidad Tecnológica del Perú