Synthetic Data–Based Approach for Supercapacitor Characterization and Areal Capacitance Optimization Using Cyclic Voltammetry Data

Abstract

Optimizing areal capacitance for supercapacitors (SCs) using cyclic voltammetry (CV) involves complex, iterative experiments. Multiple tests are necessary to account for variations in electrode–electrolyte interactions and environmental factors, ensuring thorough characterization. However, this process is time consuming and labor intensive. This study leverages machine learning (ML) to streamline the procedure by generating reliable synthetic data, thereby reducing the time and resources required by traditional methods. The reproducibility of synthetic data makes it a valuable tool for research and validation. Various ML models are used for synthetic data generation, selected based on the characteristics of the real data. This research specifically employs the XGBoost (XGB) ML model to introduce variations in scan rates, enriching the dataset within the range of 5–600 mV/s. Results show that ML algorithms effectively preserve the statistical properties of CV data for laser-induced graphene (LIG) SCs, evidenced by a high R² value of 0.97 for the synthetic dataset, confirming the data’s fidelity. Additionally, the study introduces a Python module for calculating areal capacitance, facilitating assessment in both real and synthetic datasets. This approach accelerates SC analysis while maintaining data integrity, paving the way for future research and development.