Estimation of State of Health of Li-Ion Batteries using Data Driven Techniques

• Published in: 2023 9th IEEE India International Conference on Power Electronics (IICPE) pp. 1 - 5
• Main Authors: Verma, Brijendra Kumar, Bhagtani, Sarvesh, Ram, Subhash Kumar, Abhishek, Anand, Khanna, Devansh
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28.11.2023
• Subjects: Bagging regression; Batteries; Battery health; Estimation; Learning systems; Power electronics; Predictive models; Regression model; SoH estimation; Stacking; Training
• ISSN: 2160-3170
• DOI: 10.1109/IICPE60303.2023.10474979

Battery health is a critical factor influencing the widespread adoption of Electric Vehicles (EVs). The SoH serves as a reliable indicator of a battery's capacity degradation over time, directly impacting its performance and overall lifespan. In this research paper, various ensemble based techniques have been used to estimate the state of Health (SoH) of battery. To assess the effectiveness of the proposed approach, a detailed comparison of several state-of-the-art regression models is conducted. Specifically, the performance of Bagging Regression, Gradient Boost Model, Stacking Regression Model, xGBoost Model, and Extra Trees Model is evaluated in predicting the SoH of batteries. R-squared (R2) score is utilized as the metric for assessing the accuracy and predictive capability of the developed models. The performance of the developed models are evaluated using progressively reduced amounts of data, demonstrating its effectiveness even with limited available data. Through extensive analysis of the developed models, efficacy of the data-driven technique is demonstrated in estimating battery SoH accurately. Certain models outperform others in terms of R2 score, indicating superior predictive capabilities for battery health estimation. The results offer valuable insights for selecting the most suitable regression model for battery SoH estimation, contributing to enhanced EV battery management and optimization.