Comparative evaluation of data-based methods for detection of internal short circuits on cell-level
Year: 2026
Type: article
Abstract: Due to their relevance in field failures, early detection of internal short circuits (ISC) is critical for improving lithium-ion battery safety and remains a major focus of current research. However, objective comparison of existing detection methods is hindered by non-standardized validation setups, test conditions, and limited sample sizes. This study presents a Monte Carlo–inspired simulation framework to generate a large dataset of random ISC faults in a 12s1p battery module while accounting for measurement noise and cell-to-cell variations (CtCV). Multiple established ISC detection methods are systematically evaluated with respect to fault signal clarity, parameter sensitivity, and detection performance using statistical metrics and a newly introduced detection limit. The results show that complex multi-parameter approaches, such as entropy-based methods, do not outperform simpler algorithms, whereas detection methods based on descriptive statistics and correlation achieve robust performance, with correlation-based approaches remaining insensitive to voltage and impedance variations. Furthermore, data pre-processing by z-standardization effectively mitigates noise effects, and voltage down-sampling improves detectability while reducing computational effort. Extension of the analysis to experimental ISC and thermal fault data confirms the transferability of the simulation-based findings. Overall, these results provide a quantitative basis for selecting reliable ISC detection methods and establish a general framework for detection method validation beyond complex and costly experimental testing.
Source: Future Batteries
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,950
Funder: Technische Universität Clausthal
OpenAlex ID: W7130315439