Toward Trustworthy Evaluation and Benchmarking of Deep Learning Methods for State-of-Health Estimation in Lithium-Ion EV Batteries: A Mini Review
Journal of Energy and Sustainable Environment · 06.11.2026 · 1 (1)
DOI: 10.52650/7r1hrn73 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Energy and Sustainable Environment, online erschienen am 06.11.2026. Verfasst von Adam Muzaffar Abd Razak. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Deep-learning methods for lithium-ion electric vehicle (EV) battery state-of-health estimation have expanded rapidly, yet reported performance remains difficult to compare because evaluation practices are often non-comparable and context-dependent. This mini review examines the problem through Evaluation Credibility, reproducibility, cross-condition generalization, and deployment feasibility, distinguishing evidence-supported findings from evaluation principles that remain insufficiently characterized. The review’s central finding is that state-of-health (SOH) prediction error cannot be interpreted independently of the dataset, preprocessing, evaluation split, and operating context. In a representative single-study comparison, error under leave-one-battery-out validation was an order of magnitude higher than under a 50/50 split, illustrating the sensitivity of apparent accuracy to evaluation design. To address this problem, the review proposes a four-layer operational framework spanning Ground Truth Foundation, Core Benchmark Integrity, Multi-Domain Robustness Testing, and Deployment Feasibility. It also outlines an evidence-grounded research agenda focused on transfer-learning efficiency and cross-protocol adaptation, while separating these priorities from emerging areas requiring empirical validation. This synthesis draws on seventeen primary studies identified through a structured literature search and should be read as a step toward, rather than a completed instance of, standardized trustworthy SOH evaluation.
Zitieren
APA
Razak, A. M. A. (2026). Toward Trustworthy Evaluation and Benchmarking of Deep Learning Methods for State-of-Health Estimation in Lithium-Ion EV Batteries: A Mini Review. Journal of Energy and Sustainable Environment, 1(1). https://doi.org/10.52650/7r1hrn73
Vancouver
Razak AMA. Toward Trustworthy Evaluation and Benchmarking of Deep Learning Methods for State-of-Health Estimation in Lithium-Ion EV Batteries: A Mini Review. Journal of Energy and Sustainable Environment. 2026;1(1). doi:10.52650/7r1hrn73
BibTeX
@article{Razak2026,
author = {Razak, Adam Muzaffar Abd},
title = {Toward Trustworthy Evaluation and Benchmarking of Deep Learning Methods for State-of-Health Estimation in Lithium-Ion EV Batteries: A Mini Review},
journal = {Journal of Energy and Sustainable Environment},
year = {2026},
volume = {1},
number = {1},
doi = {10.52650/7r1hrn73}
}
EUVIMED