Machine learning and whale optimization algorithm for enhanced hydrogen storage alloy design

Xiaonan Liang, Shuai Liu, Huijun Zhang, Jie Qiu, Chenyan Fan, Yike Liu, Yuhao Yan,
Machine learning and whale optimization algorithm for enhanced hydrogen storage alloy design,
Materials Today Communications,
Volume 49,
2025,
113870,
ISSN 2352-4928,
https://doi.org/10.1016/j.mtcomm.2025.113870.
(https://www.sciencedirect.com/science/article/pii/S2352492825023827)
Abstract: Hydrogen storage and transportation face significant challenges due to its physical properties. Solid-state storage, particularly using metal hydrides, presents a promising solution. This study introduces an efficient approach to exploring hydrogen storage alloys by integrating machine learning (ML), intelligent optimization algorithms and materials simulation. 444 alloy samples spanning seven categories including intermetallic compounds (A2B, AB, AB2, AB5), mischmetal intermetallic compounds (MIC) and solid solution (SS) were collected and preprocessed, with 16 key elements identified as input features. Using the AutoML tool Tree-based Pipeline Optimization Tool (TPOT), the Extremely Randomized Trees (ETR) model demonstrated excellent performance in hydrogen storage with R2= 0.84, while the Gradient Boosting Decision Tree (GBDT) model achieved a high accuracy of 96.3 % in alloy type classification. The Whale Optimization Algorithm (WOA) was then combined with these models to generate 20 high-performance alloy candidates. V0.6Ti0.2Cu0.15Y0.05 and Mg0.7Cr0.1La0.15Al0.05 were validated through first-principles calculations. The former exhibited a hydrogen storage capacity of 3.137 wt.%, while the latter achieved 4.138 wt.%, both exhibiting prediction errors within 10 %. These results underscore the significant impact of elements like Mg on hydrogen storage performance and demonstrate the potential of combining machine learning with optimization algorithms for discovering efficient hydrogen storage materials. Notably, this approach also enables the revelation of hidden interaction patterns between elements—such as mutually exclusive relationships or synergistic effects (e.g., involving Zr and Ni)—that are difficult to deduce directly from experimental observations.
Keywords: Hydrogen storage alloys; Machine learning; Whale optimization algorithm; First principle calculation