To address the key bottlenecks in the state of health prediction of lithium-ion batteries under random variable load conditions, such as poor physical consistency, insufficient feature representation, and non-physical fluctuations in the prediction curves, this paper proposes a high-precision state of health prediction model that combines physical information constraints and multi-dimensional features, aiming to enhance the safety and long-term reliability of the battery management system. A neural network-transformer framework based on physical information was constructed, converting the consistency laws of irreversible battery aging into regularization loss terms in the training process to ensure physical rationality. Multi-dimensional coupled features including voltage, current, and cycle count were extracted to characterize the aging characteristics of the battery, and Bayesian optimization was applied to adaptively adjust key hyperparameters. Experimental results show that this model outperforms the comparison models in terms of MAE, RMSE, and R² metrics, effectively suppressing non-physical fluctuations, balancing high prediction accuracy and physical rationality, and providing technical support for the full life cycle health management and safe operation of lithium-ion batteries.

\gamma-Al₂O₃ is one of the most widely used catalyst supports in heterogeneous catalysis, yet its catalytic role remains unclear due to intrinsic structural disorder and complex surface species. This paper describes the development of a high-accuracy Al–O–H machine learning interatomic potential combined with global optimization to explore the structural landscape of \gamma-Al₂O₃ and its influence on propane dehydrogenation over single-atom Pt catalysts. Two energetically favorable structures, \gamma-no aluminum vacancy (NAV) and \gamma-aluminum vacancy (AV), are identified with energies lower than the conventional model by up to 34.44 meV∙(f.u.)^–1. These optimized structures expose abundant penta-coordinated Al³⁺ sites on the (100) surface, serving as preferred anchoring sites for Pt atoms. Simulated X-ray diffraction patterns indicate that \gamma-AV shows better qualitative agreement with experimental data. Surface phase diagrams further reveal that defect-rich surfaces are thermodynamically stabilized under realistic reaction conditions. Catalytic calculations demonstrate that \gamma-NAV and \gamma-AV significantly reduce propane dehydrogenation activation barriers through enhanced metal-support interactions associated with penta-coordinated Al³⁺ sites and defect-modulated local environments. These results suggest that \gamma-Al₂O₃ is better described as an ensemble of defect-rich surface configurations rather than a single crystal structure. These findings establish a direct relationship between atomic structure, defect chemistry, and catalytic performance in \gamma-Al₂O₃.

The electrocatalytic upcycling of biomass-derived platform molecules into value-added chemicals provides a sustainable pathway for transitioning toward a circular bioeconomy. Nickel–cobalt layered double hydroxide nanosheets supported on nickel foam (NiCo-LDH/NF) were successfully fabricated as electrocatalysts for the selective conversion of isobutanol (IBA) to isobutyric acid (IBAc) under mild conditions. The optimized catalyst exhibits a three-dimensional hierarchical nanosheet architecture with an enlarged electrochemically accessible surface area and favorable charge-transfer behavior. Benefiting from the interconnected network and altered electronic structure between the Ni and Co species, NiCo-LDH/NF achieved a Faradaic efficiency of 91.3% and selectivity of 93.2% toward IBAc at 1.52 V vs. RHE. Electrochemical analyses indicated improved IBA oxidation kinetics compared with those of the corresponding monometallic hydroxide catalysts. Mechanistic studies supported a sequential oxidation pathway from IBA to IBAc via adsorbed isobutyraldehyde intermediates. This study demonstrates the potential of bimetallic LDH electrocatalysts for the selective upgrading of biomass-derived alcohols.

To address the challenge of synergistic optimization of the multi-physical property parameters for the gas diffusion layer—a core component of anion exchange membrane water electrolyzers (AEMWEs)—this study establishes an integrated analytical framework combining multi-physics mechanism simulation and active learning-based intelligent optimization. Based on a three-dimensional steady-state multi-physics coupling numerical model of AEMWEs, the regulatory mechanisms of permeability, electrical conductivity, and porosity in the mass transport and heat transfer processes were elucidated. On this basis, a multi-model weighted ensemble active learning surrogate model was constructed to optimize the electrochemical performance of the electrolyzer. The model achieved a root mean square error (RMSE) of 16.5539 and a mean absolute error (MAE) of 12.9643 during the training process. On the test set, the model achieved a coefficient of determination (R²) of 0.9994, an RMSE of 5.0240, and an MAE of 4.0694. The optimal parameter combination for electrochemical performance identified by the proposed framework (a permeability of 1.18 × 10⁻¹⁰ m², an electrical conductivity of 2208.86 S∙m⁻¹, and a porosity of 0.605) effectively suppresses local heat accumulation and eliminates mass transport bottlenecks. Detailed post-optimization analyses further demonstrated that the optimal parameter combination achieved a favorable balance between the optimal electrochemical response and acceptable multi-physics field distribution uniformity.