The dynamic behavior of hydrogen gas in the anode recirculation loop of a proton exchange membrane fuel cell (PEMFC) is governed by coupled flow‑pressure interactions within the ejector, supply manifold, and stack channels, making precise pressure regulation a critical challenge for system efficiency, stability, and lifetime. Ensuring adequate anode fuel supply and stable hydrogen supply pressure requires meticulous control of the gas handling system. This study proposes a single‑ejector‑based backstepping sliding mode controller (BSMC) for PEMFC hydrogen recirculation systems. To compensate for the unknown gas flow disturbance induced by periodic purging, a radial basis function (RBF) neural network is integrated with the BSMC, which adaptively corrects the control law under different system states. Simulation results demonstrate that the proposed RBF‑compensated BSMC effectively mitigates purge‑induced pressure fluctuations. The controller achieves a maximum tracking error of 8.173 kPa, a mean absolute error of 0.17 kPa, and a root mean square error of 0.53 kPa. Consequently, the proposed control strategy offers superior stability and robustness compared to conventional proportional‑integral‑derivative (PID) controllers under step load current variations, variable pressure targets, and periodic purge disturbances.

For gain faults in electromagnetic linear actuators (EMLAs) used in active suspension systems, this paper proposes a fault diagnosis and classified fault-tolerant control (CFTC) scheme based on an improved sliding mode learning observer (ISMLO). Considering external disturbances, actuator gain faults, and measurement noise, a comprehensive mathematical model of the faulty EMLA system is established. The proposed ISMLO integrates a proportional-derivative (PD) type learning observer with a sliding mode control law, and a forgetting factor is introduced to enhance adaptability and estimation performance. The observer enables rapid state tracking and accurate fault estimation under disturbance conditions. Based on the estimated fault information, a CFTC strategy is developed in which faults are classified into minor and major categories, and corresponding primary and secondary fault-tolerant controllers (PFTC and SFTC) are designed accordingly. To improve robustness, an H_∞ performance index is incorporated into the design of the observer and controllers. The stability of the overall system is rigorously analyzed using Lyapunov theory and linear matrix inequalities (LMIs), and the effectiveness of the proposed method is validated through Hardware-in-the-Loop (HIL) experiments. Experimental results demonstrate that the ISMLO reduces displacement and fault estimation errors by up to 88% and 86% compared with the conventional learning observer (LO) and sliding mode observer (SMO), with strong fault-tolerant performance confirmed. After applying the CFTC strategy, the average steady-state error is reduced to 0.212 mm. These results confirm that the proposed approach provides fast state tracking, accurate fault estimation, and strong fault-tolerant performance for EMLA systems.

Accurate state of charge (SOC) estimation for lithium iron phosphate (LFP) batteries across a wide temperature range remains challenging, owing to their flat open-circuit voltage (OCV) plateau and pronounced temperature sensitivity. To address this, a Long Short-Term Memory (LSTM) network is proposed that integrates internal pressure features with a dual-attention mechanism. The measured internal pressure is introduced as a new physical observation to compensate for the lack of SOC-identifying information in the voltage plateau region of LFP batteries. The model jointly learns from internal pressure, temperature, voltage, and current. The LSTM's gating mechanism captures the dynamic coupling between these multi-source signals and SOC under wide-temperature conditions, while the dual-attention mechanism further improves estimation accuracy and interpretability. Experiments on the own dataset show that the method achieves root mean square errors (RMSE) of 1.44%, 1.14%, and 1.34% at 25℃, 35℃, and 45℃, respectively. In domain adaptation tests, fine-tuning with only 15% of the target temperature samples yields RMSE of 1.55% at 35℃ and 1.97% at 45℃. Cross-domain validation on a public National Aeronautics and Space Administration (NASA) dataset, using only 10% of target domain samples for fine-tuning, results in an RMSE of 6.25% and an r² of 0.9339. Ablation studies confirm the unique contribution of the internal pressure feature. This work provides a robust, interpretable, and high-accuracy framework for LFP battery SOC estimation, paving the way for safer and more reliable battery management systems.

Amid the global energy transition, lithium-ion batteries (LIBs) serve as a cornerstone of clean energy storage systems. Robust state of health (SOH) monitoring is critical for cost reduction, safety hazard mitigation, and predictive maintenance. A systematic review of artificial intelligence (AI)-driven SOH assessment technologies is presented, with a comprehensive framework established covering feature mining, state estimation, and health management. Multi-physics health features across electrical, thermal, and mechanical domains are categorized, and corresponding feature extraction methods and dimensionality reduction techniques are classified. The modeling strengths of conventional machine learning for small-to-medium datasets are elaborated, and deep learning's automatic feature extraction capability under complex operating conditions is highlighted. Applications of reinforcement learning in charge-discharge optimization, energy scheduling, and battery health management are reviewed. Key challenges are identified, including multi-physics feature decoupling difficulties, limited model generalization and interpretability, inconsistent evaluation benchmarks, and industrial deployment barriers. Future research directions include constructing multi-physics coupled feature systems, fusing electrochemical mechanisms with data-driven models, developing lightweight reinforcement learning algorithms, and advancing engineering applications through cloud-edge collaboration and digital twin technologies.

To address safety hazards arising from trace H₂ leakage in hydrogen energy systems, catalytic elimination using a Pt/Al₂O₃ catalyst was investigated. A test system was constructed to study the effects of Pt loading, inlet temperature, and space velocity on hydrogen conversion, along with catalyst microstructure analysis. Single-factor experiments indicate that increasing Pt loading from 1 wt% to 4 wt% significantly raises conversion rate from approximately 30% to over 85% at 1 vol% H₂. Elevating the inlet temperature effectively overcomes the activation barrier, with a particularly pronounced effect at low temperatures. In contrast, the influence of space velocity is relatively weak and negatively correlated with hydrogen conversion. Multivariable synergistic analysis identifies inlet temperature as the dominant factor for overcoming surface activation limitations at low H₂ concentrations, while Pt loading enhances reaction kinetics via increased active site density. At H₂ concentrations above 2 vol%, the system enters an "interfacial reaction saturation" regime where all parameter effects diminish. Numerical simulations confirm these mechanisms, clarifying that synergistic mode of action wherein temperature dominates activation while Pt loading increases the reaction rate.

With the rapid advancement of Vehicle-to-Everything (V2X) communication and automated driving technologies, connected and automated vehicle (CAV) platoons have emerged as a promising solution for enhancing road capacity, energy efficiency, and traffic safety. Unlike the ideal fully CAV environment, real-world traffic will remain in a mixed traffic phase for an extended period, where CAVs and human-driven vehicles (HDVs) coexist. Factors such as cut-in by surrounding vehicles, HDV leader disturbances, and communication delay significantly increase the complexity of platoon control. This paper provides a systematic review of CAV platoon control methods under mixed traffic from three interrelated perspectives: control architecture, HDV interaction mechanisms, and control methods. From the architectural perspective, centralized management, distributed cooperation, and hybrid hierarchical architectures are comparatively examined in terms of decision efficiency, communication dependency, and scalability. Regarding HDV interaction mechanisms, HDV behavior modeling and prediction approaches based on model-driven, data-driven, and game-theoretic frameworks are synthesized, highlighting the trade-offs between modeling fidelity and uncertainty quantification capability. In terms of control methods, the technical characteristics and applicability limits of model-driven control, data-driven control, and hybrid model-data-driven control are systematically analyzed. Building upon this comparative assessment, the advantages, limitations, and application scenarios of these approaches are summarized to provide guidance for control strategy design and selection. Finally, future research directions for CAV platoon control under mixed traffic flow are outlined, aiming to provide theoretical support for establishing safe, stable, and efficient CAV platoons in complex mixed traffic environments.

Battery management systems (BMSs) play an undeniably critical role in both power and energy storage systems. As applications continue to expand into various complex scenarios, BMSs have increasingly become the key determinant of the overall performance of advanced battery systems. Beyond conventional basic performance metrics, lifetime and safety gradually emerge as core concerns for battery systems. However, existing BMSs are increasingly inadequate in supporting these two aspects. The development of high-safety, long-lifetime BMSs has become a common focus in battery systems across various application scenarios. This paper reviews the current state of BMS technology, analyzes its shortcomings in both hardware and software, and summarizes the latest technological advancements across four key areas: multi-dimensional parameter measurement, multi-modal fusion modeling, active management, and embedded artificial intelligence (AI) deployment. It objectively evaluates the strengths and weaknesses of these technologies for future BMS applications. Finally, the paper innovatively proposes the fundamental concept of an embedded AI BMS, highlighting its potential for deployment in both power battery and energy storage battery applications. As the core conclusion of this paper, the embedded AI BMS plays a significant role in next-generation battery systems. The discussion of key technologies in this paper also provides comprehensive references and analytical insights for the development of the next-generation smart BMS.