With the increasing drawbacks of fossil energy characterized by high consumption and emissions, the goals of carbon peaking and carbon neutrality have become central to structural transformation in the energy sector. The study addresses the dual challenges faced by regional hydrogen-supported microgrid networking, namely, uncertain source-load conditions and equipment health states. A novel networking optimization method is proposed that integrates uncertainty modeling, health assessment, and multi-criteria decision-making, aiming to achieve economically optimal, safe, and flexible collaborative operation of microgrids. First, probabilistic statistics and scenario generation techniques are employed to accurately quantify stochastic fluctuations from renewable generation and load demand, thereby enhancing the adaptability of networking schemes through multi-scenario simulations. Second, a dynamic state of health (SOH) assessment is introduced, embedding degradation models of proton exchange membrane fuel cells (PEMFCs) and proton exchange membrane electrolyzers (PEMECs) into the planning process to enable resource allocation optimization over time. Third, a multidimensional evaluation framework is constructed, encompassing operational costs, long-term health state variations, and total investment costs. Finally, under different source-load quantile scenarios, the proposed method demonstrates superior economic efficiency and robustness in PEMFC and PEMEC capacity configuration compared with conventional mixed-integer programming approaches. Moreover, the incorporation of SOH constraints significantly improves both energy output and economic performance of the system.

Solid oxide fuel cell (SOFC) is an efficient and environmentally friendly energy technology, used in distributed generation, transportation and residential systems. However, high-temperature operation, multi-physics coupling, and long-term degradation issues create significant control challenges. This paper reviews recent advances in SOFC control with key findings showing a shift from classical methods to intelligent systems. Parameter optimization currently employs algorithms such as the response surface method to obtain optimal control objectives. Fault diagnosis and performance prediction combine physical models with machine learning. Control strategies are evolving from traditional control toward intelligent control. Furthermore, application studies guide the tailoring of these advanced controllers to real-world settings. Despite these progresses, challenges remain due to insufficient control intelligence or weak fault tolerance. To address these gaps, this paper outlines two future research paths, including lifecycle digital twins for predictive health management, and explainable AI control for creating trustworthy systems by embedding physical constraints and ensuring decision transparency.

To extend the lifetime of proton exchange membrane fuel cells (PEMFCs), the two most prevalent operational faults- membrane drying and flooding-must be diagnosed and analyzed. This study presents a comprehensive fault diagnosis framework that integrates machine learning with an equivalent circuit model (ECM) for open-cathode PEMFC systems. First, based on the fuel cell fast impedance test platform built in the laboratory, water management fault experiments are conducted, and the corresponding fault impedance data are obtained. Then, the second-order R-constant phase element ECM parameters are identified adaptively using the rapid and efficient data processing method to obtain the seven-dimensional original sample set. Afterward, principal component analysis (PCA) is applied to downsize the high-dimensional state feature dataset, and four principal component parameters are selected as fault diagnosis features. In addition, Fuzzy C-Means (FCM) clustering and synthetic minority oversampling technique (SMOTE) algorithms are combined for feature enhancement. Finally, the support vector machine (SVM) optimized by the grid search technique is implemented for fault classification. Five-fold cross-validation shows that the proposed PCA-FCM-SMOTE+GS-SVM-linear framework can effectively identify normal, flooding, and drying states with 97.63% accuracy, outperforming other algorithms. Validation under varying temperatures further confirms its robustness, achieving 94.74% on unseen conditions.

With the increasing role of solid oxide fuel cells (SOFCs) in microgrids, real-time economic dispatch faces signif icant challenges due to complex transient dynamics and current-driven degradation. There is a pressing need for methods that can both preserve SOFC response characteristics and ensure long-term reliability during dispatch. To this end, this study first develops a surrogate model that captures transient response features while remaining suitable for real-time scheduling. Degradation induced by stack current variations is then incorporated into the dispatch formulation to explicitly account for lifetime impact. On this basis, to tackle the optimization problem involving nonlinear dynamics, operational targets and constraints, and load stochasticity, a safe reinforcement learning framework with two Lagrangian multipliers is proposed: one is used to regulate grid reliance and power imbalance, while the other enforces degradation targets. Case studies show that the surrogate model enables accurate dynamics reproduction, while the two-multiplier strategy achieves balanced performance under uncer tainty. Results reveal clear cost-degradation trade-offs across different degradation targets, consistent positive correlation within single training runs, and performance sensitivity to both degradation limits and load profiles. Comparative analysis further confirms that neglecting dynamics or degradation leads to overly optimistic cost estimates and the inability to effectively track and mitigate degradation.

Optimization of multi-energy system (MES) configuration under realistic operational constraints remains highly challenging due to strong nonlinearity, non-convexity, and highly fragmented feasible regions induced by multi-carrier coupling, reliability requirements, and diverse complex operational constraints. Although simulation-based meta-heuristic approaches are widely adopted to capture energy management system (EMS)-driven operational behavior, their search processes remain largely fitness-driven and structure-blind, treating EMS simulations as black-box evaluators and neglecting the system-level information revealed by operation trajectories and feasibility assessments. As a result, existing methods often suffer from slow convergence, inefficient feasibility exploration, and frequent trapping in infeasible regions when addressing complex configuration problems. To overcome these limitations, this study proposes a meta-heuristic optimization framework enhanced by a structured feasibility repair operator for EMS-driven MES configuration. The operator exploits slack in feasible regions and violation information in infeasible regions to guide the search process, complementing the intrinsic search mechanisms of the underlying meta-heuristics. Furthermore, reinforcement learning adaptively tunes the operator activation and correction magnitude based on search feedback, facilitating a balance between exploration and feasibility. The case studies demonstrate the successful application of the proposed repair operator within genetic algorithm, particle swarm optimization, simulated annealing, and tabu search. Extensive experiments based on a highly complex EMS-driven optimization model further show that the framework improves convergence performance and enhances configuration quality across all four algorithms. The repair-enhanced variants consistently identify superior configuration solutions with comparable, and in some cases reduced, computational cost relative to the baseline algorithms. Although the reinforcement learning-based parameter adaptation introduces a moderate increase in search time, it eliminates the need for extensive pre-tuning of hyperparameter ranges. Furthermore, the repair-augmented frameworks exhibit improved robustness compared with their baseline counterparts, demonstrating higher tolerance to suboptimal hyperparameter settings, reduced sensitivity to EMS rule design, and stronger performance under stochastic operating conditions.

Proton exchange membrane fuel cell (PEMFC) is a key device of the new energy industry, which is undergoing continuous development and application, but the durability is the factor that restricts its next promotion. Thus, exploring effective and accurate fault diagnosis and state monitoring method is eagerly needed. Various studies have proved that electrochemical impedance spectroscopy (EIS) offers exceptional capability in characterizing the internal states of PEMFC. To reduce the testing time of EIS, broadband signals are commonly to extract the EIS efficiently. In this work, we adopt discrete interval binary sequence (DIBS) as the excitation signal, and propose an optimization method for a specific testing environment. An equivalent circuit model of PEMFC with electrochemical noise (ECM-EN) is established, where the noise is predicted by autoregressive model based on measured voltage noise. Afterwards, the designed DIBS is applied to ECM-EN simulation, the discrepancy between simulation result and theoretical value directs the optimization of Fourier magnitude spectrum of the DIBS. Results show that the proposed method can derive a suitable DIBS for a specific PEMFC system without requiring high computational resources, enabling its deployment on embedded devices with limited resources.

To meet high-power demands during cruise operation, marine solid oxide fuel cell (SOFC) systems typically adopt modular multi-stack architectures for power scaling. However, manufacturing variability and unstable reactant supply inevitably induce inter-stack inconsistencies during long-term operation, accelerating localized deterioration and constraining overall system cost-effectiveness. To address this issue, this paper proposes an operating expenditure function (OEF)-driven progressive consistency regulation framework for multi-stack balancing of high-power SOFC systems. First, a previously developed and validated multi-physics SOFC model is employed to capture electrochemical behavior, thermal evolution, and triple-phase boundary (TPB) degradation. Then, the OEF maps conventional operating variables into a unified temperature-fuel utilization domain, creating a controllable decision space linking system inputs and key states. Furthermore, a fuzzy-cost coordinated strategy is developed to progressively regulate the direction and intensity of power shift among stacks, enabling adaptive power allocation through enhanced perception and suppression of inter-stack inconsistency. Results show that, compared with equal power sharing, the proposed strategy reduces overall inconsistency by approximately 78%-87%, mitigating cumulative deviations in temperature fields, reaction zones, and TPB degradation, while improving comprehensive system performance by 15%-30%. These findings demonstrate the effectiveness and engineering potential of the proposed method for high-power marine SOFC systems.

The internal current distribution is a critical indicator of the operational status and health of Proton Exchange Membrane Fuel Cells (PEMFC). However, existing reconstruction methods based on external magnetic fields suffer from limited accuracy and high dependence on large-scale datasets. To address these challenges, this study proposes a novel inversion approach integrating Physics-Informed Neural Networks (PINN) and Bayesian optimization. First, a multiphysics PEMFC model with a serpentine flow channel and an electromagnetic model are established to generate current-magnetic field datasets under various operating conditions. Subsequently, the Biot-Savart law is embedded into the loss function to ensure physical consistency. Furthermore, Bayesian optimization is employed to adaptively tune the weights of data and physical loss terms, resolving the hyperparameter balancing problem. Experimental results demonstrate that the proposed method significantly improves reconstruction accuracy and reduces data dependency compared to traditional data-driven methods.

Electrochemical impedance spectroscopy (EIS) is widely used in fault diagnosis of fuel cells (FCs) but faces challenges in online applications due to high cost and lengthy measurement durations. This article introduces a rapid EIS measurement technique centered on adaptive, segmented discrete-interval binary sequence (DIBS) signals that are dynamically tailored to the FCs’ actual electrochemical noise (EN). The noise characteristics are determined through autoregressive (AR) model-based power spectral estimation (PSE), allowing the excitation signal to be optimized for a high signal-to-noise ratio (SNR) across the target frequency range while respecting system linearity constraints. An advanced data processing method is also applied to the raw impedance data to further improve its quality. Comparative simulation analysis of EIS results obtained using different DIBS schemes and the traditional maximum-length binary sequence (MLBS) excitation demonstrates the accuracy and robustness of the proposed method. In addition, a practical experimental platform is established based on the programmable electronic load to verify its real-world utility. Experimental validation confirms that a relatively accurate impedance spectrum from 0.5 to 1 kHz can be obtained in a single test of 16 $\boldsymbol{\sim}$ 18 s, providing a key technological foundation for online monitoring of FCs under dynamic operating conditions.

Both the increased utilization of photovoltaic (PV) and the consequent expansion of energy storage, leading to higher costs, are critical factors in the development of a techno-economic microgrid. To address this trade-off, optimal sizing design of a multi-energy microgrid is essential. This paper proposes a novel linear search-linear programming method for optimizing the sizing of microgrids with diverse energy storage systems, including battery energy storage systems (BESS) and hydrogen energy storage systems (HESS). The method minimizes both the non-linear and non-convex life cycle cost and the PV curtailment rate. Simulation results demonstrate the feasibility and superiority of the proposed method, achieving enhanced computational efficiency: 52 times faster for PV/HESS microgrids, 124 times for PV/BESS, and 5 times for PV/HESS/BESS. Moreover, more comprehensive and accurate optimal configurations were identified, including 4289 PV units for both the PV/BESS and PV/HESS/BESS microgrids, and 6239 PV units along with 37 fuel cell units for the PV/HESS microgrid. Based on the advanced optimization outcomes, a techno-economic analysis shows that the PV/HESS/BESS microgrid optimally combines the advantages of both HESS and BESS, substantially improving overall system performance.

Fuel cell heavy-duty trucks (FCHDTs) have large cargo capacity, and their power fluctuation range can reach 3-5 times that of ordinary passenger cars. As a result, fuel cells are more prone to continuous high load, rapid load change and other deteriorated conditions, which poses a major challenge to energy management strategies. To address this issue, this paper proposes a predictive energy management strategy based on multizone Tube model predictive control (MPC). First, a multizone velocity predictor is introduced, utilizing improved transformer and K-means clustering techniques. This method includes multiple prediction submodels to handle the wide power range of FCHDTs. Then a two-stage Tube predictive control method is proposed. In the first stage of predictive control, a multizone maximal robust positive invariant set is constructed to obtain tube constraints. The multiobjective function is designed to balance fuel economy with durability and generate the optimal reference power trajectory. In the second stage, during online operation, MPC technology with short time steps is used to adjust instantaneous power in real-time, tracing the reference trajectory established in the first stage. Results demonstrate that this method effectively smooths the output power fluctuations of the fuel cell and reduces the total operating cost of FCHDTs; specifically, compared to single-stage MPC, the dual-stage MPC strategy reduces the standard deviation of fuel cell power by 0.43, 2.82, 4.86, and 6.49 under lithium battery power boundaries of +/- 24, +/- 48, +/- 72, and +/- 96 kW, respectively.

Developing efficient, eco-friendly power generation systems is crucial for future clean energy policies. Biomassdriven solid oxide fuel cell (SOFC) systems promise clean energy, but ensuring efficient, safe operation remains challenging. Additionally, multi-stack SOFC systems are an effective means to ensure fuel utilization efficiency and enhance system reliability. This study uses aggregate modeling to model the gasification-integrated parallel multi-stack SOFC system (GIMCS). Fifteen biomass-derived fuels are passed into the GIMCS to analyse the effects of gasification temperature, water vapor mass flow rate to biomass mass flow rate (S/B) on syngas fractions, and their impact, along with reaction temperature, on power generation performance. Then, the dataset of the GIMCS (15 different biomass gases as fuels) was used for genetic algorithm backpropagation (GA-BP) model training for operating condition prediction (electrical efficiency, net voltage, and current density of each stack). Additionally, CO2 emissions from waste biomass gasification were compared to those from power generation via a GIMCS. The findings suggest that the GA-BP model provides highly accurate output estimates (R-2>0.991, MAPE<0.238, RMSE<0.234) and that the GIMCS emits less CO(2 )than waste biomass gasification. This study supports predicting the performance of GIMCS to enhance waste biomass-to-electricity conversion and optimize system operating parameters for efficient and safe operation.

High-power multi-stack solid oxide fuel cells (multi-stack SOFCs) mark a significant advancement in SOFC technology, particularly for large-scale power generation. However, the slow load response of SOFC systems, which limits their ability to quickly adapt to load fluctuations, poses substantial challenges for their practical application. Additionally, issues related to system integration and variability in manufacturing tolerances result in inter-stack inconsistency, which may negatively impact the performance of the multi-stack SOFC system. Consequently, this study constructs a dual-stack SOFC/battery hybrid power generation system and proposes a two-layer energy management strategy to solve the above challenges. The first layer control strategy adopts a state-based method and uses a Battery Energy Storage System (BESS) to reduce load fluctuations, which solves the problem of slow SOFC load response. The second layer control strategy solves the problem of inconsistency between stacks in the system by using fuzzy logic control to distribute the power between stacks, and realizes the performance convergence control between stacks. In two typical scenarios, the proposed strategy is superior to the equal power allocation strategy. Scenario 1, the system’s lifespan and efficiency increased by 14.6% and 0.5%, respectively. Scenario 2, the system’s lifespan and efficiency increased by 10.8% and 1%, respectively.

Solid oxide electrolysis cell (SOEC) are a highly efficient hydrogen production technology, but thermal stresses induced by temperature gradient under high temperature may accelerate performance degradation and cause equipment damage. Optimizing operating parameters to balance efficiency and thermal safety is critical to addressing these challenges. Existing optimization methods primarily rely on numerical simulations and experimental validations, but their high time and economic costs limit practical applications. To overcome this challenge, this study proposes a performance optimization method that, unlike conventional models, considers thermal safety constraints and rapidly determines optimal SOEC operating conditions with varying power inputs. By integrating the deep neural network with a one-dimensional multiphysics field model, the method leverages data generated by complex multiphysics field model to train the network and systematically analyze the effects of different operating conditions on SOEC stack power and efficiency. Subsequently, the slime mold algorithm is employed to efficiently optimize operating parameters under power and thermal constraints, achieving maximum SOEC efficiency and providing theoretical support for future dynamic control strategies.

Accurately predicting the performance degradation trend of fuel cells helps take measures in advance and prolong the stack's service life, leading to a novel hybrid forecasting approach. The first grey model prediction method based on residual exponential smoothing optimization (ES-R-GM) can capture the voltage deterioration trend. We explore the integration of two different ES techniques, specifically the double ES (ES2) and cubic ES (ES3), to investigate their effect on enhancing the predictive accuracy of the GM model. The second method of the adaptive network fuzzy inference system (ANFIS) can characterize local nonlinear behavior. We utilize the simulated annealing (SA) algorithm to optimize ANFIS results under different fuzzy rule selection strategies. The outcomes of the two prediction methods mentioned above are combined to create a hybrid prediction using the data fusion method and the moving window technique. Various hybrid methods are evaluated under general conditions and further detailed optimization. The data collected from the experimental platform confirms the suggested hybrid framework. The results show that the hybrid ES3-R-GM + ANFIS-SC method outperformed the single models in final prediction accuracy and can effectively track both global trends and local changes. Simultaneously, it takes less time to calculate than the literature. Moreover, when applied to consistent public datasets, the hybrid approach maintains its robustness and accuracy compared with other hybrid prognostic methods.

Proton exchange membrane fuel cells (PEMFCs) will greatly shorten their lifespan due to platinum (Pt) catalyst degradation during operation. This paper proposes an optimization-based energy management method considering Pt degradation, which is an improvement over the traditional strategy that only focuses on fuel optimization to consider both the minimal fuel and the minimum fuel cell life decay. Firstly, a one-dimensional (1D) Pt degradation model is established to comprehend how various voltage situations affect Pt deterioration. Then, various strategies to suppress Pt degradation are designed using Pontryagin's minimum principle (PMP) optimization algorithm in light of the influence analysis results, and the effects of the PMP algorithm under different strategies are tested on the hardware-in-the-loop (HIL) simulation platform. The results demonstrate that the performance of the PMP algorithm in real-time strategy is extremely near to the global optimal solution generated by the offline dynamic programming (DP) algorithm. After adding the tendency to limit high potential and voltage variation in the PMP algorithm, hydrogen consumption increases by only 2%. In comparison, the stack's degradation is decreased by nearly 50%, considerably extending the stack's service life and reducing the system's comprehensive use cost.

This research aims to explore key issues in the field of walking robots and exoskeletons. By using a depth camera to record human walking data in different terrains and employing RGB and depth information for image classification, our approach utilizes both early fusion and late fusion methods to explore their impact on terrain classification. This approach achieved fast and accurate performance in recognizing terrain. We captured data for walking upstairs, downstairs, upward ramps, downward ramps and level ground and used the vision transformer for data fusion. This work compares the performance of unimodal RGB model, unimodal deep model, Pre-Fusion model and Post-Fusion model in terms of accuracy, recall and processing frequency, with accuracies of 0.954, 0.934, 0.976, and 0.976, respectively. The results show that the Post-Fusion model slightly outperforms the Pre-Fusion model in terms of accuracy and recall with more reliable classification performance. To understand the attention mechanism better, we've been visualizing the weights of the model, observing how different attention heads may capture different features of the data from various dimensional perspectives. The visualization helps in interpreting the model's focus and how it distinguishes between features of different nature, such as color texture and spatial depth. These findings can inform the ongoing refinement of terrain classification systems for applications in robotics and autonomous navigation.

There is inconsistency in the stack of multi-stack Solid Oxide Fuel Cell (SOFC) system, which can affect the efficiency and lifespan of the system. To improve the overall performance of the system when the stacks are inconsistent, the paper analyzes the impact of stack inconsistency and designs a power allocation strategy based on improved generalized predictive control (GPC). The types of stack inconsistency have been identified and online detection methods have been proposed at first. Then, through stack inconsistency analysis, it can be found that stack inconsistency can seriously affect the efficiency and lifespan of the system and the system lifespan can be maximized by adjusting the current. Furthermore, the following research also indicates that the point where the system has the highest lifespan is also the point where the overall performance of the system is optimal. Finally, to achieve the best performance of the system during actual operation, the paper designs a power allocation strategy based on improved generalized predictive control. The simulation results also show that the power allocation strategy designed in this paper can adjust the power of the stack in real-time online, ensuring maximum system lifespan and optimal overall performance.