To address the high-dimensional, strongly constrained, and priority-sensitive reconfiguration challenges faced by Integrated Power Systems (IPS) during faults, this paper proposes a Graph-Masked Advantage Actor-Critic algorithm (GMA2C). The proposed method represents the post-fault shipboard IPS as a graph composed of generator and load nodes. A multi-head graph attention network (GAT) is used to extract topological features between power sources and loads. A feasible action mask is introduced at the output of the Actor to suppress load restoration requests that have no valid power supply path under the current topology. In addition, a capacity verification mechanism based on the power management system is used to check the physical feasibility of the policy output. This avoids restoration schemes that violate generator capacity constraints. Simulations were conducted under multiple fault scenarios in two typical ship IPS topologies. The proposed method is compared with six representative baseline algorithms. The simulation results show that GMA2C provides competitive overall performance in power restoration rate (PRR), maximum load restoration rate (MLSR), critical load survival rate (CLSR), and number of service outage loads (NSO) across the evaluated fault scenarios. It also obtains feasible solutions in a stable manner. Moreover, the model contains approximately 20000 parameters, suggesting low online inference overhead. This study provides a feasible graph reinforcement learning modeling method for topology-level intelligent reconfiguration of shipboard power systems after faults.

In an isolated island environment, a hybrid power supply system of fuel, biomass, wind, solar, and storage faces multiple challenges, such as multi-source heterogeneity, strong coupling of AC and DC, and resource constraints. Its fault characteristics show cross-domain propagation concealment and multi-scale dynamic complexity. Traditional single diagnostic paradigms have difficulty balancing real-time performance, accuracy, and interpretability. Therefore, this article proposes a physics- and data-driven hierarchical fault diagnosis framework that is coordinated by signal perception. This framework builds a three-level collaborative mechanism of “end-edge-cloud:” at the device perception layer, it uses signal processing techniques such as wavelet packet transform combined with lightweight mechanism rules to achieve millisecond-level locking of hard faults in power electronic devices and transient protection. At the regional decoupling layer, it integrates the data mining capabilities of graph neural networks (GNNs) with the power grid topology model to accurately analyze the cross-domain propagation paths of faults in AC and DC hybrid systems. At the system decision-making layer, it establishes a “physics–data dual-drive” closed loop, using long short-term memory networks (LSTMs) to capture nonlinear dynamic residuals to correct physical models, and at the same time introduces physics-informed neural networks (PINNs) to constrain the training boundaries of data-driven models, ensuring that decisions comply with physical laws such as energy conservation. In addition, for the fluctuation of wind and solar power and computing power constraints in isolated island scenarios, a feature adaptive strategy driven by transfer learning and a model pruning algorithm are designed. Verification of typical cases, such as gas turbine surge warning and energy storage converter cascading fault analysis, shows that this framework can effectively break through the weak fault detection bottleneck in a strong noise background, significantly reducing the false alarm rate while ensuring a high detection rate. This framework provides a new paradigm for improving the overall resilience and intelligent operation and maintenance level of isolated island hybrid power supply systems.

The development of More Electric Aircraft (MEA) necessitates that Electro-Hydrostatic Actuator (EHA) controllers achieve exceptional power density within rigorously constrained volumes. However, the compact layout design of these controllers constitutes a challenging NP-hard problem, characterized by strong multi-physics coupling—such as electromagnetic, thermal, and structural fields—and complex nonlinear constraints. Traditional meta-heuristic algorithms frequently suffer from premature convergence and struggle to balance global exploration with local exploitation. To address these challenges, the core contribution of this paper is the proposal of a novel Fractional-Order Anteater Foraging Optimization Algorithm (AFO), which is successfully applied to an established EHA controller layout optimization model. At the algorithmic level, by incorporating the Grünwald–Letnikov fractional derivative, the algorithm exploits the inherent memory property of fractional calculus to dynamically adjust the search step size and direction based on historical evolutionary information, thereby preventing stagnation in local optima. At the engineering application level, a high-fidelity mathematical model of the EHA controller is established, comprising 11 design variables and 10 critical physical constraints, including parasitic inductance minimization, thermal radiation efficiency, and electromagnetic interference (EMI) isolation. Extensive validation against the CEC2005 and CEC2022 benchmark functions demonstrates the superior convergence accuracy and stability of the AFO algorithm. In a specific EHA case study, the proposed method reduced the controller volume by 33.9% while strictly satisfying all multi-physics constraints, compared to traditional methods. Furthermore, a physical prototype was fabricated based on the optimized layout, and experimental tests confirmed its stable operation and excellent thermal performance. The results validate the efficacy of incorporating fractional calculus into bio-inspired algorithms to solve complex, high-dimensional engineering optimization problems.

ObjectiveTo address the problems of high network losses during the operation of islanded distribution networks with a high penetration of distributed power sources and the slow convergence and poor stability of existing algorithms used for network reconfiguration, a dynamic reconfiguration strategy for island distribution networks based on an improved grey wolf optimization (GWO) algorithm is proposed, with the primary objective of minimizing network losses.MethodA dynamic reconfiguration model for islanded distribution networks is established, with the minimization of active network losses and voltage deviation as the optimization objectives. To enhance the global search capability and convergence efficiency of the GWO algorithm, several strategies are introduced, including probabilistic perturbation, a dynamic tabu list, and adaptive parameter adjustment. ResultsThe results of the dynamic reconfiguration modelling and simulation of an islanded distribution network with a high penetration of distributed power sources show that the improved GWO algorithm achieves better stability, higher accuracy and greater computational efficiency compared to other algorithms, including the original GWO algorithm and the improved particle swarm optimization (PSO) algorithm. Under static reconfiguration conditions, the active power loss in the islanded distribution network is reduced by 21.8%, and the minimum bus voltage is increased by 2.03%. Under dynamic reconfiguration strategy, the active power loss is reduced by 27.98% over a 24-hour period. Furthermore, under extreme weather and line fault scenarios, the reconfiguration strategy continues to ensure stable network operation, with intra-day power losses reduced by 22.16% and 26.30%, respectively. ConclusionThe results show that the improved GWO algorithm provides a novel theoretical framework and optimization approach for the dynamic reconfiguration of islanded distribution networks.

Direct-starting of industrial motors has problems such as large current impact (five to eight times the rated current), mechanical stress damage, and low energy efficiency. This paper explores the technological innovations in motor soft-start driven by intelligent control and wide-bandgap semiconductors, and constructs a highly reliable and low energy consumption solution. Firstly, based on a material–device–algorithm system framework, a comparative study is conducted on the performance breakthroughs of SiC/GaN in replacing silicon-based devices. Secondly, an intelligent control model is established and a highly reliable system architecture is developed. A comprehensive review of recent literature indicates that SiC devices can reduce switching losses by up to 80%, and intelligent algorithms significantly improve control accuracy. System-level solutions reported in the industry demonstrate the capability to limit current to 1.5–3 times the rated current and achieve substantial carbon emission reductions. These technologies provide key technical support for the intelligent upgrading of industrial motor systems and the dual-carbon goal. In the future, development will continue to evolve in the direction of device miniaturization and other directions.

ObjectivesTo address the acute contradiction between the extremely high power demands of high-energy pulsed weapons and the requirement for full-spectrum stealth capability, this paper systematically reviews research progress and challenges in integrating energy adaptation and stealth coordination within shipboard integrated power systems. MethodsA systematic review methodology is employed to establish a three-dimensional collaborative analysis framework integrating energy, stealth, and intelligence. Within this framework, an in-depth assessment is provided of energy management strategies for pulsed power adaptation, global signal management techniques, and intelligent collaborative design methods based on digital twin technology. Furthermore, the study systematically reviews the current state of research on key technologies such as hybrid energy storage system topologies, multi-physics-based characteristic signal suppression, and cross-domain collaborative decision-making mechanisms. ResultsIn terms of energy matching, hybrid energy storage systems are widely recognized as a mainstream solution for mitigating pulsed power impacts and ensuring grid stability. Regarding stealth coordination, active control technologies such as digital degaussing and active noise control are increasingly replacing traditional passive suppression methods. At the system integration level, the introduction of artificial intelligence and digital twin technologies shows strong potential for addressing challenges related to millisecond-level dynamic response requirements and inefficient R&D iteration cycles. ConclusionsTheoretical research and technical analysis indicate that establishing an integrated collaborative design framework for precise energy supply and intelligent signal control is a fundamental approach to addressing compatibility challenges associated with high-energy weapons aboard naval vessels. In particular, intelligent dynamic trade-off control and cross-domain collaborative optimization are expected to become key future research directions, carrying significant strategic importance for enhancing the combat effectiveness and survivability of next-generation naval vessels.

Accurate identification of key nodes in complex networks is vital for optimizing system robustness and controlling information spread. Existing centrality metrics struggle to balance the continuous extraction of global topological features with the fine-grained perception of local structures, while traditional heuristic algorithms also face severe resolution limitations. To address these issues, this paper proposes a node importance evaluation method based on fractional-order topological propagation and local information entropy (FSEC). This method overcomes the limitations of discrete integer-order propagation inherent in traditional graph walks. It constructs a continuous fractional-order topological propagation operator within the spectral graph theory framework. This enables the smooth projection of node degree features into the global topological space, thereby yielding high-order global impact factors. Furthermore, an information theory mechanism is introduced to quantify the probability distribution of a node’s information contribution within its local neighborhood. The local structural information entropy is then calculated to reflect the node’s asymmetric control over micro-level information flow. Deliberate attack simulations were conducted on nine real-world networks and three types of artificial network models. The results show that the proposed FSEC algorithm significantly outperforms baseline algorithms like Autoencoder and Graph Neural Network (AGNN), Degree Centrality, k-shell, PageRank, and Mixed Degree Decomposition (MDD) in degrading the largest connected component (LCC) and global network efficiency (NE). The proposed method also achieves the minimum Area Under the Curve (AUC) values globally. Its monotonicity is slightly lower than that of AGNN but superior to all other baseline algorithms. In addition, SIR simulations further confirm the effectiveness of the FSEC method. This approach successfully resolves the ranking tie problem among nodes in the same topological layer.

With the rapid evolution of a new power system characterized by a high proportion of renewable energy, system operations have become increasingly random, variable, and uncertain. The system model exhibits features such as high dimensionality, multiple time scales, stochastic behavior, and nonlinearity. This paper proposes a large-scale computational power system model architecture based on cloud-edge-terminal collaboration. By defining functional roles within the cloud-edge-terminal structure and implementing a global model coordination mechanism, the approach enables an organic integration of global awareness, local adaptation, dynamic training, and online optimization for power system problem models. At the cloud level, various object models and the power grid topology are constructed. The edge generates typical problem models for the power system, while the terminal devices produce lightweight models adapted to local grids. This architecture supports collaborative modeling for key business scenarios such as power flow analysis, stability assessment, and reactive power optimization. The study focuses on the training methods of distilled parameters within the terminal models to enhance their adaptability for real-world deployment in power systems. Simulation results demonstrate that the cloud-edge-terminal model offers excellent scalability, adaptability, and real-time performance for computations in new power systems, effectively supporting localized, intelligent operations and decision-making within the system.

Shipboard integrated power systems (IPS) are evolving toward multi-zonal architectures with high power density. Consequently, fault diagnosis faces dual challenges regarding topological complexity and high-dimensional features. Traditional data-driven methods often overlook the physical topology of power systems. This oversight limits diagnostic accuracy. To address this, this paper proposes a fault diagnosis framework based on a residual graph attention network (ReGAT). First, the method utilizes discrete wavelet transform (DWT) to construct node features. The graph topology is established according to physical connection relationships. Subsequently, a residual connection mechanism is introduced to enhance the graph attention network (GAT). This approach effectively mitigates the over-smoothing problem and vanishing gradient phenomenon in deep networks. Experimental results demonstrate that the proposed method outperforms traditional machine learning models in average accuracy across 10 random trials. Furthermore, the model exhibits extremely low variance. These results confirm the superior robustness and generalization capability of the framework.

To enhance the operational efficiency of rural multi-energy microgrids, this paper constructs an integrated electricity-gas-heat-cold multi-energy coupled optimal scheduling model. The optimization framework is formulated as a multi-objective problem, targeting the minimization of operational costs, the maximization of comprehensive benefits, and the reduction of the multi-energy loss of power supply probability(LPSP). To solve this complex problem, an improved Multi-Objective Artificial Hummingbird Algorithm(MOAHA)is proposed, incorporating diverse strategies to enhance its global search capability and convergence performance. Simulation results demonstrate that the proposed algorithm achieves scheduling schemes with superior comprehensive performance. It significantly improves the system's economic efficiency, environmental sustainability, and power supply reliability, thereby providing an effective solution for the low-carbon and high-efficiency scheduling of rural multi-energy systems.

Abstract Shipboard Direct Current (DC) propulsion systems exhibit complex dynamic characteristics spanning multiple time scales, ranging from microsecond-level power electronic switching to second-level electromechanical transients. Traditional real-time simulation methods often struggle to balance high-fidelity switching details with large-scale system complexity. This paper proposes a high-fidelity real-time cooperative simulation platform based on an FPGA-CPU heterogeneous architecture. A mapping strategy between time scales and computing architectures is established: microsecond-level ultra-fast switching models, including a 24-pulse rectifier and a Neutral Point Clamped (NPC) inverter, are deployed on the FPGA’s parallel logic to capture high-frequency harmonics and switching dynamics. Concurrently, millisecond-level electromechanical models, such as the six-phase Permanent Magnet Synchronous Motor (PMSM) and propeller load, are executed on a multicore CPU. To ensure temporal consistency across heterogeneous platforms, a deterministic data exchange mechanism based on Direct Memory Access (DMA) is implemented to resolve cross-platform synchronization issues. Furthermore, enabling technologies, including Vector Space Decoupling (VSD) modelling and power feedforward cooperative control strategies, are integrated to enhance simulation fidelity under complex operating conditions. Experimental results demonstrate that the proposed heterogeneous architecture effectively overcomes the simulation bottlenecks induced by multi-time scale coupling. It accurately reproduces system-level harmonic oscillations and fault propagation, significantly improving the R&D efficiency and verification reliability for modern marine integrated power systems.

With the continuous expansion and complexity of Integrated Power Systems (IPS) on ships, even localized faults can trigger cascading failures. Therefore, reliable and real-time fault diagnosis is a critical component for ensuring navigation safety. This review highlights major developments in IPS fault diagnosis over the past decade (since 2016, covering 79 references), with a focus on four representative methodological frameworks: modelbased, data-driven, knowledge-based, and hybrid-based. The review also outlines emerging technologies and their current applications. Related research has been preliminarily validated on various full-scale ships, demonstrating promising engineering applicability. In reviewing the evolution of diagnostic methods, the paper identifies three pressing challenges: difficulties in data acquisition and imbalance, limitations in sensing and onboard computing capacity, and reduced diagnostic reliability under coupled fault scenarios. To address these issues, this review outlines future research directions, including enhancing data quality and intelligent preprocessing, leveraging data augmentation, transfer learning, and unsupervised modeling to mitigate sample scarcity and label deficiency, and developing modeling and inference frameworks tailored to coupled fault conditions. This review aims to provide theoretical guidance and practical reference for the design and optimization of fault diagnosis techniques for integrated power systems on ships.

The vibration signals of Electro-Hydrostatic Actuators (EHAs) exhibit strong non-linearity and non-stationarity, particularly under complex coupling mechanisms, making the extraction of intrinsic fault features computationally challenging. Conventional deep learning approaches often lack mathematical interpretability and struggle to decouple superimposed fault signatures from incomplete datasets. To address these issues, this paper proposes the Enhanced Continuous Wavelet Transform Capsule Network (ECWTCN), an intelligent decoupled diagnosis framework designed for multiscale signal analysis. The architecture integrates a wavelet-kernel convolution layer to extract physically interpretable time-frequency features across multiple scales, effectively capturing transient impulses associated with incipient faults. Furthermore, a novel maximized aggregation routing algorithm is introduced to optimize the dynamic routing process, enhancing global feature aggregation. A distinct advantage of the ECWTCN is its capability to generalize distinct fault patterns, enabling the identification of unseen compound faults by training exclusively on normal and single-fault samples. Comparative experiments show that the proposed method delivers strong multi-label classification performance under operating condition A, achieving a Subset Accuracy of 93.7% and a Label Ranking Average Precision of 0.998. Complexity analysis further confirms the method's efficiency in terms of FLOPs and parameter size. This work presents a robust, lightweight, and mathematically interpretable solution for the analysis of complex signals in high-reliability equipment.

The power equipment defect text is important data generated by the operation and maintenance of the power grid system. It is characterized by having many professional words from the power field and high complexity. During model training, there are some problems, such as difficulty in semantic understanding, gradient disappearance, negative information loss, and data imbalance, which hurt the text quality and defect analysis effect. To deal with these issues, this paper proposes a ULF-BI-LSTM text quality improvement algorithm integrating UCNN, LeakyRelu activation function, and Focal Loss function. Then, we correctly separate professional vocabulary, delete invalid vocabulary, and normalize the object description by text data preprocessing. Finally, we fill in missed data and correct the wrong data using the ULF-BI-LSTM algorithm. Experimental results show three improvement strategies that effectively improve the accuracy, precision, F1, and recall of the algorithm. The proposed algorithm outperforms the mainstream algorithm TextCNN, SVM, and BI-LSTM, which alleviates the above problems.

Large pulsed degaussing coils usually require full-cycle thermal simulation to determine the temperature rise limit, but the associated computational cost is extremely high and severely restricts engineering design efficiency. To address this issue, this paper proposes an efficient evaluation method based on sparse simulation and model extrapolation. A first-order lumped-parameter thermal model is first established and rapidly calibrated using finite-element results from only a few key nodes, so that the dynamic thermal behavior of the system can be captured with limited simulation effort. A quantitative guideline for selecting sparse sampling nodes is further introduced by partitioning the pulse sequence into calibration, validation, and equilibrium-oriented zones and assigning explicit range and error criteria to each zone. The calibrated model is then employed to predict the temperature evolution over the entire operating cycle, and the temperature rise limit is finally obtained by solving the thermal equilibrium condition. An electro-thermal coupled degaussing coil model is used as a case study for validation. In addition, a sensitivity analysis with respect to the equivalent thermal resistance, equivalent thermal capacitance, and ambient temperature is performed to examine the robustness and transferability of the proposed framework. The results show that the proposed method achieves a prediction error of less than 5% while reducing the computational cost by about one order of magnitude compared with conventional full-cycle simulation. The sensitivity analysis further indicates that the saturation temperature is mainly governed by the equivalent thermal resistance and ambient temperature, whereas the equivalent thermal capacitance primarily affects the transient heating rate. The proposed method therefore provides an accurate and efficient tool for evaluating the temperature rise limit of pulsed degaussing coil systems and offers a reproducible practical solution for thermal design and safety margin assessment of electrical equipment under massive pulsed operating conditions.

The reliability of the electrical systems of ships, as the main carriers of goods worldwide, is critical to maritime safety. Equipment malfunctions, if not detected and addressed promptly, it may lead to serious safety accidents. While various fault diagnosis methods exist, they often struggle with insufficient accuracy and weak noise immunity in handling ship multivariate time series data. This study aims to develop a more accurate fault diagnosis method for ship power grids. To achieve this objective, this study proposes a novel approach combining complex network, intelligent algorithm and machine learning (Barabasi-albert model-enhanced genetic algorithm for optimizing LGBM, BGL) for ship electrical grid fault diagnosis. The methodology consists of three main steps: feature extraction using wavelet transform (WT), feature subset selection through complex networks combined with intelligent algorithms and node importance algorithm, and finally, fault diagnosis implementation using a machine learning model. The experimental validation is carried out on a simulation dataset containing normal and multi-class fault state ship power grids. The results show that BGL can accurately identify faulty and normal states and distinguish different fault types. When compared with 14 benchmark algorithms and 3 LGBM-based improved algorithms, BGL improves the accuracy, precision, recall, and F1-Score by at least 0.72 %, 0.68 %, 0.90 %, and 0.90 %, respectively, in a noise-free environment, and by at least 1.86 %, 2.09 %, 2.08 %, and 2.06 % in a noisy environment. It is shown that BGL can effectively capture the characteristic patterns of different fault types with strong noise resistance and generalization ability, and provides a general framework for fault diagnosis of multivariate time-series data in ship power grids.

As a key component in the motor system, the stability and reliability of the entire system is directly influenced by the operating condition of the bearings. With the improvement of industrial automation, the importance of diagnosis of bearing faults technology is becoming more and more prominent. In this paper, for the diagnosis problem of bearing fault, the diagnostic efficacy of three feature extraction methods, namely, time-domain (TD), Discrete Wavelet Transform (DWT), and Fast Fourier Transform (FFT), is systematically investigated. For these three feature extraction methods, this paper proposes six fault diagnosis algorithms, which are decision tree (DT), random forest (RF), support vector machine (SVM), convolutional neural network (CNN), ensemble learning (EL), and extreme learning machine (ELM). The results indicate that the DWT-based feature extraction method is the best overall, achieving a diagnostic accuracy of 99.79% under the EL; the time-domain features provide stable diagnostic results, and the RF achieves an accuracy of 98.83%; whereas the FFT-based features obtain an accuracy of 97.58% under the ELM. Further analyses show that different feature extraction methods affect different algorithms to different degrees, and there is a significant difference in the sensitivity of the algorithms to the feature extraction methods: the traditional machine learning algorithm performs well with DWT features, while the deep learning method obtains better results with time-domain features. The research results provide a theoretical basis for the optimisation of bearing fault diagnosis algorithms, and at the same time provide principle support for the selection of systematic feature extraction methods.

As a core component of both the ship propulsion system and mission-critical equipment, shipboard motors are undergoing a technological transition from traditional fault diagnosis to multi-physical-field collaborative modeling and integrated intelligent maintenance systems. This paper provides a systematic review of recent advances in shipboard motor fault monitoring, with a focus on key technical challenges under complex service environments, and offers several innovative insights and analyses in the following aspects. First, regarding the fault evolution under electromagnetic–thermal–mechanical coupling, this study summarizes the typical fault mechanisms, such as bearing electrical erosion, rotor eccentricity, permanent magnet demagnetization, and insulation aging, and analyzes their modeling approaches and multi-physics coupling evolution paths. Second, in response to the problem of multi-source signal fusion, the applicability and limitations of feature extraction methods—including current analysis, vibration demodulation, infrared thermography, and Dempster–Shafer (D-S) evidence theory—are evaluated, providing a basis for designing subsequent signal fusion strategies. With respect to intelligent diagnostic models, this paper compares model-driven and data-driven approaches in terms of their suitability for different scenarios, highlighting their complementarity and integration potential in the complex operating conditions of shipboard motors. Finally, considering practical deployment needs, the key aspects of monitoring platform implementation under shipborne edge computing environments are discussed. The study also identifies current research gaps and proposes future directions, such as digital twin-driven intelligent maintenance, fleet-level PHM collaborative management, and standardized health data transmission. In summary, this paper offers a comprehensive analysis in the areas of fault mechanism modeling, feature extraction method evaluation, and system deployment frameworks, aiming to provide a theoretical reference and engineering insights for the advancement of shipboard motor health management technologies.

As the core equipment of power distribution management, smart meter is becoming the control center of intelligent house. Accurately predicting the replacement number of smart meters is of great significance to effectively reduce the operation and maintenance cost of smart meters and greatly improve the operation security of smart grid. This paper focuses on how to accurately predict the replacement number of smart meters with few samples and insufficient data statistics, and studies the improved gray-Markov model. Firstly, an improved grey model based on weighted average weakening buffer operator modified data predicts the number of smart meter replacements. Markov model is used to optimize the prediction results of the improved grey model, and maximize the prediction accuracy of the replacement number of smart meters. The actual replacement case of smart meters further verifies the validity of this model in predicting the replacement quantity accurately.

Active Neutral Point Clamped three-level converters find extensive application in high-power scenarios. However, under conventional modulation strategies, they suffer from severe imbalance in power dissipation and thermal generation between inner and outer tubes, thereby compromising system reliability. To address this issue, novel modulation techniques that do not increase hardware complexity require investigation. This paper provides an in-depth analysis of the operating principles and multiple typical operating modes of the ANPC three-level converter. Building upon this foundation, a hybrid modulation method is proposed that cyclically alternates and sequentially cascades two drive modulation strategies with complementary loss characteristics. This approach achieves dynamic transfer and balancing of switching device loss stresses by rotating high-loss components across different time periods. Simulation results demonstrate that, compared to conventional single-modulation strategies, the proposed hybrid modulation scheme effectively balances power dissipation between the inner and outer tubes, significantly improving the system's temperature distribution without necessitating increased switching frequencies or additional hardware. This provides an effective solution for enhancing the operational reliability and engineering applicability of high-power ANPC converters.