Accurate and rapid detection of characteristic fault gases such as H2 and CO is essential for transformer condition monitoring and early fault diagnosis. However, the efficient identification of high-performance gas-sensing materials remains challenging due to the high cost and low efficiency of conventional trial-and-error screening methods. In this work, an AI-assisted screening strategy combining machine learning and density functional theory (DFT) calculations is employed to identify suitable two-dimensional materials and metal-doped systems for transformer fault gas detection. Through high-throughput screening, Pd-MoS2 is identified as the optimal candidate, exhibiting adsorption energies of-0.65487 eV for H2 and-0.84679 eV for CO, which fall within the desirable range for reversible gas sensing. Experimental results demonstrate that the Pd-MoS2 sensor shows high sensitivity, fast response and recovery times (22 s/29 s for H2 and 24 s/32 s for CO), as well as good repeatability and stability. By integrating data-driven screening with experimental validation, this study provides a systematic and efficient pathway for the rational design of gas-sensing materials for transformer monitoring applications.

Precise, dynamic, and in situ characterization of material microstructures is essential for understanding failure mechanisms and tailoring material properties. In polymer insulating materials, prolonged electrical stress induces the formation of microscale dendritic defects, known as electrical trees. However, existing characterization techniques such as scanning electron microscopy (SEM) cannot capture the dynamic three-dimensional (3D) evolution of these trees under strong electric fields, thereby limiting mechanistic understanding and materials design. Here, we report that electric field excitation induces stimulated autofluorescence within polymer microstructures, enabling an in situ, non-destructive, and dynamic characterization approach. This technique achieves submicron spatial resolution, allowing for the first real-time visualization of electrical tree growth and morphological evolution under high electric fields. It demonstrates broad applicability across various polymer types and aggregation states. Using this method, we reveal the suppression effect and spatial action mechanism of SiO2 microfillers on electrical tree propagation. The observed fluorescence mechanism arises from electric field-induced molecular rearrangement in defect regions, leading to autofluorescence upon laser excitation. As the first method capable of dynamically characterizing microstructures under strong electric fields, it complements conventional tools such as SEM and optical microscopy, offering a new pathway for addressing key challenges in dynamic microstructural analysis.

This two-part paper proposes a device that integrates dc fault isolation with ac/dc fault ride-through capacities. A four-terminal MTDC simulation system and a simplified low-voltage hardware experimental platform are established for demonstrating the functions of clearing the dc fault current and dissipating surplus power. The experimental results are compared with those of a simulation to analyze the feasibility of the proposed device and verify the reliability of the simulation. Using the simulation system, the proposed device competes against representative dc circuit breaker (DCCB) schemes with dc chopper schemes. The proposed device can effectively clear dc fault current and dissipate the surplus power. The proposed device lowers the cost of components by at least 33.7% from those of independently configured DCCBs with dc choppers.

In the flexible multiterminal dc (MTDC) transmission systems, during dc short-circuit faults on the dc overhead line (OHL) and ac short-circuit faults on the ac network of the grid-side converter, dc circuit breakers (DCCBs) and dc choppers are reliable solutions to handle faults, as they ensure uninterrupted operation. However, their high costs challenge their widespread applicability. In this two-part paper, a device with high performance and low cost is proposed. Part I describes the principles of handling ac and dc short-circuit faults, and provides a theoretical analysis of the performance of dc fault isolation (DC FI) and ac/dc fault ride-through (AC/DC FRT). In Part II, the feasibility of the proposed device is validated on a four-terminal MTDC simulation system and a scaled-down hardware experimental platform. The performances and cost-effectiveness of the proposed device with the optimal configuration are compared with those of representative DCCB and dc chopper schemes. The theoretical analysis, simulation/experiment verification, and comparison showcase the potential of the device for enhancing the cost-effectiveness of MTDC systems.

Achieving high-precision, in situ, three-dimensional (3D) characterization of microscale defects within material interiors and at interfaces through non-destructive methods has remained a persistent challenge. Herein, this study reports a novel phenomenon: electric fields induce autofluorescence in various intrinsically non-fluorescent polymers. This enables non-destructive, in situ 3D imaging of surface and internal microstructures and facilitates real-time characterization of dynamic evolution of internal electrical tree defects in polymers. Studies integrating quantitative fluorescence spectroscopy with molecular dynamics and quantum chemical calculation modeling demonstrate that the application of moderate-strength electric fields induce conformational changes in polymer molecules, thereby generating fluorescent emission. Under laser excitation, these molecules transition from the ground state to excited states, releasing fluorescent photons during spontaneous relaxation. By precisely capturing these photons and mapping their spatial relationship with radiation sites, the 3D structural characteristics of materials within the excited radiation regions can be reconstructed. This approach provides a robust, non-destructive tool for polymer microstructure imaging, representing a significant complement to established techniques such as scanning electron microscopy and X-ray computed tomography. Validated across glassy, rubbery, and semi-crystalline polymers, it exhibits broad applicability in polymer electronics, composite durability assessment, and failure mechanism investigations, thereby expanding the scope for advanced material characterization.

Ensuring the long-term reliability of polymeric insulators under high electric fields remains a major challenge in modern power and electronic systems. Although thermochromic/self-reporting strategies have been explored for dielectric diagnostics, most existing approaches primarily provide qualitative or macroscopic indications and cannot capture the transient temperature evolution within the micro-regions where electrothermal damage initiates. Here, we present a high-resolution quantitative mapping framework for in-situ thermal dynamics during electrical treeing using a thermochromic microcapsule–epoxy composite. Thermochromic microcapsules are uniformly dispersed within the epoxy matrix to act as local optical transducers, converting temperature surges induced by partial discharges into quantifiable color variations. By integrating high-speed optical imaging with image-processing algorithms, the transient temperature field during electrical treeing can be reconstructed with sub-millimeter spatial sampling (≈14 µm per pixel) and millisecond-level temporal resolution. This composite design enables a direct and quantitative correlation between the colorimetric response and local thermal dynamics, providing a means to visualize and analyze the critical thermodynamic stages of electrical tree propagation. The developed intelligent dielectric composite offers a new approach for in situ monitoring of electrothermal degradation and represents a promising strategy for predictive maintenance in power insulation systems.

This article proposes a physics-informed multi-scale adaptive graph neural network (PI-MSA) for rapid event detection, localization, and classification (ED-L-C) of single and multi-source coupling events in power systems to prevent large-scale failures. Prior work decomposes ED-L-C tasks into separate classification subtasks. Instead, this work frames ED-L-C as a unified regression problem through a spatial-aware and task-coupled three-dimensional target formulation, enabling one-shot inference of all abnormal events and their class probabilities at both node and edge levels. By incorporating physical grid topology and shortest-path admittance encoding through parallel multi-scale networks, this unified one-stage pipeline enables joint optimization and learning of shared graph features as well as progressive inter-task dependencies. Experimental validation on five power systems of varying scales and configurations demonstrates that PI-MSA achieves over 75% mean average precision and 90% localization accuracy across various scenarios, outperforming traditional methods by over 40% and exceeding state-of-the-art methods by at least 5%. The model maintains robust and accurate event analysis under topology changes and reduced observability conditions while achieving computational efficiency at 225 frames per second, enabling real-time monitoring and analysis for modern power systems.

In the field of lightning protection in power systems, improving the accuracy of primary side lightning overvoltage inversion and reconstruction of capacitive voltage transformers (CVTs) is of great significance. Therefore, this article innovatively proposes a lightning overvoltage inversion method based on the CVT black box inverse model. In the research process, firstly, scattering parameter measurement technology is used to accurately obtain the frequency domain response characteristics of CVT in the frequency range of 0.1 Hz–10 MHz, and the vector matching method is used to construct its voltage transfer positive function and analytical inverse function; Secondly, based on the voltage waveform data of the CVT secondary side distortion measured on site, corresponding algorithm programs are written according to the inverse function to achieve accurate inversion and reconstruction of the primary side lightning overvoltage waveform. Experimental verification shows that this method significantly improves the inversion accuracy, with the measured peak error of lightning impulse wave greatly reduced from 20.7

This article proposes a short-circuit model of transformers for analyzing internal short-circuit faults. The model accurately captures the coupling between subwindings using a coupled leakage inductance matrix. This matrix can be directly computed from the self and mutual inductance matrix, which is typically provided by the manufacturer. In addition, the model includes linear resistances, linear inductances, and ideal transformers, enabling direct use in most electromagnetic transient simulation software. The model is validated with a 10 kV single-phase test transformer and applied to analyze internal short-circuit faults in a +/- 800 kV converter transformer and a 220 kV three-phase five-limb transformer in the field. The analytical results closely align with the field recordings, confirming the model's accuracy and practical applicability.

Electric fields are fundamental physical quantities that describe electromagnetic phenomena. Achieving passive, high-precision, vector-resolved sensing is essential for both fundamental electromagnetics and the precise regulation of electrified systems, yet it remains a central technical challenge. Inspired by the bioelectrosensory mechanism of treehoppers, this study designs and realizes a passive electroluminescent vector electric field sensor (PELVEFS) based on a bioinspired dielectric heterostructure. The three-dimensional high-permittivity architecture within the device couples with the external electric field and reshapes it into a direction-dependent surface field distribution, which subsequently drives an electroluminescent (EL) coating to emit optical signals that encode both the magnitude and the direction of the vector field. A shared-weight one-dimensional convolutional neural network inversion scheme then reconstructs the full vector information from the EL spectra, enabling the precise measurement and decoupling of field strength and direction without any external power source. Experimental results demonstrate that the PELVEFS provides a nonoverlapping omnidirectional response across a wide dynamic range of 0.20-1.00 kV/mm, achieving a mean absolute error of 0.015 kV/mm and a mean relative error of 2.2% for field strength, as well as a mean absolute angular error of 3.62° for direction. This work establishes an integrated approach for vector electric field sensing in complex electromagnetic environments, encompassing the entire process from the sensing mechanism and device architecture to data-driven information reconstruction.

Microparticulate fillers constitute an effective approach for mitigating tree-like microscale damage and consequent failure in polymeric insulating materials during prolonged service. Nevertheless, existing characterization techniques remain inadequate for directly resolving the internal evolution of such damage under strong electric fields and elucidating its interactions with microparticulate fillers. Consequently, theoretical and experimental evidence supporting performance regulation in microscale dielectric materials remains insufficient. In this study, the electric-field-induced photoluminescent autofluorescence generated within polymer materials was utilized. On the basis of this mechanism, an evolution observation platform was developed for autofluorescence imaging of electrical trees. This platform enabled time-lapse in situ 3D imaging of microscale electrical tree damage evolution under strong electric fields in both SiO2-doped epoxy composites and pure epoxy. Geometric skeletons extracted from the electrical trees were used to construct a multidimensional, time-dependent feature-parameter framework encompassing “point–line–surface–volume” characteristics. This framework quantitatively differentiated the electrical tree growth behaviors exhibited by the two material systems. Complementary phase-field simulations further corroborated the experimentally observed evolution patterns and investigated the suppression mechanism associated with SiO2 microparticles. Through the coupled effects of physical obstruction and electric-field regulation, electrical trees transitioned from sparse dendritic structures to multibranched, cluster-like morphologies that promoted a more uniform electric-field distribution and even led to growth stagnation. This study establishes a quantitative characterization methodology that facilitates the evaluation of the degradation behavior and insulation stability of dielectric materials, with potential applications in insulating-material optimization and defect diagnosis.

Large-scale photovoltaic (PV) systems with numerous power electronic units impose heavy computational burdens on conventional electromagnetic transient (EMT) simulation. This paper presents a GPU-oriented implementation framework for fast electromagnetic transient (EMT) simulation of photovoltaic (PV) systems. Each PV unit is reduced to a port-level Norton equivalent at every fixed EMT time step, enabling the global network to be solved as a 3-node circuit while preserving detailed device and control dynamics inside each unit. The computation is organized as a three-stage pipeline: circuit preparation (constructing per-unit Norton parameters and back-substitution maps), small-system network solution, and massively-parallel state update. To avoid per-step nonlinear iterations of the PV array, the array equation is converted into an explicit Norton-type injection by evaluating/linearizing it using previous-step quantities, consistent with companion-model discretization of dynamic elements. All EMT states and history terms are stored and updated in GPU global memory with a double-buffer scheme, yielding a scalable and reproducible implementation for large PV systems.

In dual-active-bridge (DAB) converters, sudden power variations generate transient DC bias currents that threaten reliable operation. Therefore, elimination strategies are required. Existing elimination strategies generally assume that the high-frequency transformer operates in the linear region and neglect the core nonlinearity. Ideally, transformers should be designed close to their maximum utilization to minimize size, weight, and cost. In this context, even if the transformer operates in the linear region, its nonlinearity cannot be ignored. Under such conditions, the magnetizing branch produces significant DC bias currents, leading to the failure of existing elimination strategies. To solve this challenge, this article proposes a transient DC bias elimination strategy for single-phase-shift-controlled DAB converters considering core nonlinearity. The strategy inserts an additional modulation cycle to ensure seamless mode transitions. Experimental results demonstrate that the proposed strategy eliminates transient DC bias currents more effectively than existing strategies and confirm the necessity of considering core nonlinearity in the operation of DAB converters.

Ground Flash Density (GFD) serves as a critical parameter for evaluating the lightning strike risk of transmission lines.  It is typically derived from the statistical analysis of data monitored by Lightning Location Systems. However, constrained by the limited sample size of ground flashes and the detection accuracy of Lightning Location Systems, the spatial resolution of GFD is relatively low. To enhance the spatial resolution of GFD along transmission line corridors, this paper proposes a downscaling method of GFD based on the Monte Carlo method considering micro-terrain conditions. For a given original grid area, considering the influence of terrain on lightning, the lightning strike point is determined by the lightning leader fractal model, thereby generating a simulated dataset of ground flash samples dataset. The original grid is further divided into multiple subunits, and the GFD of subunit is calculated by combining the grid average ground flash density statistics, ultimately yielding the high spatial resolution ground flash density. The proposed method is used to calculate GFD in the actual grid area and the effectiveness of the method is verified.

Aging-induced evolution of crosslinking density critically determines the long-term performance and failure behavior of epoxy coatings. However, conventional characterization methods are generally destructive, indirect, and incapable of quantitatively monitoring microstructural degradation in situ. In this work, an amino-terminated tetraphenylethene derivative with aggregation-induced emission (AIE) characteristics, 4′,4″‘,4″“‘,4″““‘-(Ethene-1,1,2,2-tetrayl) tetrakis(([1,1′-biphenyl]-4-amine)), (ETTBA), was rationally designed and synthesized as a fluorescent probe for self-assessing the aging state of epoxy coatings. Owing to its reactive amino groups, ETTBA was covalently incorporated into the three-dimensional epoxy crosslinked network, endowing the coating with stable fluorescence response while maintaining excellent service performance. Benefiting from the restricted intramolecular rotation mechanism of AIE luminogens, the fluorescence intensity exhibited high sensitivity to aging-induced variations in crosslinking density. At an ultralow loading of 0.20 wt%, the coating achieved a fluorescence quantum yield of 47.02% and a large Stokes shift of 105 nm. Accelerated thermo-oxidative aging results revealed a strong positive correlation between fluorescence intensity and crosslinking density. Furthermore, the established quantitative prediction model demonstrated excellent fitting performance (R2 = 0.99891) and high prediction accuracy, with a maximum relative error of only 2.22%. This study provides a novel molecular-level optical strategy for nondestructive and quantitative monitoring of epoxy coating aging.

The heat generated during the high-power energy conversion requires that polymer dielectrics have superior capacitive energy storage performance at high temperatures. However, the thermal stability of polymer dielectrics relies on the dense packing of molecular chains, whereas efficient dipolar orientation requires sufficient free volume between segments. This inherent coupling conflict has long been a bottleneck for advancing hightemperature energy storage in polymers. In this work, we report a class of polymer dielectrics incorporating mechanically interlocked molecules (MIM) that effectively decouple thermal stability from high-temperature energy storage capability. Cyclic crown ethers are anchored to the polyetherimide backbone via noncovalent mechanical bonds, providing relaxation space for dipole orientation while spatially confining chain segments to maintain dense molecular packing. The resulting MIM polymer exhibits a resistivity at 200 degrees C that is more than two orders of magnitude greater than that of conventional polyetherimides, achieving a discharge energy density of 6.53 J/cm3 with an energy efficiency of 90 %, which is 1941 % higher than that of the classic hightemperature polymer medium polyetherimide. Furthermore, we demonstrate that this strategy is broadly applicable to different polar polymer systems, highlighting its potential as a general design paradigm.

Dielectric materials are essential in electrical equipment, yet environmental moisture infiltration disrupts internal electric field uniformity, degrading dielectric performance and potentially causing insulation failure. Traditional humidity monitoring methods are unsuitable for live dielectric systems and struggle to achieve the high-resolution spatial localization of microscale moisture. Inspired by the biological response of reef-building corals, we propose a field-activated microsensor (FAM) based on the electroluminescence effect. Integrating FAMs endows dielectric materials with real-time microscale moisture sensing capabilities. Through precisely designed micro-nano structures, FAMs sustainably respond to local electric field distortions induced by moisture. Furthermore, FAMs autonomously construct cross-linked interpenetrating networks to optimize the intrinsic properties of dielectric material. Ultimately, FAMs enable the real-time, in situ, and high-precision detection of microscale moisture under high-humidity and high-field conditions, demonstrating broad application prospects for the condition monitoring and early fault warning of energized dielectric systems.

Dead-time control is essential for modular multilevel converters (MMCs), but it negatively impacts MMC performance. To support the development of control strategies to mitigate the adverse effects of dead-time, electromagnetic transient (EMT) simulations are crucial for analyzing MMCs’ dead-time behavior and developing strategies to mitigate these effects. However, simulating the impact of dead-time in high-level MMCs remains a challenge. The complex modular cascaded circuit significantly slows simulation speed, while the freewheeling conduction of submodule diodes during dead-time disrupts the integrity of submodules in each bridge-arm. This makes it difficult to represent the circuit as a unified Thevenin equivalent for simplification. To address the issue, the dead-time effect is modeled using a diode-H-bridge in this paper. Submodules are categorized into those affected by the dead-time effect and those not. This paper also proposes the capacitor state mapping approach, referred to as “twin mapping method”, to restore the submodules’ behavior during and outside the dead-time, eliminating the isolation of submodules and enabling the application of the Thevenin equivalent theorem. Finally, a Thevenin equivalent model (EM) is developed and compared with a detailed model (DM) and state-space model (SSM). PSCAD/EMTDC simulations demonstrate that the proposed EM effectively captures dead-time spikes and notches while significantly accelerating simulation speed.

Recognition methods of electromagnetic transients (EMT) have been widely used in power systems with the assumption that training and testing data are drawn from the same probability distribution. However, that assumption is hard to satisfy in industrial applications because the distribution of measured EMT testing data generally changes over time. The performance of these methods gradually deteriorates with the distribution shift. The phenomenon limits application of EMT recognition methods. Therefore, this paper proposes a transfer learning-based recognition network (TLRN) for EMT to break the limitation. It consists of a feature extractor, EMT recognizer, domain recognizer, and maximum mean discrepancy (MMD). The feature extractor is constructed to learn features of EMT automatically. The domain recognizer and MMD make features learned by the feature extractor domain invariant. Based on domain invariant features, the EMT recognizer achieves accurate EMT recognition, despite the distribution discrepancy between EMT training and testing data. TLRN maintains satisfactory EMT recognition performance by updating periodically with an unsupervised learning strategy. Using EMT datasets measured from different substations, scenario experiments, and experiment comparisons are conducted, and the recognition performance of the proposed TLRN is demonstrated.

Microcapsule-based self-healing is an effective strategy for addressing multiscale damage in polymeric materials. However, conventional methods exhibit limited repair capabilities for large-scale damage (at the submillimeter level) under low microcapsule doping concentrations (<5 wt%). This study proposed a dual self-healing mechanism through the development of a double-shell microcapsule/modified epoxy resin composite (DMMEC). The topological reconstruction of the epoxy matrix was achieved by incorporating a 5-wt% terephthalaldehyde crosslinker, endowing the system with dynamic covalent adaptive network characteristics. Polyurethane/polyaniline (PU/PANI) double-shell microcapsules with photothermal conversion enabled infrared-triggered localized viscoelastic matrix flow, synergistically inducing damage propagation through minimal repair agent release. Notably, permittivity disparities at the microcapsule-matrix interface were found to guide the targeted development of micrometer-scale electrical tree damage, achieving efficient microscale electrical tree repair at a low doping concentration (2 wt%). Moreover, when subjected to large-scale mechanical scratch damage (2-mm length x 100-mu m width), recovery rates of 91.3 % in mechanical strength and 90.4 % in electrical strength were attained. The proposed method overcomes the inherent trade-off between healing efficiency and matrix performance degradation in low-load microcapsule systems, thereby providing a novel paradigm for efficient multiscale damage restoration in polymeric materials.