Conference Agenda

The reliability of solid-solid dielectric interfaces in HVDC insulation systems is a critical concern due to their vulnerability to degradation and breakdown under DC stress. This paper investigates a defective epoxy-silicone interface representative of internal regions in composite insulators, subjected to DC excitation under positive and negative polarities using a stepwise energization scheme. The study combines partial discharge (PD) measurements, optical inspection of interfacial damage, and time-frequency (TF) analysis. The results show a clear polarity dependence. Negative polarity produces higher PD repetition rates and larger accumulated charge, whereas positive polarity leads to breakdown at a slightly lower voltage and more dispersed interfacial damage despite lower overall PD activity. TF analysis reveals similar spectral characteristics for both polarities, suggesting comparable discharge mechanisms. These results highlight the importance of polarity-aware interpretation of PD measurements for the diagnostic assessment and reliability evaluation of solid-solid interfaces in HVDC insulation systems.

Wide bandgap (WBG) devices enable higher efficiency and more compact power module designs. However, the associated high electric fields and rapid voltage transients impose substantial stress on insulation systems. Under these electrical stresses, partial discharge (PD) originates from structural defects (e.g., voids, triple junctions, and surface irregularities) and accelerates insulation aging. While distinct phase-resolved partial discharge (PRPD) patterns have been identified under sinusoidal excitation, the lack of a well-defined phase reference in fast-switching power electronic systems makes PD source classification ambiguous and limits accurate diagnostics. To address this limitation, this paper develops a PD source classification approach under power electronic switching voltage using high dv/dt excitation applied to solid insulators with artificially generated surface and internal PD defects. An Amplitude-Width-Area (AWA) approach is then used to construct PD patterns directly from measured pulse signals, resulting in distinguishable feature distributions for surface and internal. Using the resulting AWA image dataset, an InceptionV3 convolutional neural network (CNN) is trained to classify PD sources and achieves over 99% classification accuracy, demonstrating reliable PD source identification under switching voltage conditions.

This paper presents the conceptual design, multiphysics modeling, and partial experimental validation of a terahertz-based signal chain for detecting space charge in polymeric insulation materials. The proposed framework establishes a complete conversion pathway from terahertz excitation through charge redistribution, transient stress formation, elasto-optic modulation, and optical readout, forming a coherent mechanism that links the microscopic evolution of space charge to measurable macroscopic optical signals. A coupled physical model is developed to describe each conversion stage, focusing on the dynamic interactions between the electromagnetic, mechanical, and optical domains. The analysis indicates that localized electric field perturbations can induce stress waves in polymeric dielectrics, and that these stress waves lead to measurable refractive index variations through the photoelastic effect. This indirect sensing principle provides a potential approach for detecting internal charge activities without electrical contact or destructive sampling.To verify the feasibility of the core transduction mechanism, a laboratory setup is developed to reproduce the pressure-wave-induced photoelastic effect using a controlled mechanical excitation source instead of terahertz irradiation. The optical phase response is captured through an interferometric detection system, and the measured phase variations show good consistency with theoretical predictions of the stress-optic model. These results confirm that the optical subsystem can accurately capture weak stress perturbations corresponding to space-charge-related field variations. Although the full terahertz-induced pressure wave has not yet been realized, the partial experimental validation substantiates the functional viability of the signal chain’s key elements and provides essential calibration data for future terahertz integration.The proposed study demonstrates that the terahertz-based signal chain is both theoretically sound and experimentally feasible in its fundamental components. It establishes a methodological foundation for future implementation of non-contact, high-resolution, and physically interpretable diagnostic systems for polymeric insulation. The presented work provides quantitative insight into the coupling between terahertz excitation, mechanical stress, and optical modulation, offering a reproducible pathway toward advanced sensing technologies in insulation condition monitoring and space charge diagnostics.

PD is measured for 2 gases CO2 and and 80% N2 + 20% O2 (synthetic air). BDV is measured in 3 gases between 1.5 and 6 bar: 62% N2/38% CO2, 100% CO2, and air. A needle plane arrangement is used with needle lengths: 1.5 mm, 3 mm, and 10 mm, and gap distances: 18 mm, 25 mm, and 45 mm. Measured partial discharge inception voltage (PDIV) and partial discharge extinction voltage (PDEV) are compared to streamer inception predicted values. Streamer inception predictions are compared to BDVs to assess detectability of protrusions in each gas. Corona stabilization is observed to increase BDV thus increasing the predicted ratio of BDV to PDIV and allowing for safer detection of protrusions. Results show an agreement between PDIV and predicted streamer inception. CO2 allows for the best detection of protrusions 3 mm and smaller, most times performing better than the other gases. Pure CO2 without O2 shows better detectability by itself than with O2 added. The 62% N2/38% CO2 mixture shows the best withstand strength at around 3 bar to 5 bar for 10 mm protrusions. PD behavior of 62% N2/38% CO2 could not be assessed

High voltage motors used in industrial drives and power generators require high insulation reliability to prevent failures. Insulation failure is mainly caused by degradation of the mica-epoxy main insulation in stator coils. Recently, however, failures due to degradation of the semiconductive surface layer have also been reported. To maintain reliable operation of these critical machines, diagnostic methods that can detect degradation in the semiconductive layer are required. We have developed a machine learning model to identify semiconductive layer degradation by distinguishing it from main-insulation degradation based on partial discharge (PD) signals. The model uses the oscillation frequencies of PD signals as features without relying on phase information. In this study, PD signals were obtained from samples simulating both types of degradation to train a preliminary model. The model achieved high classification accuracy, demonstrating its feasibility.

To validate this approach, experimental measurements and model training were conducted as follows. Artificial insulation specimens were prepared to simulate two degradation modes: surface deterioration in the semiconductive layer and internal degradation in the main insulation. AC voltages with peak values up to several kilovolts were applied to the specimens. PD signals were detected using a high frequency current transformer and captured with a digital oscilloscope at a sampling rate of 100 MS/s. Each captured waveform was denoised, segmented for every AC cycle, and standardized before feature extraction. The inverse of twice the zero crossing interval of each PD signal was defined as the oscillation frequency. The cumulative frequency distribution of these oscillation frequencies within each AC cycle was then used as the input feature for the machine learning model.

For model development, a support vector machine was employed. The model was trained and validated using five fold cross validation to prevent overfitting. Classification performance was evaluated using the Fβ score as a metric to balance precision and recall. The developed model achieved an accuracy exceeding 90% in distinguishing between semiconductive layer and main-insulation degradation. These results confirm that oscillation frequency based features are effective indicators of insulation degradation modes.

These findings demonstrate the potential of a PD phase independent machine learning model that utilizes the cumulative frequency distribution of PD oscillation frequencies for early detection of insulation degradation in high voltage motors. Future work will focus on applying this approach to on site measurements and developing a real time diagnostic system for industrial applications.

In form-wound rotating machines, the coil conductor is enclosed within an insulation system comprising a main insulation and corona-preventing layers. The corona-preventing layer is implemented using corona armor tape (CAT), a semiconductive material composed of a binder resin and carbon black applied onto an insulating substrate. To enhance dielectric performance, the insulation system undergoes vacuum pressure impregnation with insulating varnish. However, degradation of CAT has been reported due to ozone generated by partial discharges (PDs) occurring in the gap between the CAT and the main insulation. This degradation compromises the electric field mitigation function and may ultimately lead to insulation failure. Therefore, accurate estimation of CAT’s lifespan under actual operating conditions is essential for ensuring long-term reliability.This study investigates the degradation behavior of varnish-impregnated CAT under PD-induced stress using a controlled evaluation model. An artificial air gap was introduced between the CAT and the main insulation to induce PD, and the discharge energy was quantified using the V–Q Lissajous method. The surface resistance of CAT was monitored over time, and its degradation lifespan was defined as the time required for the resistance to reach a predefined failure threshold. Temperature-accelerated aging tests were conducted, and an Arrhenius plot was constructed to evaluate the activation energy and predict the CAT’s lifespan at arbitrary temperatures. The results revealed a clear correlation between discharge power and degradation rate. Higher discharge power led to increased ozone generation, which accelerated the oxidative decomposition of the binder resin and the detachment of carbon black. Arrhenius analysis identified two distinct degradation regimes: ozone-induced degradation dominated at lower temperatures (343–363 K), while active oxygen atoms generated by ozone autolysis were predominant at higher temperatures (363–423 K). The activation energy was lower in the high-temperature regime, indicating a faster degradation reaction. The degradation behavior of varnish-impregnated CAT closely matched that of standalone CAT, confirming the applicability of the lifetime prediction method based on PD discharge energy and Arrhenius analysis. Furthermore, the proposed method is expected to be extendable to CATs with different material compositions, enhancing its utility in insulation design and predictive maintenance. In conclusion, the degradation lifespan of varnish-impregnated CAT can be accurately predicted using discharge power as a parameter in the Arrhenius model. This approach enables realistic simulation of operational conditions and provides a reliable methodology for evaluating the durability of CAT in form-wound rotating machines. The proposed technique contributes to predictive maintenance strategies and improves the reliability of insulation systems.

In this article, a numerical prediction model is established to analyze the temperature distribution of windings in dry-type transformers. The study focuses on transformers with foil-winding for the low-voltage side and disc-winding for the high-voltage side, which are commonly used in modern distribution systems due to their mechanical strength, thermal performance, and compact design. Accurate prediction of winding temperature is critical for transformer design, as excessive heat can degrade insulation, reduce lifespan, and increase the risk of failure. Traditionally, two main methodologies are employed to estimate winding temperature: Computational Fluid Dynamics (CFD) simulation and zero-dimensional (0D) calculation. CFD simulations are known for their high accuracy in capturing detailed thermal behavior, but they are computationally intensive and time-consuming. On the other hand, 0D calculations offer faster results but often lack precision due to their simplified assumptions and limited spatial resolution. To bridge the gap between accuracy and efficiency, this study introduces a metamodel developed using optiSLang software. The metamodel is trained on a comprehensive database of approximately 200 CFD simulation samples. It mimics the structure of 0D calculations while leveraging the accuracy of CFD-derived data, effectively combining the strengths of both approaches. The metamodel functions as a surrogate model, enabling rapid thermal predictions without the need for full-scale CFD simulations. After extensive training and optimization, the metamodel demonstrates exceptional performance. It achieves an average prediction error of only 5×10−11 when compared to full CFD simulations. Moreover, it delivers results within 3 seconds, making it highly suitable for iterative design processes, real-time applications, and digital twin implementations. This level of performance allows engineers to conduct rapid thermal assessments during the early stages of design, significantly reducing development time and computational costs. The implementation of this predictive tool holds significant value for transformer design and optimization. By enabling rapid and accurate thermal performance assessments, it enhances the efficiency of electrical design workflows. Additionally, it supports cost reduction by allowing for tighter design margins without compromising reliability. The model can be applied to various transformer configurations, guiding engineers toward more efficient and robust solutions tailored to specific operational requirements. Furthermore, the metamodel contributes to sustainability goals by optimizing material usage. With more accurate thermal predictions, designers can avoid over-dimensioning components, thereby reducing the consumption of copper, aluminum, and insulation materials. This not only lowers production costs but also minimizes the environmental impact of transformer manufacturing. In conclusion, the proposed metamodel represents a practical and innovative approach to transformer thermal analysis. It offers a promising pathway for improving the design, reliability, and cost-effectiveness of dry-type distribution transformers in the power industry. By integrating advanced simulation data with efficient predictive modeling, this approach supports the development of next-generation transformers that meet the growing demands of energy efficiency, reliability, and sustainability.

Based on molecular dynamics simulations and density functional theory, this paper establishes a mixed insulating oil model composed of soybean-based oil and palm oil with varying proportions. The thermal conductivity and dielectric properties of the mixed insulating oil are studied. The results indicate that when the mixing ratio of soybean oil to palm oil is 20%:80%, its comprehensive performance is relatively optimal, both flowability and dielectric properties being excellent. The diffusion coefficient is approximately 0.02155 (MSD), and the mean square radius of gyration is about 5.6 Å, meeting the requirements of heat dissipation and insulation of liquid insulating materials. This work reveals the thermal conductivity and dielectric property enhancement mechanism of mixed natural ester insulating oil from a microscopic perspective, providing an important theoretical reference for the development of environmentally-friendly liquid insulating materials.

Mica tape employed for stator ground insulation in turbo and hydro generators is composed of mica paper, glass cloth backing materials, and bonds such as epoxy resin. The stator winding built of form-wound coils generates heat due to power losses in the conductor, which is dissipated by the cooling system. Since the mica insulation layer of the stator coil exhibits low thermal conductivity, improvement of this property has recently been reported by several Japanese generator manufacturers as a means to simplify the cooling systems of turbogenerators and to enhance both the output and efficiency of hydrogenators. Nippon Rika Group has accumulated approximately two decades of experience with high-thermal-conductivity mica tapes. The present report provides a representative example of the characteristics of this mica tape and cured test specimens derived from it. Initially, the general properties of high-thermal-conductivity mica tape, classified as glass-cloth-backed epoxy prepreg mica, were comparatively evaluated against those of conventional mica tape. The evaluation demonstrated that the high-thermal-conductivity mica tape exhibited properties comparable to those of conventional products. Moreover, the tape was confirmed to possess adequate flexibility and strength, thereby facilitating standard winding operations. Subsequently, cured test specimens simulating coil insulation layers were prepared, and their thermal conductivity as well as electrical, mechanical, and thermal properties were examined. Furthermore, the operational lifetime, evaluated in accordance with the relevant industrial standards, demonstrated that the insulation possesses sufficient service durability. In conclusion, high-thermal-conductivity mica tape was confirmed to exhibit electrical, thermal, and mechanical properties equivalent to those of conventional mica tape, while achieving an approximately twofold improvement in thermal conductivity. This indicates that stator coils employing the high-thermal-conductivity mica tape can enable higher performance of generators in the same footprint and/or simplification of auxiliary equipment, thereby contributing to cost reduction and environmental impact mitigation for the generator as a whole.

Accurately attributing partial discharge (PD) activity to its physical origin remains a central challenge in high-voltage diagnostics because phase-resolved partial discharge (PRPD) patterns frequently contain superposed contributions from multiple defect mechanisms together with disturbance signals and noise. As a result, expert visual interpretation is time-consuming and can be ambiguous when clusters of pulse data overlap in phase and amplitude. This work presents an unsupervised, three-stage processing pipeline that separates mixed PD activity into source-specific groups and reconstructs distinct PRPD patterns for each group, thereby reducing pattern overlap and improving interpretability.

In order to create measurement data used for testing our method, we acquired digital measurements of PD pulses in a high-voltage laboratory at a sampling rate of 200 MHz. The measurements were conducted on several well-characterized specimens, both individually and in parallel arrangement to emulate concurrently active mechanisms while intentionally including realistic disturbance conditions. The resulting dataset consists of time-domain pulse signals with ground truth information at the level of “source type present/absent,” which is sufficient to assess unsupervised separation performance without the need for prescribing labels for individual pulses.

The method proposed in this paper comprises: (i) feature extraction, mapping each pulse from the time domain to a compact feature space; (ii) dimensionality reduction; and (iii) clustering. For feature extraction we investigated Fourier- and wavelet-based representations. Across the examined configurations, spectral features derived from the Fourier Transform provided the clearest separability for the signals considered. For dimensionality reduction we applied principal component analysis (PCA) to mitigate redundancy and concentrate discriminative variance. A sensitivity study on the number of retained components identified ranges that preserve cluster structure while avoiding relevant information loss. In the clustering stage we evaluated widely used algorithms: Gaussian mixture models, k-means, k-means++, DBSCAN, HDBSCAN, and OPTICS. HDBSCAN consistently yielded the most robust separation across irregular cluster shapes and variable point densities, and it handled noise points without requiring a priori specification of the number of clusters.

Experimental results show that the proposed pipeline isolates PD sources and disturbance/noise in superposed measurements and creates source-specific PRPD patterns that are more compact and interpretable than the original superpositional PRPD. The separated patterns align with expected phase positions for the underlying mechanisms and reveal weak sources that are obscured in composite PRPDs. We also outline practical parameter choices—frequency-domain feature settings, PCA dimensionalities, and HDBSCAN hyperparameters—that support stable performance across datasets.

Overall, this study provides a practical workflow for separating overlapping PD activity and offers experimental evidence that unsupervised clustering can recover source-level structure directly from routinely acquired pulse data. The approach reduces reliance on subjective pattern reading, supports clearer diagnosis in the presence of simultaneous mechanisms, and can be integrated with existing PD measurement systems. The final paper will present representative laboratory cases and discuss pathways toward online deployment.

This work presents a methodological guide for analyzing and interpreting physicochemical tests of dielectric oil to improve consistency between laboratory results and maintenance recommendations for oil-immersed transformers. The methodology integrates technical criteria, standards from ASTM, IEC, and IEEE, and operational practices commonly used in Colombia. It emphasizes correlating multiple test types and interpreting results within the context of the transformer’s actual operating conditions, avoiding isolated evaluations that may lead to inaccurate conclusions. The guide aims to support more reliable diagnostics, better-informed maintenance decisions, and comprehensive insulation management. A brief technical description is provided for each physicochemical test included, followed by an association between test results and recommended maintenance actions. Key technical parameters and procedural steps of the methodology are outlined, including a case study. Finally, the main conclusions of the work are discussed, highlighting the guide’s role as a practical tool for improving transformer maintenance and insulation assessment.

The reliability of high-voltage insulation systems depends not only on material performance and diagnostic techniques, but also on the competence of personnel involved in operation and maintenance. This paper presents an avatar-based training framework that combines a browser-based human-like instructor with conversational artificial intelligence to support learning related to electrical insulation systems. The system integrates a ReadyPlayerMe avatar with a large language model backend and provides interactive, scenario-driven guidance focused on insulation inspection, dielectric stress awareness, partial discharge interpretation, and safe operational decision-making. Scalability across different insulation technologies and asset types (e.g., transformers, cables, switchgear, and outdoor insulation) is achieved through modular, standards-based training scenarios that can be adapted without modifying the core system architecture. Incorrect or unsafe user responses are handled through guided corrective feedback, emphasizing underlying insulation principles and associated risks rather than reinforcing erroneous conclusions. While no existing IEEE or IEC standards explicitly address conversational AI for high-voltage insulation analysis, the proposed framework is designed to align with established insulation testing and diagnostic practices. A pilot study involving engineering students and staff evaluated usability and acceptance using the Technology Acceptance Model, indicating high ease of use, trust, and engagement. Key challenges include ensuring domain accuracy, integrating real diagnostic data, addressing LLM limitations, and establishing validation procedures suitable for safety-critical training applications.

In the design of underground transmission systems, it is necessary to consider induced voltages and currents on cable metallic sheaths (screens), as they can affect cable ampacity, insulation integrity, temperature, and operational safety, among other effects.

A method used to suppress the sheath voltage while limiting the sheath current is the cross-bonding technique. This method is used in power lines with three or more cable sections, and the screens are ideally transposed into three consecutive minor sections of equal length.

However, in practice, transpositions are often non-ideal due to space or structural constraints, resulting in unequal minor section lengths. This paper presents a comparative analysis of the behavior of induced voltage and current in power cable screens with ideal and non-ideal cross-bonding under different operating conditions. The methodology involves solving the telegrapher's equations in the space-time frequency domain. A two-dimensional Numerical Inverse Laplace Transform (TNLI-2D) is then applied to obtain voltage and current profiles along the cable length. This approach enables continuous spatial sampling, which helps to identify the location and duration of the most important voltage and current peaks.

The results obtained in this work indicate that, in steady-state operation, non-ideal transposition increases circulating currents in some of the minor sections of the screen, compared to the ideal transposition case. For a real system studied, currents in specific sections were observed to increase by approximately four to six times compared to the ideal scenario; this condition can accelerate degradation in an unbalanced manner due to thermal effects. The steady-state voltage distribution remained similar between ideal and non-ideal conditions, with non-ideal sections showing increments no greater than 7% in magnitude.

The study also examines the cable system behavior during transient conditions, such as fault conditions, and in cable system configurations: with and without Sheath Voltage Limiters (SVLs). Based on simulation results, under a fault condition, it was observed that the non-ideal transposition of screens does not produce a significant change in the current values difference across the minor sections when compared to the ideal transposition.

With this technique, a high-resolution quantification of the spatial-temporal distribution of electrical stresses along the cable screens is possible, enabling identification of where these stresses reach higher levels. By identifying regions prone to higher induced voltages and currents on the screens, the analysis can be used to correlate these conditions with potential insulation-degradation mechanisms, such as partial discharges and electrical treeing.

It is widely accepted that partial discharge testing can be used to measure the insulative health of many dielectric mediums, including rotating machines, switchgear, cables, and transformers. Partial discharges are the result of localized defects in insulation, such as voids, delamination, and age or environmentally induced degradation. PD can also be indicative of contamination and are both a sign and source of insulation deterioration. Through the analysis of phase resolved partial discharge (PRPD) patterns, the presence of faults, as well as the fault type, can be determined, allowing for condition-based maintenance strategies. The most effective use of PD testing is in its ability to catalog the trending health of the insulative materials over time, but the effectiveness of trending can be mitigated by PRPDs that have poor resolution or high signal to noise ratios. PRPDs with higher resolutions can determine more than the presence, phase angle, and general magnitude of the PD. Bipolar PRPD’s with enough resolution can providing critical information about polarity and the ability to distinguish between multiple types of PD within the same PRPD plot, such as a void and surface discharge.

This paper attempts to define the qualities of a PRPD with good resolution, as it is a balance of appropriate gain, distinguishable PRPD patterns, and enough measurement time to allow for data collection without oversaturation. For that purpose, this paper explores various measurement windows, using different PD detection equipment with various resolutions, to compare the resulting PRPDs and the analyses that can be made from them.

High-voltage circuit breakers are essential components of modern power grids. While sulfur hexafluoride (SF₆) has long been the industry standard for arc quenching and insulation, recent efforts to adopt more sustainable alternatives have drawn attention to supercritical carbon dioxide (scCO₂). Understanding contact degradation under arcing conditions is critical to determining the operational lifetime and reliability of circuit breaker technology. This study investigates whether scCO₂ alters arc energy distribution and reduces contact degradation compared to air and gaseous CO₂ environments. It will analyze the erosion behavior of copper–tungsten (Cu–W) arcing contacts exposed to high-voltage arcs in scCO₂ at pressures ranging from 8.3 to 10.3 MPa. A synthetic test circuit and capacitor bank are employed to generate controlled current interruptions up to 5 kA. Arc voltage and current waveforms are captured using a high-speed data acquisition system, while post-test surface morphology will be characterized using a VP Axia Scanning Electron Microscope (SEM) and mass-loss measurements. Additionally, mass spectrometric analysis will be conducted to assess the chemical composition of the eroded surface and arc by-products. The results are expected to demonstrate that scCO₂ conditions significantly influence arc stability and contact degradation compared to tests in air and gaseous CO₂. This is one of the first experimental studies to examine arc-induced contact wear in sCO₂ under high-current conditions. These findings will contribute to a deeper understanding of material behavior under scCO₂ arcing and support the ongoing development of sustainable, SF₆-free high-voltage switching technologies.

The reliability of electrical insulation in all-electric and more-electric aircraft is critically influenced by multiple operating stresses, including electrical, thermal, and environmental factors such as voltage, temperature, pressure, and frequency. These stresses govern the initiation of partial discharges and the long-term degradation of insulation materials, directly affecting the safety margins and service life of aircraft electrical systems. Partial discharge inception voltage (PDIV) is therefore a fundamental parameter for defining insulation design limits and developing accelerated ageing protocols. This study presents a comparative experimental and computational investigation of discharge inception behavior in enameled wire insulation systems subjected to thermal, pressure, and combined thermo-environmental stresses representative of altitude operation. Experimental measurements were conducted over a wide range of temperatures and pressures to quantify their individual and coupled influence on discharge inception and breakdown strength. The measured results are analyzed using a physics-based streamer model that predicts discharge initiation by integrating ionization coefficients along electric-field lines, and by statistical evaluation using Weibull probability analysis to assess insulation reliability. Empirical regression models are further developed to represent the dependence of discharge inception voltage on multiple stress parameters, while an electrical ageing model based on the inverse-power relationship is employed to relate insulation lifetime to field and stress conditions. The combined approach provides both physical and statistical validation of discharge inception behavior, enabling prediction of insulation performance and degradation trends under realistic multi-stress operating environments. The results can contribute to the development of comprehensive lifetime models for aerospace electrical insulation operating under thermal and environmental extremes.

Surface charge accumulation on insulation surfaces under high-voltage direct current, abbreviated as HVDC, is closely linked to local electric field distortion and may contribute to surface flashover. This issue is especially relevant in DC insulation structures such as DC gas-insulated switchgear, DC gas-insulated transmission lines post insulators, and ultra-high-voltage DC wall-bushing support insulators, whose stable surface insulation is essential for long-term operation. Compared with polymeric insulation materials including epoxy resin and polytetrafluoroethylene, ceramic materials exhibit higher thermal and mechanical stability and are commonly used in harsh environments. However, their surface charging behavior under DC electric stress has not been sufficiently examined. Most previous studies concentrated on charge transport at polymeric gas–solid interfaces, while the processes of charge injection, spatial evolution, and attenuation on ceramic surfaces remain less well characterized. To investigate these behaviors, a DC electric field surface charge accumulation model was established, and controlled charge-injection measurements were carried out on three representative ceramics: Al₂O₃, AlN, and Si₃N₄. Localized charge deposition was achieved by a needle–plate electrode configuration, and an electrostatic probe was used to measure surface charge density and its spatial distribution. Measurements were conducted in air and in sulfur hexafluoride, abbreviated as SF6, to examine the influence of gas environment. In both gases, charges with the same polarity as the applied DC voltage accumulated on ceramic surfaces, and the measured surface charge density decreased radially from the injection point, indicating that charge transport and trapping are strongly localized. Compared with air, all three ceramics showed significantly lower surface charge density in SF6, which is attributed to the strong electron attachment capability of this gas that suppresses charge retention at the surface after injection. Additional measurements were conducted on Si₃N₄ samples with different surface roughness. Increasing roughness reduced the peak charge density and produced a flatter spatial distribution near the injection region, demonstrating that surface morphology affects local charge retention and radial spreading. Although the attenuation trend remained similar across materials, the magnitude and spatial extent of charge accumulation were influenced by both gas type and surface condition. These observations clarify the distinct charge-injection behavior of Al₂O₃, AlN, and Si₃N₄ under DC voltage and describe how surface morphology and surrounding gaseous medium alter the distribution and attenuation of deposited charges.

As the demand for cryogenic pumps grows across various sectors, the use of cryogenic feedthrough has expanded accordingly. In a liquid nitrogen environment at -196 °C, phenomena such as thermal contraction, gas bubble formation, and degradation of dielectric strength are commonly encountered. Ensuring reliable operation under these conditions requires understanding of the thermal and dielectric properties of insulating materials and maintaining adequate insulation performance. Glass insulators exhibit low coefficient of thermal expansion, high manufacturing precision, and stable hermetic sealing with metals. These properties make them well-suited for cryogenic feedthrough applications. However previous research on liquid nitrogen insulation has focused on evaluating the insulation properties of polymer-based synthetic insulators or examining the thermal properties of glass insulators. Analysis of the insulation properties of glass insulators under cryogenic conditions is lacking. Therefore, research is needed to evaluate the dielectric properties of glass insulators under cryogenic conditions. This paper conducted creepage discharge experiments on glass insulating specimens in a liquid nitrogen environment to secure basic data that can be utilized in the design of cryogenic devices. Barium aluminosilicate glass insulating materials were used for the tests. Specimens with different creepage distances of 11, 9, and 7mm were fabricated and subjected to creepage breakdown tests under AC voltage. The experiments were repeated under nine different conditions at pressures of 1, 3, and 5 bars. Change in breakdown voltage according to pressure and creepage distance was measured and compared and analyzed. The test results confirm that the flashover voltage rises with pressure, and higher breakdown voltages are obtained from specimens with longer creepage distance. These findings are expected to enable the determination of appropriate creepage distances and pressure conditions for the application of glass insulators in cryogenic feedthroughs.

The long-term reliability of electrical equipment, particularly motors and transformers, depends critically on the durability of their enameled wire insulation. Enameled wire insulation experiences simultaneous multi-stress aging from mechanical, thermal, and electrical sources during operation. In this study, online partial discharge (PD) monitoring techniques are used as a primary diagnostic tool to investigate the deterioration of insulation of high-temperature enameled wires. This research utilizes enameled twisted pair samples, prepared in accordance with the IEC 60851-5 standard and employing insulating materials from Thermal Class 200°C.

According to this methodology, the twisted pair samples are subjected to simultaneous multi-stress aging inside a controlled chamber. For a certain aging duration, the specimens are exposed to a range of elevated temperatures with constant electrical stress. PD activity is continuously monitored by capturing phase resolved partial discharge (PRPD) patterns in real-time using a calibrated acquisition system.

The focus of this research activity relies on the diagnostic analysis of the acquired PD patterns. The progression of essential quantitative quantities such as discharge amplitude, pulse repetition rate and phase distribution is carefully monitored and correlated with the applied stress and aging time. The analysis reveals that elevating thermal stress produces characteristic transformations in the PD patterns. Likewise, a significant increase in PD amplitude, a surge in repetition rate, and an evident migration of discharge activity within the phase window. Such variations are related to degradation processes. A comparison of different insulating materials reveals varied responses to applied stress, identifying distinct PD patterns corresponding to a particular aging mechanism.

The diagnostic based on the amplitude-number-phase relationship in PD patterns provides a highly sensitive degradation indicators. This research identifies a correlation between measurable PD pattern features and the health state of the insulation. As a result, this methodology enables predictive maintenance strategies and improves the operation of reliability and lifespan of high-performance electrical machines under stress conditions.

This work presents the results relevant to the monitoring of electrical insulation aging through the analysis of partial discharge signals acquired under controlled thermal stress conditions.

The study utilizes twisted pair specimens prepared according to IEC 60851-5, simulating turn-to-turn insulation configurations typical of Type I windings. These specimens were subjected to accelerated thermal aging in a laboratory oven to replicate high-temperature operational environments. During the aging process, high-frequency partial discharge measurements using a high frequency current transformer were performed at defined intervals, enabling real-time tracking of insulation degradation. The acquired signals were subjected to post-processing techniques to isolate relevant partial discharge events from background interference and remove the contributions of noise.

This study uses artificial intelligence (AI) algorithms, specifically unsupervised clustering methods, to support the classification and temporal analysis of partial discharge patterns. These algorithms were applied to identify signal clusters that appear to correspond to different aging mechanisms or relevant to different locations of occurrence within the insulation system.

The clustering approach provides a structured way to interpret large volumes of partial discharge data. The temporal evolution of these clusters offers a potentially useful framework for observing changes in insulation condition over time. This method also allows for the tentative identification of possible end-of-life indicators, contributing to a more informed assessment of insulation health.

The proposed system can distinguish between noise and genuine partial discharge activity. This Artificial Intelligence driven clustering enables the tracking of insulation condition across the aging timeline.

The results demonstrate that the combination of single signal partial discharges acquisition, rigorous signal processing, and intelligent clustering algorithms can improve the efficiency of insulation ageing conditions over time. The methodology supports continuous monitoring and long-term trend analysis, making it suitable also for integration into industrial monitoring platforms. By enabling early detection of insulation defects and tracking their evolution, this approach can contribute to enhanced operational safety, to reduced maintenance costs, and to extended equipment lifespan. In conclusion, this work underscores the transformative potential of AI-enhanced diagnostics in electrical insulation monitoring and provides a basis for future developments in predictive maintenance strategies.

With the long-term use of power equipment that was massively introduced during Japan’s period of rapid economic growth, the advancement of maintenance and diagnostic technologies has become an urgent issue. Among these, insulation degradation of busbars is one of the major factors that reduces the reliability of power systems. To assess their soundness, highly accurate detection and localization of partial discharge (PD), which is an early-stage degradation phenomenon, are essential. However, since the interior of a busbar consists of complex metallic conductors, the reflection and attenuation of electromagnetic waves make it still difficult to precisely identify the discharge source location. Conventional PD detection methods employ current sensors, TEV sensors, or UHF antennas. Nevertheless, these traditional approaches sometimes fail to ensure sufficient detection sensitivity and localization accuracy.

In this study, we propose a novel non-contact diagnostic method focusing on the magnetic field components generated by discharge current pulses, using multiple miniature loop antennas (LS). Loop sensors are less affected by electric field components and exhibit high sensitivity to magnetic fields. The target of investigation is an epoxy-molded busbar that combines high insulation performance and mechanical strength. Discharge signals caused by internal defects such as voids were captured. Several loop sensors were arranged around the busbar, and the discharge source location was estimated three-dimensionally based on the analysis of signal arrival time differences and phase inversions observed at each sensor. This method extends the conventional Time-of-Flight (ToF) technique; however, the application of loop sensors to epoxy-molded busbars has been extremely limited.

The experimental target was a three-phase epoxy-molded busbar with an interphase voltage of 22 kV, measuring 35 cm in height, 90 cm in width, and 20 cm in depth. Three loop sensors, each with a diameter of 10 mm, were initially installed around the busbar at 40 cm intervals. The spacing was later reduced to 10 cm, and measurements were performed while shifting the sensor positions by 5 cm increments. The output signals from each sensor were simultaneously recorded using an oscilloscope (Tektronix MSO58, 1 GHz bandwidth, 6.25 GS/s, 8 channels). As a result, it was confirmed that the proposed method could locate the discharge source with an accuracy of approximately 5 cm. The primary cause of error is considered to be the non-uniform propagation path of electromagnetic waves inside the busbar.

Future work will include direct verification of defect locations through X-ray and CT imaging, as well as systematic measurement of PD characteristics under long-term voltage application up to insulation breakdown. These efforts aim to elucidate the degradation mechanism and establish lifetime prediction for epoxy-molded busbars.

With HV cables of considerable length, the output node of HV power supplies (HVPS) is subjected to cable reflections in shape of MHz fast-decaying bipolar aperiodic pulses (BAP). Previous work depicted the conditions of such voltages generation and empirically collected and/or surmised evidence of their deleterious effect. If breakdown (BD) under dc and/or unipolar pulses was studied experimentally, and corresponding BD voltage (BDV) values have been relied upon for decades, literature appears to be scarce, if available, on BD studies or BDVs at BAP; BD mechanism appears to be unclear. It is known that BDV increases at short pulses compared to dc and/or line frequency but is lower at high frequency (HF). This is extensively discussed elsewhere for gases, liquids and solids. A unifying trait of HF BD in solids is a) enhanced heating by dielectric losses; b) accumulation of space charges (SC) “frozen” in the gap because of fast polarity changes, hence field distortion and the resulting BDV drop. With BAP bursts having just a few periods, heating is hardly a dominant effect; SC accumulation seems a likelier culprit of lower BDV. This paper studies BD at BAP in a dedicated test rig comprising a 200-kV pulser charging a 10-15-m HV cable connected to an adjustable spark gap (SG) at the distal end. Device Under Test (DUT) was connected at the cable proximal end, between the pulser and the cable. With HV cable disconnected, or SG nor breaking down, DUT was subjected to unipolar pulses of positive polarity. Upon SG BD, cable reflections are sent to DUT. HV cable physical arrangement and shield termination had significant impact on the frequency and shape of the resulting reflections. First, DUTs comprised a 40kV silicone wire laid on a metal plane. These DUTs are an air-solid insulation system, which BD is affected by both gas and solid phase. Then needle-to-plane ½” gaps potted in a silicone compound were tested. The next DUTs were brass rods-to-plane, potted in the ¼” gap in the same compound, filleted by 0.15-mm radius. A step-stress method was used with voltage varied typically from 40kV to 135kV. In brief, the test results showed the following.

1. Tests with the 40-kV silicone wire showed that BDV at BAP was ≈40% lower than that on unipolar pulses, and the number of shots withstood was at least an order of magnitude less. Although these tests involve also air, they indicate to a lower BDV of the cable solid insulation at BAP compared to unipolar pulses.

2. All-solid insulation DUTs also had 30-70% BDV reduction at BAP with large spread between the rod samples, smaller with needles, but all over, the number of shots withstood was 3-5 orders of magnitude less than that on unipolar pulses.

We conclude that insulation subjected to BAP needs to be crafted more carefully than working at dc and/or unipolar pulsed voltages. Understanding BD mechanism at short BAP bursts beyond the generic supposition of dominance of SC similar to HF BD would be beneficial for insulation design.

Voltage supplies providing high dc voltages (up to one hundred kV) at relatively low

currents (up to few Amperes) are widely used for testing materials and components as

well as for initiating and supporting core processes in various technologies, e.g., for

generating gas discharge plasmas for ozone production, surface treatment and other

plasma-chemical applications. Such voltage sources consist of a step-up transformer and

a rectifier unit. Existing devices typically utilize transformers with magnetic cores made of

iron or several nano-composite rings with windings directly wound on them, which are

bulky and expensive, requiring sophisticated construction and complex circuitry. An

alternative approach is proposed where a single core design is implemented by packing

multiple ferrite U-cores together to create a window for a stack of multiple secondary

windings implemented as traces etched on PCBs with integrated rectifying elements. The

PCBs rectified outputs are connected in series to achieve the required high dc voltage.

The modular solution of the secondary winding reduces complexity, size, material

requirements and costs also facilitating the construction and assembly procedures.

However, it introduces additional stress on the insulation system due to the dc offset

voltage (which also increases with the number of PCB windings connected in the stack)

and superimposed high frequency voltages on each PCB induced by switching

operations. In the present study, the electric field patterns and its strength levels that

stresses the insulation system were identified and quantified by means of computer

simulations. The results of the simulations are presented and discussed in the paper.

Based on the performed analysis, a prototype of the transformer (90 kV / 0,4 A / 36 kW)

was built and successfully tested. Specific features of the simulations corresponding to

different test conditions are outlined in this work.

Reliable clustering of partial discharge (PD) wave-

forms is critical for source separation and insulation assessment,

especially without labeled data. However, waveform variability

due to propagation effects and noise limits conventional feature-

based methods. This paper proposes an unsupervised graph-

based framework that combines convolutional autoencoder fea-

ture extraction with GraphSAGE embedding and Leiden commu-

nity detection. PD pulses are represented as nodes in a similarity

graph, enabling clustering based on both waveform morphology

and neighborhood structure. A stability-based resolution selection

ensures robust and scalable identification of PD sources without

prior knowledge of their number.

Insulating oil play a crucial role in transformer operation, it requires a special attention. Control tests, and sampling methods on new or in-service oil, are carried out in accordance with current standards. These analyses are carried out in the laboratory, and their interpretation requires the involvement of qualified specialists. The decision resulting from these tests is of capital importance: an incorrect assessment can lead to significant material damage, expose personnel to the risk of serious accidents, cause considerable economic losses, and even result in the shutdown of the network and the removal of the transformer. The main of this study focuses on the classification and diagnosis of the transformer used oil. The state of this oil can be assessed from the results of six standardized tests: color index, kinematic viscosity, acidity index, dissipation factor, dielectric strength and water content. Based on the obtained values, the oil is classified in four categories, thus determine the appropriate treatment: conservation, filtration, regeneration, or disposal. Although contaminated oil can be treated. The decision to reform an oil is therefore mainly based on economic considerations. Moreover, the use of artificial intelligence algorithms, such as neural networks, offers the possibility of automating and optimizing the transformer oil classification process. These approaches provide faster and more reliable analysis, while reducing potential human error. Furthermore, this algorithm can be continuously improved through learning from labeled databases, thus enhancing their accuracy and efficiency over time.

The backbone of any high voltage insulation system is mica paper. Due to the inherent weakness of the paper additional substrates are bound to the paper with epoxy or similar resins to form a laminate. Substrates including glass, polyester film, polyimide film, or polyester fleece are commonly used in several different combinations. The taping method needs to match the materials and the end use application. Improving productivity and automating certain aspects requires analyzing the combination between the process, materials, and equipment.

Endless combinations of mica paper style, glass weights, or binder contents can be explored. Minor variations in these materials are heavily scrutinized due to the sensitivity of the end use application. This study intends to explore the additional variables in the process including taping tensions, lapping, and manufacturing equipment to apply the mica tape rolls. Taping technology varies across the industry. The scope of this paper is to share findings from voltage endurance testing utilizing identical raw material lots that were taped on different machines.

The rapid global shift toward electrified transportation has accelerated research in almost every subsystem of electric vehicles (EVs). The electrical insulation system often perceived as a passive element has emerged as a critical enabler of safety, reliability, and performance. With EV architectures transitioning from conventional 12V and 48V systems to high-voltage platforms of 400V, 800V, and beyond, the insulation materials used in traction motors, inverters, high-voltage cables, and battery systems must meet increasingly stringent electrical, thermal, and mechanical demands. The integrity of these insulation systems ultimately dictates not only the operational safety of the vehicle but also its lifetime efficiency and cost of ownership. Insulation materials also serve dual roles, former for electrical isolation between sensor power lines, communication buses (CAN-FD, Ethernet, LVDS), and high-voltage circuits while the latter one for electromagnetic shielding and dielectric management to minimize coupling between high-frequency components.

Conventional insulation materials, such as polyimides, epoxies, and silicone rubbers, are being re-engineered to meet the new performance envelope. Modern developments emphasize polymer composites and hybrid materials reinforced with nano- or micro-scale ceramic fillers such as aluminum oxide (Al₂O₃), silicon dioxide (SiO₂), and boron nitride (BN). These fillers improve dielectric strength, thermal conductivity, and tracking resistance. Furthermore, epoxy and polyurethane systems with controlled cross-linking densities are being optimized to enhance thermal endurance and mechanical flexibility vital for components exposed to vibration, high switching transients, and temperature cycling.

The introduction of wide-bandgap (WBG) semiconductor devices such as silicon carbide (SiC) and gallium nitride (GaN) has revolutionized EV power electronics, enabling higher switching frequencies and increased efficiency. However, these advances also generate extreme electrical stresses, characterized by high voltage gradients (dv/dt) and partial discharge phenomena that accelerate insulation degradation. Traditional materials and geometries are insufficient to withstand these transients. Consequently, the industry is exploring field-grading dielectrics, encapsulation resins with enhanced partial discharge resistance, and advanced potting techniques to stabilize electric fields and improve lifetime performance.

In high-voltage battery packs, insulation serves multiple roles: electrical isolation, mechanical separation, and thermal containment. Polymers such as polyether ether ketone (PEEK), polypropylene (PP), and polyethylene terephthalate (PET) are being tailored for chemical compatibility, flame resistance, and vibration tolerance. Meanwhile, for high-voltage cables, cross-linked polyethylene (XLPE) and thermoplastic elastomers (TPEs) are replacing heavier traditional materials, offering reduced weight and improved recyclability. The growing adoption of wireless battery management systems (wBMS) further reduces harness complexity, but demands higher insulation performance per unit distance to maintain safety margins.

Electrical insulation in electric vehicles is transitioning from a passive safeguard to an active design parameter influencing system efficiency, miniaturization, and sustainability. The materials are critical to the success of new vehicles, providing thermal protection, electrical insulation, and mechanical durability. The convergence of materials science, electrical engineering, and environmental innovation will define the next decade of insulation development. Advances in polymer chemistry, nanocomposite engineering, and smart diagnostic technologies will not only enhance reliability but also expand the functional role of insulation as a cornerstone of next-generation electric mobility.

Abstract

This poster compares two energy pathways driving dielectric decomposition in power transformers: thermal stress (arcing) and electric-field stress (ionization). Thermal stress breaks strong molecular bonds and produces characteristic gas patterns consistent with heat-driven bond scission. Electric-field stress converts field energy into charged-particle kinetics; repeated low-energy PD events preferentially generate H₂ and CH₄ without significant bulk heating, initiating micro-scale carbonization/tracking.

A deep understanding of these two energy pathways helps select the right diagnostic techniques and set realistic expectations for testing and monitoring. However, because transformers experience combined electrical, chemical, and thermal stresses, continuous monitoring can reveal dynamics and trends that offline tests may miss.

The poster’s aim is evaluation, not prescription: to clarify how thermal vs. electrical stress differently decompose mineral oil and cellulose and to set realistic expectations for what each diagnostic (DGA, online PD, PF/DF, offline PD) should reveal, case by case. This comparative view supports more coherent test planning and interpretation across monitoring and maintenance workflows.

This study investigates the improvement of impulse partial discharge inception voltage (PDIV) in laminated polyimide (PI) insulation films incorporating a boehmite alumina (BA) nanocomposite layer for inverter-fed motor applications. With the increasing use of high-speed switching in inverter-driven systems, steep surge voltages are applied to motor windings, leading to partial discharge (PD) and subsequent insulation degradation. To enhance insulation reliability, nanocomposite technology using inorganic fillers such as BA has attracted attention. In this work, laminated films consisting of a pure PI layer and a BA-filled PI layer were employed to clarify the effect of layer configuration and thickness ratio on PD characteristics. The specimens included single-layer PI and laminated BA/PI and PI/BA films with BA thickness ratios of 13% and 28%. Impulse PDIV was measured using an IEC(b) electrode system under repetitive impulse voltage conditions. Experimental results revealed a clear dependence of PDIV on both the configuration and thickness of the BA layer. Among the tested samples, the 13% BA laminated film exhibited the highest PDIV when the BA layer was placed on the high-voltage electrode side (BA/PI configuration), achieving approximately a 12% increase compared with the reverse configuration and neat PI. In contrast, this improvement was not observed in the 28% BA samples, indicating the presence of an optimal BA thickness ratio. The observed PDIV enhancement could not be explained solely by conventional factors such as dielectric thickness and relative permittivity, as predicted by Dakin’s equation. Therefore, the results were interpreted based on the Volume–Time theory, which considers the probability of initial electron generation leading to PD. Two primary mechanisms were proposed: (1) suppression of electron emission due to the higher first ionization energy of BA compared with PI, and (2) electric field relaxation within the laminated structure caused by the higher permittivity of the BA layer. These effects reduce the availability of initial electrons and require higher applied voltage to initiate discharge. The findings demonstrate that the PDIV performance of laminated nanocomposite insulation is strongly influenced by the BA layer configuration and thickness. In particular, an optimal BA layer ratio exists that maximizes PD resistance. This study provides important insights for the design of advanced insulation systems aimed at improving the reliability and durability of inverter-fed motor insulation.

The present paper reports simulation-based investigations of electric field distributions in different geometrical configurations of the corona electrode. Previously, various corona electrode arrangements have been used in the literature for corona aging of insulating samples. However, to date, no conclusive comparative study has been performed to choose the best corona electrode design that would result in a uniform electric field distribution. So, in the present study, simulation investigations have been carried out using Finite Element Methods (FEM) to explore the impact of corona discharges that affect the electric field distribution at various gap spacings between the HV and GND electrodes. An understanding of the effect of such conditions that leads to the intensification of the electric field is needed to better design the corona electrode for experimental purposes. The effect of the electric field distribution in the vicinity of the corona-treated insulating surface is developed. This study provides information on the selection of the best-suited electrode design to achieve a homogeneous electric field.