Conference Agenda

This paper investigates fault detection in 36 kV Air-Insulated Switchgear (AIS) systems through online Partial Discharge (PD) measuring. The study utilizes portable PD detection equipment equipped with Transient Earth Voltage (TEV) and ultrasonic acoustic sensors. It details the operational experiences of TEİAŞ (Turkish Electricity Transmission Corporation) in conducting periodic PD scans using unconventional, non-invasive methods. Two significant case studies—involving a vacuum circuit breaker (VCB) failure and solid-insulated current transformer (CT) defects—are analyzed. The correlation between TEV and acoustic discharge patterns is evaluated alongside offline diagnostic tests, including AC dielectric loss (tan δ) and DC insulation resistance (IR), and further validated through visual inspections. The results demonstrate that fault identification is achievable with portable sensing technologies, suggesting that integrating TEV sensors into advanced, stationary PD measuring systems can further enhance diagnostic accuracy.

Metal-clad switchgears distribute electrical energy in Medium Voltage (MV) distribution networks and play an important role in the reliability and safety of the power supply to customers. According to statistics, 40% of metal-clad switchgears faults are attributed to insulation failures originating from cracked supporting insulator, loose electrical connections or vibrating components, flawed insulation material, cable terminations defects or surface contamination combined with moisture. These defects can excite thermal degradation or Partial Discharges (PDs) under normal working stress which are hazardous to switchgears’ insulation systems and may eventually result in catastrophic failures. Therefore, condition monitoring based on PD detection has been recently recognized as an effective, nondestructive, and noninvasive predictive maintenance technique, which provides an early indication of potential deterioration or damages in the switchgears. In this paper, a case study of online PD measurements conducted at 34.5 kV MV metal-clad switchgears installed at company’s three substations is discussed. The PD measurements were performed in three measurement sets to build a trend using portable PD monitor with ultrasonic contact sensors. The measurement analysis was performed using established criteria based on best in-class industry practices and hands-on experience gained during several tests. The measurement analysis concluded the presence of unacceptable PDs as per the industry practice acceptable limits. The insulation faults were detected and accurately localized after analyzing several PD test results from different measurement points based on the amplitude comparison method. The source of the discharges were surface irregularities and sharp edges in the Voltage Transformers’ (VTs) solid bus bar connections which caused corona discharges in the surrounding air. The solid bus bar connections in all substation switchgears were replaced with insulated flexible connecting leads (lead kit) to bring down the PDs level to acceptable level. The PD measurements were repeated, and found at an acceptable level, confirming the integrity of the insulation systems in the switchgears. There is recent trend to install permanent online condition monitoring systems for real-time continuous monitoring based on temperature, Partial Discharge (PD), and humidity sensors. The data retrieved from these sensors can be transported and integrated to a remote center for analysis and visualization and establish trends for timely decision-making. The parameter values are set in the form of alerts (warnings and alarms) to indicate any abnormal conditions in the switchgear. Thus prompt fixes can be planned accordingly to reduce equipment damage, flashovers, and personnel injuries, while enhancing the reliability and safety. This advanced online condition monitoring system will improve predictive maintenance applications in the journey of digital transformation of power systems in the future smart grids.

Power distribution switchgear constitutes the critical infrastructure for control and protection in distribution networks; its operating state directly governs system safety and reliability. Mechanical faults make up a large share of switchgear failures. Yet current monitoring techniques suffer from clear limitations: approaches based on current, vibration or acoustic signals furnish only indirect measurements, are vulnerable to electromagnetic interference, and displacement sensors demand physical contact and intricate installation. High-speed cameras and other optical systems avoid contact, but introduce new constraints—they are expensive, bulky, require intense ambient lighting and consume considerable power. Moreover, these methods often fail to detect every type of mechanical fault. Previous research has shown that Dynamic Vision Sensing (DVS) can perform non-contact, dynamic measurements of key mechanisms in high-voltage circuit breakers under healthy conditions. However, existing studies have concentrated on factory tests and normal operation, leaving a systematic investigation of switching dynamics under fault conditions unexplored. Building on this foundation, the present work pioneers the application of event-based vision technology to fault-simulation experiments in distribution switchgear. We have developed a multi-condition mechanical characteristic measurement platform for 10kV SF₆ fully-insulated distribution switches, enabling visual detection and quantitative parameter analysis for energy-storage spring fatigue fault. Regarding algorithmic design, this study departs from conventional approaches that rely on image reconstruction followed by optical flow estimation. Instead, we propose a Direct Event-to-Flow Estimation (DEFE) model, which employs an end-to-end neural network to directly regress dense optical flow fields from event data, circumventing the need for intermediate image reconstruction. Comparative results demonstrate that while maintaining measurement accuracy for opening and closing times (error below 2%), the DEFE model improves inference speed and reduces computational resource demands, contributing to its improved real-time performance and engineering adaptability. The findings confirm that event vision technology can reliably capture high-speed dynamic characteristics of critical switch components even under fault conditions, providing a new technical approach for mechanical fault mechanism analysis and online condition monitoring of distribution equipment. This work not only expands application scenarios for DVS in non-contact power equipment detection but also advances event vision methodology from reconstruction-based analysis toward end-to-end dynamic perception at the algorithmic level.

Overhead line insulators are subjected to electrical, mechanical and environmental stresses which lead to insulating defects. Live line condition monitoring of such assets is therefore useful for maintaining grid reliability. This paper presents a fibre optic probe that can quickly assess cap and pin insulators by scanning the electric field profile along the string. A 400 kV insulator string comprising 24 discs with and without artificially placed defects was scanned by attaching the probe to a moving lift. Each scan was completed in approximately 30 seconds and the measurement profile indicated distortions due to the presence of defects. When such defects are detected, a more detailed electric field profile can be subsequently measured by moving the probe to specific points at the edge of each insulating disc. The results show that, if combined with autonomous UAVs, the probe has the potential to perform fast condition assessment of several insulator strings in an area, improving efficiency of grid maintenance operations.

Supercritical carbondioxide (scCO₂) has emerged as a promising alternative to sulfur hexafluoride (SF₆) as an arc-quenching medium in high-voltage switchgear. Unlike SF6, whose high dielectric strength originates from its strong electronegativity, scCO₂ attains an even higher dielectric strength because of its much shorter mean free path. scCO₂ also exhibits outstanding arc-quenching capability due to the combination of its high dielectric strength and excellent thermal properties. It offers a substantially lower global warming potential, fewer toxic decomposition products, and is much cheaper compared to SF₆. To achieve these benefits, scCO₂ must be maintained at its supercritical state via elevated operating pressures on the order of 100 bar and be accelerated through an optimized nozzle geometry to a speed sufficient for extinguishing the arc and preventing restrike. A 72-kV class circuit breaker utilizing scCO₂ is currently under development. However, the extreme high-enthalpy environment generated during arcing events hinder local diagnostics and flow visualization. This limitation constrains model validation efforts, as the modelling approaches of interest rely on transport and thermophysical properties that remain experimentally inaccessible. Our previous work developed a finite-volume based framework to predict fluid phase change during an arcing event, but the time- and length-scales conventional to Navier-Stokes based methods were too large for characterizing the dissociated species and their respective particle dynamics. To address these gaps, both a finite-volume (FV) and particle-in-cell (PIC) approach were used to investigate the underlying physical mechanisms that drive arc-fluid interactions at supercritical pressures. A recent experiment utilizing a sub-scale testbed with scCO₂ provided input conditions for the corresponding computational models, which in turn produced flow field distributions of density and temperature for comparison and analysis. While each framework has been separately validated for non-ionized and partially ionized supercritical flows, current findings underscore the need for their coupling to accurately resolve the coupled plasma-fluid dynamics that underlay scCO₂ arc discharges. Our FV results show that turbulent jet theory overpredicts the spreading angle for cold supercritical flows, which is detrimental for accurately characterizing nozzle outlet characteristics needed for arc-quenching. PIC results show that high-pressure arcs in gaseous flows are accurately characterized using the kinetic theory approach. However, the method is insufficient for predicting breakdown for high-density liquid-like scCO₂ flows, or thermodynamic states above the fluid’s Widom line. Further work for the final paper will investigate coupling these models to elucidate how variations in nozzle and contact geometry affect arc-fluid interactions, to inform improved experimental design of our test apparatus.

High-voltage circuit breakers (HVCBs) are crucial components in power systems, ensuring stability and safety by interrupting current during abnormal conditions. However, their mechanical complexity makes them prone to faults, particularly mechanical failures that account for nearly 80% of all interruptions. Traditional vibration analysis and manual inspections are time-consuming and unreliable. With the rise of artificial intelligence (AI), machine learning (ML), and signal processing, intelligent fault diagnosis techniques have gained significant attention. This paper reviews 100 research studies published between 2005 and 2025, covering vibration-based, acoustic-based, current-based, and voiceprint-based fault diagnosis methods. Each study is analyzed in terms of methodology, key results, advantages, and limitations. The comparative discussion shows a clear transition from classical handcrafted-feature methods to advanced multimodal fusion and deep-learning-based frameworks. The paper concludes with insights into existing challenges and potential directions for real-time, adaptive, and data-efficient HVCB monitoring systems.