A Study on the Effectiveness of Partial Discharge Models for Various Electrical Machines’ Insulation Materials

When detecting the presence of partial discharge (PD) activity in the insulation system in high-voltage equipment, the excitation voltages at variable frequency have been widely used instead of power-frequency (50/60 Hz) sinusoidal voltage in order to reduce the charging power. This work reviews the relevant research on PD activity at very low frequency (VLF) method, including sinusoidal or cosine-rectangular voltage shape, and damped AC (DAC) method. Based on the research history and development status, some major PD characteristics, such as PD inception voltage (PDIV), PD amplitude, PD charge, PD phase-resolved pattern, and several hot issues, such as surface charge decay and statistical time lag, have been discussed. Moreover, the advantages, disadvantages, and applied conditions of two reviewed methods has been summarized. Finally, the prospects have been made on the main development trends of this research field in the future.

In this paper a novel approach to model partial discharges (PD) activity taking place inside a spherical void in epoxy resin system has been traced. The approach is based on a time varying conductance of the inner void, subjected to multistress conditions: voltage, temperature and pressure. A simple lumped circuit macro-model simulates the global effects of PD activity: the different parameters influencing the discharge phenomenon in the void are taking into account by using a physical approach resulting in a time varying conductance inside the circuit. The evaluated PD activity has been compared with experimental and simulated one for the accessible and inaccessible part of the system. A discussion on the observed changes in PD activity has been reported.

Among the factors influencing partial discharge (PD) activity in insulation systems, deposition of electric charges on dielectrics significantly impact the subsequent behaviour. This work presents a measurement technique that allows studying the charge released during PD events by measuring the resulting excess current. The method, being similar to the dielectric response analysis, measures the current flowing through a test object before, during and after a PD event. By comparing the dielectric (capacitive) and the total currents, the charging process as well as its decay can be explored. The applicability of the method is demonstrated when analysing PD activity in a twisted-pair test object exposed to a square voltage waveforms.

On-Line Continuous Partial Discharge (PD) measurements of medium voltage substations are an excellent way to determine the overall health of the equipment. Partial discharge measurements can provide maintenance alerts to allow scheduling of equipment outages based on actual condition data as opposed to scheduling based on time intervals. PD measurements provide information about the insulation system that is impossible to extract with other methods. Remote monitoring adds additional value to continuous monitoring by relieving customers of the burden of data analysis, readily involving remote experts. This paper discusses the effectiveness of PD measurements using multiple sensors. Multiple sensors provide additional noise reduction and the ability to localize PD activity for measurements made in higher frequency bands. The importance of continuous PD monitoring vs. periodic walk in PD measurements for switchgear is also discussed. PD activity in switchgear can be unstable and dependent on ambient conditions. Negative periodic PD tests in switchgear can result in a false sense of security. Case studies from field experience are presented to support the discussions.

Partial discharge (PD) online measurement is considered as one of the most effective means to assess stator insulation state and prevent sudden accidents of hydro-generator. However, the existence of external interferences has greatly reduced the credibility of PD online measurement technology of hydro-generators. In this paper, a set of high reliable hydro-generator PD on-site measurement system was developed, which made use of PDA technique to obtain PD pulses, and in order to eliminate the external interferences and ensure the reliability of PD pulses acquisition, the combination of directional double sensors and pulse time-delay technique are adopted, which can effectively suppress the external pulse shaped interferences (PSI) from power network. At present, the system has been applied to several hydro-generators in a huge-size hydropower plant to assess the stator insulation state, and a lot of on-site measurement results were achieved and explained. An actual failure of the stator insulation was checked out by the system and a great deal of economic loss was avoided, which greatly increases the credibility of the PD online monitoring technology.

The principal cause of damage in underground power cable installations is partial discharge (PD) activity. PD is a localized non-linear phenomenon of electrical breakdown that occurs in the insulating medium sitting between two conducting materials, which are at different potentials. The damage to the insulating material is induced by the AC voltage to which the insulator is subjected during the discharge process, and it can be directly or indirectly measured by the charge displacement across the insulation and the cavity defect. Non-invasive detection techniques that help in identifying the onset of the discharge process are required as PD is a major issue in terms of maintenance and performance of underground power installations. The main locations of failure are the accessories at points of connection such as terminals or splices. In this article, a study of electrical detection of PD and image processing of thermal pictures is presented. The study was carried out by controllably inducing specific failures in the accessories of the installation. The temporal evolution of the PD signals was supported with thermal images taken during the test in order to compare the PD activity and thermal increase due to failure. The analysis of thermographic images allows location of the failure by means of intensity-based texture segmentation algorithms. This novel technique was found to be suitable for non-invasive detection of the PD activity in underground power cable accessories.

The partial discharge measurements have been widely used in the field of insulation diagnostics. The presence of partial discharges inside the cable indicates the degradation of insulation material. This paper deals with the development of insulation diagnostic method based on the partial discharge (PD) measurements. The useful information about the partial discharge activity and insulation defects is extracted by the experimental results. Artificial cavity was introduced inside the medium voltage XLPE cable by using the traditional needleplane configuration. The statistical characteristics formulated with the help of PRPDA technique and ultra-wideband discharge characteristics by using HFCT sensor was studied and analyzed. Paper is related to study about the partial discharge phenomenon in detail. The aim is to interpret the variations in the partial discharge characteristics over the insulation ageing period in terms of physical phenomenon's taking place in PD sources. The results of this work are based on the partial discharge measurements for lifetime tests conducted over the medium voltage (MV) XLPE cable samples at high voltage laboratory.

Due to environmental concerns regarding the use of mineral oil, biodegradable oil is increasingly being used as an alternative dielectric fluid. This paper presents results of experiments performed in the laboratory on the use of biodegradable oils for transformer application. The investigations cover two important diagnostic techniques for insulation assessment: (i) an investigation of partial discharge (PD) activity and characteristics in such oils and (ii) an investigation of the standard hydrocarbon dissolved gas products produced due to transformer faults. The fundamental aim of the investigation was to provide information as to whether the existing analysis techniques of PD fault pattern recognition and dissolved gas analysis methods, developed for mineral oil, are valid when used for insulation assessment with biodegradable transformer oil. The experiments were conducted on test samples to simulate three common types of transformer faults: low-energy PDs, high-energy arcing/sparking, and overheating. For the purpose of comparison, the same tests were performed using a mineral oil (Shell Diala MX) and a biodegradable oil (Envirotemp FR3). Also, each oil type was tested at three different moisture levels and with or without the presence of solid insulating materials (pressboard). PD activity was monitored using the standard IEC60270 phase-resolved analysis method. The fault gases produced were extracted and analysed by standard gas chromatography methods. Test results indicate that the PD phase resolved patterns are, in general, similar for the two oil types and thus existing PD pattern interpretations can be used to distinguish different types of PD faults, e.g. corona versus surface discharges. However, the values of various discharge quantities (PD magnitudes, repetition rate, current, etc) are very different under the same test condition. The quantities and the trend of dissolved gases for faults in biodegradable oil are substantially different compared with mineral oil. For the PD fault, biodegradable oil was found to release only a limited number of gases. For arcing and overheating faults, the main key gas produced is different between the two oils. Also, the presence of cellulosic materials such as pressboard influences the extent of hydrocarbon gases dissolved in the oil. It was found that for biodegradable oil, fewer amounts of gases are produced as the oil gets wetter.

This paper deals with the study of the AC breakdown voltage (BDV) and partial discharges (PDs) activity of synthetic ester-based nanofluids (NF) with two kinds of carbonic nanoparticles (NPs), namely graphene (Gr) and fullerene (C₆₀); the synthetic ester (SE) being Midel 7131. The BDV measurement was achieved at various concentrations of NPs and different electrodes gap distances, while the partial discharges activity was studied only at the optimal concentration that gave the best improvements of BDV. First, a detailed improved procedure for preparing our NFs is presented. Then, the zeta potential measurements are performed on these NFs, and their stability is checked. After the preparation and characterization of samples, the BDVs and PDs parameters are measured according to IEC 60156 (modified gap distance) and IEC 60270 standards, respectively. Finally, conformity of the experimental data with Normal and Weibull distributions is examined, and the BDV at cumulative probabilities of 1% and 50% are then deduced. It is shown that BDV outcomes for all studied liquids obey both Normal and Weibull distributions, and the BDVs at cumulative probabilities of 1% and 50% are improved. Moreover, adding Gr and C₆₀ nanoparticles at different concentrations enhances the BDV values for investigated electrode gaps (0.1 to 2.0mm). The best improvement is obtained with a concentration of 0.4 g/L for 0.5mm and 0.7mm electrode gap with fullerene and Graphene nanoparticles, respectively. For a 2mm gap distance, the best improvements are of about 12.67% and 16.64% with 0.4g/L of C₆₀ and 0.3 g/L of Gr, respectively. It is also shown that the addition of C₆₀ significantly reduces the activity of partial discharges compared to pure SE, while the addition of Gr destroys the partial discharges resistance of pure SE.

It is well known that the occurrence and evolution of partial discharge phenomena within electrical insulation can lead to their destruction. Early detection of this phenomenon can avoid catastrophic failures of electrical equipments installed on transmission, distribution and generating stations. In this paper, two practical experiences using radio frequency scanning to detect partial discharge activity in electrical equipments are described. In the first case, partial discharge activity in two of three phases of a 13.8 kV air isolated bus duct was detected using radio frequency scanning after suspicions of PD activity in a 115 kV GSU power transformer and a 160 MVA generator was detected using acoustic emission and electromagnetic interference measurements tests with radio frequency current transformers. The second case describes the use of radio frequency scanning to detect partial discharge and corona activity inside a 34.5 kV metal clad switchgear located on a 500 kV substation. Although the electrical interference in the area was the main concern to get a reliable measurement, digital filtering and wide band scanning of the radio frequency scanning technique showed high efficacy to discriminate partial discharge activity inside the switchgear. Final results confirmed the abnormal condition of a circuit breaker in the switchgear.

Previous observations on electrical tree growth in epoxy resins has shown that different types of tree growth structure, electrically conducting and non-conducting, can occur dependent on the state, glassy or flexible, of the epoxy resin. In this current study, the partial discharge characteristics were characterized experimentally at a temperature of 20°C within two different epoxy resins systems having glass transition temperatures of 0°C and 50°C. The partial discharge activity (determined from apparent charge measurements) was characterized in terms of φ~q~n patterns using statistical tools. The aim was to compare the apparent charge measurements obtained from conducting and non-conducting electrical tree structures with computer simulations of the partial discharge activity in both conducting and non-conducting electrical trees. The results show that there is a significant relationship between the local extent of the partial discharge phenomena, as determined by the conductivity of the tree channels, and the apparent charge, as shown by the experimental and simulated partial discharge patterns. The implications of this work for partial discharge detection as well as for condition monitoring in real insulating systems are discussed.

International audience

Partial Discharge (PD) measurements are very promising for insulation condition monitoring in Gas Insulated Switchgears (GISs). However, for accurate PD detection and diagnosis, there is a need for better understanding of physical mechanisms behind PD activity. Accordingly, we have been investigating the PD characteristics and associated mechanisms for electrode/epoxy delamination of GIS spacers as one of the most serious defects in GIS. The gas pressure inside delamination usually changes from sub-atmospheric pressure, just after delamination initiates, to filled SF 6 gas pressure in the GIS tank, after SF 6 gas permeates into the delamination area. Therefore, in this paper, the PD activity at different gas pressures, from 0.02 MPa to 0.4 MPa, is acquired. The relative PD generation rate and its phase appearance are analyzed at different pressures and then discussed from the viewpoint of PD mechanism and PD detection. Experimental results show that, with increasing pressure, the peak of the PD generation rate shifts from around zero crossing like void-type discharge to around a peak of ac voltage, before PD pulses vanish at high pressure for the applied electric field stress. The obtained results could be useful for delamination diagnosis in GIS spacers.

Internal insulation defects of oil-immersed transformers are common because of flaws in both manufacturing and transportation process. Partial discharge (PD) activity caused by insulation defects may occur at rated voltage condition, leading to deterioration of material, which results in breakdown of insulation. Feature parameters of PD are closely related to the performance of insulation materials. Therefore, the parameter changes under long-term AC voltage due to the decrease in electrical properties of material. It is effective to acquire insulation condition of materials by analyzing changes in the parameters of PD. The present thesis studied partial discharge in insulation defect of oil-paper insulation under long-term AC voltage. The PD tests were performed on specimens of 1 mm thickness which was placed between the needle-to-plane electrodes in an oil cup. The specimen was impregnated with oil in vacuum with treatment of hot air drying in order to remove moisture. Feature parameters of PD, such as phase-resolved partial discharge analysis and equivalent time-frequency, were recorded during the test process. When the feature parameters of PD are significantly changed, some specimens were removed. The surface appearance of these specimens was studied by scanning electron microscope. The measurement provides a theoretical basis for the analysis of the changes in the feature parameters of PD, which is helpful for the estimation of remaining life of insulation material. Test results indicate that as the AC voltage applied, cellulose depolymerisation continuously happened at the surface of paper. Cellulose in some area is carbonized by PD, and particulates are created. As the PD developed, the discharge magnitude is increased at the beginning while decreased rapidly just before breakdown of the paper, and the distribution phase of discharges is almost unchanged.

Hydrogenerators are strategic assets for power utilities. Their reliability and availability can lead to significant benefits. For decades, monitoring and diagnosis of hydrogenerators have been at the core of maintenance strategies. The main cause of hydrogenerator breakdown comes from failure of its high voltage stator, which is a major component of hydrogenerators. In our previous study, it was shown that more than 85% of stator failure mechanisms indicate the presence of Partial Discharges (PD) activity. PD are minute sparks that occur within voids inside high voltage insulation or in the air around the insulating system. Each PD event does not cause immediate failure, but it will slowly erode the insulation system and will lead to breakdown in years to decades. PD signal can be detected from the lead of the hydrogenerator while it is running, thus allowing for on-line diagnosis. Hydro-Québec has been collecting more than 33.000 unlabeled PD measurement files over the last decades including two types of measurement instruments. One of the instruments used is the Partial Discharge Analyzer (PDA). PDs have several different sources and each source is characterized by a specific signature. It is therefore essential to be able to automatically recognize the nature of the DPs in order to anticipate possible degradation. Expert rules to characterize these PD signals exist but these rules cannot be used as an automatic classification tool. Indeed, for certain ambiguous cases (conflict between classes) or the presence of several signatures at the same time (multi-classes), the expert's judgement remains essential.
 A Variationel AutoEncoder (VAE) is used for dimension reduction and projection into a 2D latent space to analyze the training data and then classify it. The 2D latent space of the VAE allows the data space to be restructured and reorganized to anticipate the performance of the classifier. The problem is how to optimize this latent space and thus obtain the best distribution of the different sources of PDs to maximize the chances of good responses from the classifier? To our knowledge, there is currently no method in the literature that clearly answers this question. What is certain, however, is that the optimization of the VAE learning is directly related to the quality of the learning base, i.e. a large size and a perfectly balanced database (all the cases are present and sufficiently represented). The objective of this paper is to compare the quality of the latent space obtained from the experts' rules with a latent space obtained directly from the input signal in an « End-to-End » approach.
 The first method concerns an original unsupervised deep learning method for PD recognition. Instead of labelling the PD measurement files by the expert for a supervised learning process, we use the rules developed by the experts of Hydro-Québec to create a feature vector from recognizable PD signatures. Indeed, labelling a sufficient quantity of signatures for a supervised approach is very time-consuming and therefore cannot be implemented. This is a common problem in the industry where more and more operational data is available but cannot be labelled by experts, who are busy with other tasks. Expert knowledge is then injected into a characteristic and feature vector. In the second method, several Generative Adversarial Networks (GANs) associated with several types of PDs are thus used to generate artificial signals in order to increase the size of the learning base and especially to balance it. A new latent space is thus obtained from the learning of the VAE exclusively carried out on data generated by the GANs.
 Validation tests based on the 33,000 measurements with metrics for evaluating the performances of the various latent spaces are used.