Real Transformer Research Articles

Numerical Calculation and Direct Measurement of Local Hot Spot Temperatures in Transformer Clamping Plates

A verified multiphysics model is developed to simulate local hot spot temperatures in clamping structures. Stray field losses in clamping plates are obtained with the use of nonlinear impedance boundary condition through finite element method (FEM) while hot spot temperatures are simulated with the use of computational fluid dynamics (CFD). This multiphysics coupling is validated through direct measurements of local hot spot temperatures that were carried out with fiber optic temperature probes in real transformer. The local hot spot temperatures obtained with the simulations were in good agreement with the measured values, thus verifying overall approach and indirectly the calculation of local stray field losses.

Verification tests of a novel ferroresonance detection technique

Ferroresonance is a complex electromagnetic phenomenon that affects voltage transformers. The appearance of the ferroresonant phenomenon provokes important oscillations that are responsible for serious thermal and dielectric overstresses, which may lead to catastrophic failures in the electrical power system. This paper analyzes the behaviour of a novel ferroresonance detection technique, based on artificial neural networks, through the results obtained in several verification tests conducted in a real voltage transformer.

Simulation of Magnetizing Inrush Current of Three-Phase Three-Legged Transformers by Direct Measurement of Coils Inductance

This paper presents a method for simulating magnetizing inrush current for three-phase three-legged transformers. A method is showed for easy and fast needed transformer parameters obtaining. It is shown that magnetizing inrush current is strongly dependent on transformer core saturation, and a straightforward method is presented for taking into account core saturation. Finally, a real transformer simulation results are presented and compared with those obtained by direct measure of magnetizing inrush current, discussing method suitability.

Numerical study of the cooling capacity of several alternative liquids in zig-zag cooling system of a power transformer

Mineral oil is the usual coolant of the power Transformers. Nonetheless, the high risk of fire of this liquid due to its low ignition point and its scarce biodegradability has promoted the development of dielectrics alternative fluids with better physical-chemical properties. However, these new liquids have to be as good coolants as mineral oil in order to maintain the power transformer performance. This numerical study analyzes the coolant capacity of some of these new commercial liquids (a natural ester, a synthetic ester, and a silicone oil) by using the zig-zag cooling system of the low voltage winding of a real power transformer. Two types of thermal-fluid studies (parametric and time-dependent studies) have been made by using a 3D model of this system. The comparison of temperatures obtained from the parametric study (different inlet flow rates) allow us to say that the worst coolant is the silicone oil. Also, it can be say that a significant increase in the inlet flow rate produces mild drops of hot-spots temperatures and no modification in the initial temperature distribution. For the time dependent study (25% of overload during 1 hour), the hot-spot temperature exceed the limits, which negatively affects the aging speed of the insulating paper and therefore the transformer lifetime.

Vibration-Signal-Based Deep Noisy Filtering Model for Online Transformer Diagnosis

Machine learning methods are effective for the diagnosis of power transformer faults. However, influenced by uncertainty and noise in data, machine learning-based diagnostic methods are still in the initial phase of certain assets in power systems. To mitigate this gap, a deep noisy filtering diagnostic (DNFD) model is proposed for accurate and rapid evaluations of power transformer faults using noisy vibration signals. A balanced isolation forest method is employed to detect abnormal data from the original vibration signals. Two deep noisy filter networks suppress the level of noise, based on which contrastive learning obtains the transformer fault states. Datasets collected from a 10 kV real power transformer validate the proposed model. The results demonstrate that the proposed method acquires a higher fault diagnostic accuracy with respect to the compared algorithms, showing the superiority and efficacy of the proposed model.

Accurate Assessment of Moisture Content and Degree of Polymerization in Power Transformers via Dielectric Response Sensing.

Power transformers are essential apparatuses used to transfer electrical energy from one voltage-level circuit to another. For reliable systems, preventive maintenance of the transformers is required to ensure good services of all mechanical, electrical, and insulation parts. Oil-immersed paper is most often used for transformer insulation. To ensure such good insulation performance and for assessing insulation conditions, advanced transformer sensing, monitoring, and effective assessment techniques are required. This paper introduces an effective technique for assessing the insulation conditions in power transformers, which are crucial for ensuring reliable energy transfer. The method utilizes advanced transformer sensing and monitoring, focusing on oil-immersed paper insulation commonly used in transformers. The technique employs dielectric response sensing, obtained from frequency-domain spectroscopy tests, to estimate degrees of polymerization (DP) and percentages of moisture content (PMCs) in the oil-immersed paper insulation. These parameters are well-known indicators of insulation performance. The approach is based on the weighted k-nearest neighbor regression, using a database of dielectric loss factors at low frequency and oil conductivities. To overcome limited data availability, linear interpolation and extrapolation techniques are applied to enlarge the database. Experimental verification and comparison with a previously developed method demonstrate the proposed technique's superiority in accuracy and complexity. The maximum deviations of DP and PMC in the validation cases are 6.2% and 18.7%, respectively. In addition, to evaluate the validity of our proposed method in the case of a real power transformer, a comparative analysis of the DP and PMC values determined by the proposed method with those obtained through a previously developed and complicated approach was performed. The predicted results indicate that the DP and PMC values of the oil-immersed insulation fall within the ranges of 800 to 1000 and 1.5 to 2.0, respectively, which agree with the results determined by the complicated approach and closely align with real conditions. By offering a reliable and advanced means of assessing insulation conditions, this technique contributes to the preventive maintenance and overall efficiency of power transformers.

Experimental design and data analysis and optimization of mechanical condition diagnosis for transformer sets

Abstract The typical power transformer diagnosis approach is imprecise and unstable. A support vector machine classification algorithm is proposed, by designing an algorithm program that can improve the accuracy and speed of energy transformer diagnosis, the vibration signals of the surface twisting in different states are extracted by wavelet packet energy spectrum signal processing method, it is verified that the curve similarity between the vibration simulation model and the measured data is greater than 0.98, proving the simulation model’s validity. The calculation technique of online short circuit inductance is developed from the equivalent transformer model, and the variation error of simulation results is less than 0.05% when compared to the real transformer characteristics. The suggested state diagnostic technique successfully compensates for the drawbacks of the reactance method, which is incapable of detecting and judging the slightly loose or faulty winding. The method’s accuracy and superiority, as well as the practicability of the state diagnosis system, are demonstrated.

Novel Technique for Location Identification and Estimation of Extent of Turn-to-Turn Fault in Transformer Winding

This paper presents a novel technique for diagnosing a turn-to-turn fault at the incipient level inside the power transformer winding using frequency response analysis (FRA). Utilizing lower frequency region of FRA in case of pre-fault and post fault conditions of transformer windings, the proposed technique determines the deviation in equivalent inductances. Based on the deviation, a location factor (<i>lf</i>) is derived, which indicates location of fault, by calculating accuracy factor (<i>x</i>) corresponding to each section of the winding. Thereafter, the extent of turn-to-turn fault and number of faulted turns are determined based on the accuracy factor of the faulted disk. The performance of the proposed technique has been validated through a simulation model as well as from two different laboratory-developed winding models and 400 kVA real filed transformer. The results designate that the developed technique is proficient to estimate location of turn-to-turn fault even when number of faulted turns are lower than 5&#x0025; of the total turns. Its response remains immune against winding configurations and lead error encountered during measurement. Eventually, a comparative assessment of the proposed approach with various other approaches indicates its eminence in terms of turn-to-turn fault localization and estimation of fault extent and faulted turns.

Convolutional sparse filter with data and mechanism fusion: A few-shot fault diagnosis method for power transformer

In actual industrial scenarios, fault data is rare and fault labels are difficult to obtain, which brings many obstacles for fault diagnosis. For this situation, this research proposes a novel semi-supervised convolutional sparse filter with local mechanism similarity regularization (CSF-LMSR) to construct a more reliable few-shot diagnosis method for power transformer. First, a sparse filter with local mechanism similarity regularization term (SF-LMSR) is designed as a more interpretable unsupervised feature extractor with prior knowledge. This unsupervised process enables the model to extract satisfactory features from the whole dataset even with a lower proportion of labeled data. Second, SF-LMSR is combined with convolutional neural network (CNN) by a novel coupling mode of kernel replacement, which enhances the learning ability of CNN. This classification model still adopts supervised learning training, but the demand for labeled fault data is greatly reduced, which reduces the burden of labels. The effectiveness of the proposed method is verified using real power transformer dissolved gas analysis (DGA) datasets. It can be seen from experimental results that the proposed method does provide a new perspective for transformer fault diagnosis, and it is a successful attempt for the power industry few-shot diagnosis problem.

Dynamic Incipient Fault Forecasting for Power Transformers Using an LSTM Model

Dissolved gas analysis (DGA) is a traditional approach for power transformer fault diagnostics based on measurement of gas contamination. Hydrocarbon gases generated and dissolved in transformer oil during operation can increase in density as fault conditions predominate. Critical determination of gas concentration changes and assessment trending of dissolved gases for fault prediction and prevention of transformer damage is essential. In this paper, a dynamic fault prediction approach is proposed using a long short-term memory (LSTM) model with intelligent classification to determine the running state of a transformer for prediction and avoidance of potential transformer damage. In the paper, the LSTM model processed DGA data collected from real on-site transformer field measurements and predicts future dissolved gas concentrations in time sequence. Four artificial intelligence diagnostic models (support vector machine (SVM), K-nearest neighbors (KNN), decision tree, and artificial neural network (ANN)) were rendered and used for comparative fault prediction assessment. By comparing experimental results from the different LSTM-based models, this paper asserts that the LSTM-KNN model provides the highest and most reliable prediction accuracy for power transformers.

Reproducing Transformers’ Frequency Response from Finite Element Method (FEM) Simulation and Parameters Optimization

Frequency response analysis (FRA) is being employed worldwide as one of the main methods for the internal condition assessment of transformers due to its capability of detecting mechanical changes. Nonetheless, the objective interpretation of FRA measurements is still a challenge for the industry. This is mainly attributable to the lack of complete data from the same or similar units. A large database of FRA measurements can contribute to improving classification algorithms and lead to a more objective interpretation. Due to their destructive nature, mechanical deformations cannot be performed on real transformers to collect data from different scenarios. The use of simulation and laboratory transformer models is necessary. This research contribution is based on a new method using Finite Element Method simulation and a lumped element circuit to obtain FRA traces from a laboratory model at healthy and faulty states, along with an optimization method to improve capacitive parameters from estimated values. The results show that measured and simulated FRA traces are in good agreement. Furthermore, the faulty FRA traces were analyzed to obtain the characterization of faults based on the variation of the lumped element’s parameters. This supports the use of the proposed method in the generation of faulty frequency response traces and its further use in classifying and localizing faults in the transformer windings. The proposed approach is therefore tailored for generating a larger and unique database of FRA traces with industrial importance and academic significance.

FPDT: a multi-scale feature pyramidal object detection transformer

Object detection is a fundamental part of autonomous driving algorithms, and with the promotions of transformers in a couple of years, numerous computer vision tasks are integrating transformers into object detectors to acquire a better generalization ability. Building a pure transformer-based detector seems to be a wonderful choice; however, transformers are not omnipotent, and they come with painful drawbacks. Its fundamental operator, multi-head self-attention (MHSA), suffers from the need for computational resources due to its quadratic complexity, which demands an unreasonably high memory usage and critically low throughput. To address this issue, we use a convolution operation to simulate MHSA from transformers by referencing the philosophy and principle of MHSA and making an application migration on convolutional neural networks (CNNs). This gives a detector with power and speed simultaneously. Furthermore, a multi-scale pyramidal feature extractor gives the detector a better view over various scales. In general, our proposed object detector mainly follows the philosophy of attention mechanism, which is implemented by a multi-scale feature pyramidal CNN encoder that simulates the transformer, and a real transformer query neck to extract all of the objects once and, eventually, feed them to the output heads. After training on the COCO2017 dataset, by combining the construction philosophy of the object detector and the philosophy and characteristics of the transformer, our FPDT-Tiny gives an average precision (AP) of up to 34.1 in 150 lower epochs, which is 16.0 and 10.8 higher than CNN-based YOLOv3-Base and SSD-300, respectively. Also, the AP given by our FPDT-Small is up to 37.7 under the same epoch, which is 10.4 and 7.9 higher than the transformer-based detector YOLOS-Small and DETR-ResNet-152, respectively, also demonstrating a comparable performance.

Proof of the Concept of Detailed Dynamic Thermal-Hydraulic Network Model of Liquid Immersed Power Transformers

The paper presents a physics-based method to calculate in real time the distribution of temperature in the active part of liquid immersed power transformers (LIPT) in a transient thermal processes during grid operation. The method is based on the detailed dynamic thermal-hydraulic network model (THNM). Commonly, up to now, lumped models have been used, whereby the temperatures are calculated at a few points (top-oil and hot-spot), and the parameters are determined from basic or extended temperature-rise tests and/or field operation. Numerous simplifications are made in such models and the accuracy of calculation decreases when the transformer operates outside the range of tested values (cooling stage, loading). The dynamic THNM reaches the optimum of accuracy and simplicity, being feasible for on-line application. The paper presents fundamental equations of dynamic THNM, which are structurally different from static THNM equations. The paper offers the numerical solver for the case of a closed-loop thermosyphon. To apply the method for real transformer grid operation, there is a need to develop details as in static THNM, which has been used to calculate the distribution of the temperatures in LIPT thermal design. The paper proves the concept of dynamic THNM using the experimental results of a closed-loop thermosyphon small-scale model, previously published by authors from McGill University in 2017. The comparison of dynamic THNM with measurements on that model are presented in the paper.

A hybrid ensemble learning approach for the aging-dependent reliability prediction of power transformers

As the most critical equipment for electricity transmission, power transformers (PTs) generally have irreversible insulating degradation, which makes them susceptible to aging breakdowns. Predicting the aging-related failure probability (AFP) of PTs as precisely as possible is beneficial for reliability-centered maintenance and retirement strategies. This paper proposes a Long Short Term Memory (LSTM) network and Particle Swarm Optimization (PSO) combined prediction approach for this essential task. Firstly, a classic AFP evaluation method and computation-based degree of polymerization (DP) estimation method are combined to construct the historical AFP time series of PTs. Then, the produced AFP time series are treated as input to train the LSTM neural network, which can be used to predict the AFPs of individual PTs within a predefined horizon. To enhance the prediction performance of the LSTM network, the PSO algorithm is introduced to search the optimal LSTM network parameters adaptively. Numerical tests based on real transformer data from China’s Chongqing Electric Power Company have been conducted to show the efficacy and utility of the suggested approach.

Transformer's frequency response analysis results interpretation using a novel cross entropy based methodology

Transformer defects can be identified by the FRA (frequency response analysis) that is a promising diagnostic technique. Despite the standardization in FRA measuring technique, its results interpretation is yet a research area. Because different faults types can be identified in various frequency bounds of the FRA signatures, it is necessary to identify the possible relationships between specific failures and frequency ranges in this contribution. For this purpose, a real transformer is used to conduct the essential tests, which include both healthy and faulted circumstances (axial displacement (AD), radial deformation (RD), and short-circuits (SC)). To identify efficient characteristics from the produced frequency response traces and improve interpretation accuracy of such traces, a new hyperbolic fuzzy cross entropy (FCE) measure is demonstrated and then utilized for the aim of discrimination and classification of transformer winding defects in pre-defined frequency ranges. After normalizing FRA results of the transformer under healthy and various fault circumstances the lower bounds from such responses have been extracted and then utilized to construct the desired form of the fuzzy sets of healthy and faulted circumstances. Then, a new hyperbolic FCE measure-based discrimination and classification of winding faults methodology is offered on the basis of highest and lowest FCE measure values. The highest FCE measure value between the fuzzy sets of healthy and faulted circumstances such as AD, RD and SC is designated to confirm the occurrence of winding faults in a suitable frequency range. The suggested methodology ensures smart interpretation of FRA signature and accurate classification of winding faults as it can effectively discriminate both healthy and faulted circumstances in the desired frequency ranges. The proposed approaches' performance is tested and compared by applying the experimental data after feature extraction.

Outlier detection and data filling based on KNN and LOF for power transformer operation data classification

The missing and abnormal data in power transformer operation and monitoring greatly affect the accuracy of fault diagnosis and thus threaten the stable operation of power systems. To conduct outlier detection and improve data quality for safety warning, this paper proposes a transformer operation data preprocessing method based on KNN (K-nearest neighbor) and LOF (local outlier factor) for power transformer operation data classification. Firstly, this paper analyzes the characteristics of transformer operation data. Secondly, the local reachable density of the input data is calculated by LOF algorithm. The local outlier factor score of the data is derived according to the local reachable density, and the abnormal data is output according to the abnormal score. Then, KNN algorithm is utilized to classify the relevant data around the abnormal value and missing value of the transformer. The data are filled or corrected according to the classification results. Thirdly, the elbow method is used to determine the optimal K value and cluster operation data by K-Means algorithm. Finally, the proposed method is applied and verified with real transformer operation data in case study. The results show the method can effectively detect and correct the abnormal and missing data, conduct transformer data cleaning and preprocessing and provide accurate and effective data samples for transformer fault diagnosis.

Electromagnetic Vibration Characteristics of Inter-Turn Short Circuits in High Frequency Transformer

As a common fault of transformer winding, inter-turn short circuits cause severe consequences, such as excessive current and serious deformation of winding. Aiming to solve the problem of inter-turn short circuit at the end-winding and middle-winding of high frequency transformers (HFT), this paper considers the electromagnetic vibration characteristics of inter-turn short circuits (interleaved winding and continuous winding) at different positions, and the HFT is established by the multi-physical field coupling principle. Coupling equations for the inter-turn short circuit, as well as electromagnetic force and sound pressure level, are established to characterize the vibration noise mechanism of inter-turn short circuits. Furthermore, the HFT equivalent model is simulated in 3D finite element method (FEM) to emulate the real transformer operation and investigate the impact of interleaved winding and continuous winding under inter-turn short circuit faults. The short-circuit current and axial flux leakage, as well as the harmonic response of vibration acceleration and sound pressure level distribution, are obtained when inter-turn short circuits occur at different positions. Finally, the results show that the electromagnetic effect of the inter-turn short circuit in end-winding is worse than it is in middle-winding. Advantages in resisting impulse current make interleaved winding superior to continuous winding in terms of vibration and noise.

Research on Transformer Voiceprint Anomaly Detection Based on Data-Driven

Condition diagnosis of power transformers using acoustic signals is a nonstop, contactless method of equipment maintenance that can diagnose the transformer’s type of abnormal condition. To heighten the accuracy and efficiency of the abnormal method of diagnosing abnormalities by sound, a method for abnormal diagnosis of power transformers based on the Attention-CNN-LSTM hybrid model is proposed. This collects the sound signals emitted by the real power transformer in the normal state, overload, and the discharge condition. It preprocesses the sound signals to obtain the MFCC characteristics of the sound signals. It is then grouped into a set of sound feature vectors by the first- and second-order differences, and enters the Attention-CNN-LSTM hybrid model for training. The training results show that the Attention-CNN-LSTM hybrid model can be used for the status sound detection of power transformers, and the recognition of the three states can achieve an accuracy rate of more than 99%.

Assessment of Real Time Transformer Oil Images of Different Temperatures and Ages Using Pca-Pnn and Pso-Pnn Techniques

Assessment of Real Time Transformer Oil Images of Different Temperatures and Ages Using Pca-Pnn and Pso-Pnn Techniques

LUMDE: Light-Weight Unsupervised Monocular Depth Estimation via Knowledge Distillation

The use of the unsupervised monocular depth estimation network approach has seen rapid progress in recent years, as it avoids the use of ground truth data, and also because monocular cameras are readily available in most autonomous devices. Although some effective monocular depth estimation networks have been reported previously, such as Monodepth2 and SC-SfMLearner, most of these approaches are still computationally expensive for lightweight devices. Therefore, in this paper, we introduced a knowledge-distillation-based approach named LUMDE, to deal with the pixel-by-pixel unsupervised monocular depth estimation task. Specifically, we use a teacher network and lightweight student network to distill the depth information, and further, integrate a pose network into the student module to improve the depth performance. Moreover, referring to the idea of the Generative Adversarial Network (GAN), the outputs of the student network and teacher network are taken as fake and real samples, respectively, and Transformer is introduced as the discriminator of GAN to further improve the depth prediction results. The proposed LUMDE method achieves state-of-the-art (SOTA) results in the knowledge distillation of unsupervised depth estimation and also outperforms the results of some dense networks. The proposed LUMDE model only loses 2.6% on δ1 accuracy on the NYUD-V2 dataset compared with the teacher network but reduces the computational complexity by 95.2%.