Transformer Insulation Diagnosis Research Articles

Influence of Data Balancing on Transformer DGA Fault Classification With Machine Learning Algorithms

The application of artificial intelligence algorithms for transformer incipient fault classification using dissolved gas analysis (DGA) is an interesting engineering approach. However, there are various factors that affect the performance of artificial intelligence algorithms. This article presents the influence of the data balancing approach on transformer DGA fault classification with the machine learning (ML) approach. In this work, a total of 4580 DGA samples from in-service transformers are considered for training various ML models. The main challenge for the DGA problem lies in the availability of the normal degradation transformer data and its uniformity corresponding to different faults is almost impossible. This is because DGA is not an exact science, but an empirical approach subjects to variability. Thus, it is a usual practice to apply data sampling techniques that largely influence the efficiency of the algorithms. The present work reports the impact of the data balancing schemes on the performance of the fault classification and demonstrates that a careful choice of the data sampling method and ML algorithm is essential for DGA problems. To demonstrate the global scale ability of the propose model, the model is tested on the IEC TC ten data (field inspection data), while these data are not exposed to the machine during the learning stage. The ability of the ADASYN method in significant enhancement of the global scale capability of AI-based transformer DGA fault classification is reported. This approach will be helpful for condition monitoring engineers in transformer insulation diagnosis in implementing the monitoring modules for large transformer fleets and understanding the insulation oil behavior over years.

Fault diagnosis technologies for power transformers during the on‐site inductive oscillating switching impulse voltage withstand test

Although oscillating switching impulse (OSI) voltage is recommended for the transformer field test, the lack of corresponding fault diagnosis technologies limits its application in the field. In this study, the inductive OSI test technology and diagnostic methods are investigated. The front time and the oscillation frequency of the OSI are altered by changing the additional inductance, whereas the tail time is affected by both the front and tail resistance. The OSI voltage withstand test is carried out on a 110 kV three-phase double-winding transformer on site. Furthermore, transfer function (TF) analysis is used for the diagnosis of transformer insulation with high sensitivity during the OSI test. The fault winding is distinguished by comparing the change in the voltage TF and current TF at different voltage levels. The partial discharge (PD) detection method has the highest sensitivity, which realizes the early diagnosis of insulation faults. The insulation status of the windings is ascertained based on the number and amplitude of PDs. The test results prove that the inductive OSI voltage withstand test is easy to perform on site and can successfully diagnose the insulation faults of the transformer in more detail compared with other routine insulation tests.

Estimation of Conductivity at Reduced Time for Sensing Moisture Content of Oil-Paper Insulation

Conductivity is an important parameter which has definite correlation with the amount of moisture present within the transformer insulation. However, in existing methods conductivity is usually estimated from time domain (PDC) measurement, which is inadvertently noise prone and time consuming. Considering this issue, in this article presents a novel technique for conductivity estimation for moisture sensing of oil-paper insulation using frequency domain spectroscopy measurement (FDS). To this end, FDS measurement of oil-paper insulation was performed from 50mHz to 1kHz to obtain the imaginary complex permittivity $\varepsilon ''$ ( $\omega$ ), which is a function of conductivity ( $\sigma$ ) as well as imaginary complex susceptibility $\chi ''$ ( $\omega$ ). From the knowledge of frequency variation of both $\sigma $ and $\chi ''$ ( $\omega$ ), we developed an electrical equivalent model of the insulation. The equivalent model parameters were optimized until $\varepsilon ''$ ( $\omega$ ) computed from the proposed model and that obtained from original FDS measurement were almost identical. The conductivity values obtained from the equivalent model showed good agreement when compared with PDC measurement. To validate the practicability of the proposed method, experiments were conducted on various test samples and also on a few real life transformers. Most importantly, the measurement time was found to be reduced by 78.5%, enabling diagnosis of transformer insulation at a significantly reduced time.

Estimation of de‐trapped charge for diagnosis of transformer insulation using short‐duration polarisation current employing detrended fluctuation analysis

Researchers have shown that the value of charge carriers, de-trapped from the oil–paper interface of power transformer insulation, is useful in carrying out the diagnosis. However, the evaluation of the de-trapped charge requires the analysis of polarisation–depolarisation currents. Being an off-line time-consuming process, the measurement and analysis of polarisation and depolarisation current (PDC) data are not practically advantageous. The study presents a detrended fluctuation analysis-based technique to estimate the magnitude of normalised de-trapped charge using the polarisation current measured for a short duration. Using the proposed technique, the requirement of measuring the complete PDC data, for diagnosis purposes, can be eliminated. Further, the technique also eliminates the requirement of depolarisation current which in turn facilitates a reduction in equipment shutdown time. The applicability of the proposed technique is tested on the data obtained from several real-life power transformers.

Condition Assessment of Power Transformer Insulation Using Short-Duration Time-Domain Dielectric Spectroscopy Measurement Data

Utilities prefer noninvasive methods for assessing the condition of power transformer insulation. Analysis of polarization–depolarization current (PDC) is one such popular method. One such analysis involves the estimation of trapped charge released from the interfacial region of oil–paper insulation. The literature shows that such charges can be reliably used for the diagnosis of transformer insulation. However, such analysis requires a complete profile of PDC. PDC measurement (an offline technique) takes a large amount of time (several hours) to complete. The magnitude of PDC data for a larger value of time is also sensitive to changes in environmental conditions and field noise as its magnitude is low. Hence, a reliable estimation of detrapped charge may require numerous PDC measurements. This situation is not convenient for utilities as it prolongs shut down time. In this article, a method has been proposed which is capable of estimating detrapping charge using PDC data measured for a short span of time. The proposed method is tested on data collected from several real-life in-service transformers.

Corrigendum: Effect of charge accumulated at oil–paper interface on parameters considered for power transformer insulation diagnosis

Corrigendum: Effect of charge accumulated at oil–paper interface on parameters considered for power transformer insulation diagnosis

Effect of charge accumulated at oil–paper interface on parameters considered for power transformer insulation diagnosis

Polarisation and depolarisation current (PDC) measurement and analysis is one of the popular tools for effective diagnosis of power transformer insulation. Normally, it is assumed that polarisation current is the combination of the current due to dipole movement and conduction current. Similarly, the depolarisation current is only due to the relaxation of dipoles. However, it is found that after eliminating the effect of dc conduction from polarisation current the resulting current is not similar to that of measured depolarisation current. This shows some non-linearity is present in the system. This non-linearity occurs due to movement of trapped charge that resides in the interfacial region of oil–paper insulation. This study shows the effect of de-trapping charge on various performance parameters that are used for insulation diagnosis like paper moisture and dielectric dissipation factor (tanδ).

Effect of measurement temperature on power transformer insulation diagnosis using frequency‐domain spectroscopy

Frequency-domain spectroscopy (FDS) is a widely accepted method for the estimation of moisture content in oil-paper insulation of power transformer. Researchers have shown that measurement temperature is an important factor that not only shifts the tan δ curve both horizontally and vertically with respect to frequency axis but also affects the value of paper-moisture content of transformer insulation. This implies FDS data measured at different temperatures yields different results even if the insulation under consideration remains more or less unaffected. Availability and construction of master curve for in-service real-life units might not always be available. This study proposes a method that is capable of predicting the profile of tan δ curve at different measurement temperatures. Existing expressions available for predicting paper-moisture does not consider the effect of change in measurement temperature. This study also proposes modification of existing expression to predict the moisture content of insulation paper at any measurement temperature. In addition, this study outlines a method to evaluate the value of activation energy ( E a ) from the insulation response. The discussed technique is first tested successfully on data recorded from a laboratory sample. Thereafter, the method is applied on data collected from a real-life power transformer.

Improvement of power transformer insulation diagnosis using oil characteristics data preprocessed by SMOTEBoost technique

This paper proposes a novel method for power transformer insulation assessment using oil characteristics. A hybrid algorithm, named as SMOTEBoost is implemented in the paper to improve the diagnosis accuracy and consistency. The SMOTEBoost can significantly enhance the generalization capability of artificial intelligence (AI) algorithms for transformer insulation diagnosis. This will provide important benefits for applying AI techniques in utility companies, i.e., an AI algorithm with its model built upon on a local dataset can be utilized globally to make transformer insulation diagnosis. The SMOTEBoost adopts Synthetic Minority Over-sampling Technique (SMOTE) to handle the class imbalance problem, in which data points belonging to different fault types or insulation conditions are unevenly distributed in the training dataset. By using this boosting approach for reweighting and grouping data points in the training dataset, the SMOTEBoost facilitates AI algorithms consistently attaining desirable diagnosis accuracy. To verify the advantages of SMOTEBoost algorithm, it is integrated with a number of representative AI algorithms including support vector machine (SVM), C4.5 decision tree, radial basis function (RBF) network and k-nearest neighbor (KNN) to make transformer insulation diagnosis using various oil characteristic datasets collected from different utility companies. A statistical performance comparison amongst these algorithms is presented in the paper.

Pattern recognition techniques for power transformer insulation diagnosis-a comparative study part 2: implementation, case study, and statistical analysis

Transformer oil tests such as breakdown voltage, resistivity, dielectric dissipation factor, water content, 2-furfuraldehyde, acidity, and different dissolved gasses have been adopted in utility companies for evaluating the conditions of transformer insulation. Over the past 20years, various pattern recognition techniques have been applied for power transformer insulation diagnosis using oil tests results (oil characteristics). This paper investigates a variety of state-of-the-art pattern recognition algorithms for transformer insulation diagnosis. To verify the applicability and generalization capability of different pattern recognition algorithms, this paper implements 15 representative algorithms and conducts extensive case studies on eight oil characteristics datasets collected from different utility companies. A statistical performance (in terms of classification accuracy) comparison among different pattern recognition algorithms for transformer insulation diagnosis using oil characteristics is also conducted in the paper. Copyright (c) 2014 John Wiley & Sons, Ltd.

Pattern recognition techniques for power transformer insulation diagnosis-a comparative study part 1: framework, literature, and illustration

The condition of the insulation system of a power transformer has a significant impact on its overall reliability and serviceability. Transformer oil tests including breakdown voltage, acidity, dielectric dissipation factor, 2-furfuraldehyde, water content, and dissolved gases analysis have been commonly performed in utility companies to provide information regarding the conditions of transformer insulation. Over the past two decades, various pattern recognition techniques are proposed to interpret the oil tests results and make diagnosis on transformer insulation. However, there are still considerable challenging issues to be investigated before the pattern recognition technique can become a ready-to-use tool at utility companies. This paper provides a comparative study of pattern recognition techniques for power transformer insulation diagnosis using oil tests results. A general pattern recognition application framework will be outlined in the paper. And a comprehensive literature review on various pattern recognition techniques for transformer insulation diagnosis will be provided in the paper. The important issues for improving the applicability of pattern recognition techniques for transformer insulation diagnosis will also be discussed. A case study will be presented to demonstrate the procedure of applying pattern recognition techniques to practical transformer insulation diagnosis using oil test results. Copyright (c) 2014 John Wiley & Sons, Ltd.

An expert system approach for transformer insulation diagnosis combining conventional diagnostic tests and PDC, RVM data

Search for a reliable and efficient insulation diagnostic tool has always been the interest of power utilities. Today a large number of methods are available that can be used for insulation condition monitoring. These methods include both traditional and newer techniques. However due to complex aging process of oil paper insulation under the influence of different types of stresses, insulation condition assessment is generally performed by experts after carefully evaluating different measurement data. Furthermore, measurement data are influenced by various factors (like conductive aging byproducts, furanic compounds, paper and oil-moisture) in addition to measurement error (if any). This makes prediction of insulation condition based on single type of measurement rather difficult. This paper presents an Expert System designed to perform insulation diagnosis. The Expert System considers measurement data obtained using both traditional and newer techniques in order to come to a definitive conclusion. The Expert System extracts insulation condition sensitive information from data obtained using different techniques and then uses these to devise an optimized insulation model. This optimized model is used to predict paper-moisture content and other insulation condition sensitive parameters. Since these values are predicted using optimized model, they are not dependent on a single type of measurement and hence are less likely to be affected by error of any specific measurement. The performance of the developed Expert System is first tested on a laboratory sample and then on several real life power transformers belonging to NTPC Ltd.

Experimental investigation of insulation parameters affecting power transformer condition assessment using frequency domain spectroscopy

In recent years, dielectric relaxation methods in time and frequency domains have been evolved as a tool for the assessment of transformer insulation. However, it has been observed that the results of these tests are highly influenced by several factors viz. the temperature of the insulation, the magnitude of the test voltage, the geometry of transformer insulation structure. In the present work, a detailed analysis of the influence of the above-mentioned factors on the dielectric response data is presented by performing measurements on a prorated transformer model. A novel experimental procedure is suggested to acquire a database to study the effect of insulation duct geometry on the relaxation measurements. Possible corrective actions are suggested to improve transformer insulation diagnosis.