Dissolved Gas Analysis Dataset Research Articles

Transformer fault diagnosis based on the improved QPSO and random forest

Abstract Although dissolved gas analysis (DGA) is an effective method for transformer fault diagnosis, problems with the quality and accuracy of DGA characterization datasets often arise in practical industrial applications and face difficulties in adjusting the parameters of fault diagnosis models. To address the above problems, this paper proposes a fault diagnosis model (MD-IQPSO-RF) based on Mahalanobis distance (MD) data cleaning and improved quantum particle swarm (IQPSO) optimization of random forest (RF) parameters. Specifically, the abnormal samples of the DGA dataset are first processed by MD to improve the quality and accuracy of the dataset. Then, the RF parameters were optimized using the IQPSO algorithm to adjust the model parameters in order to improve the diagnostic performance of the RF. Finally, the optimal parameters of RF are output, and the training data are used to train the RF algorithm to construct the MD-IQPSO-RF transformer fault diagnosis model. The experimental results show that the model achieves an average accuracy of 93.631% for fault diagnosis, which is 6.92% higher than the unoptimized RF model. Comparison with other similar methods also achieved good results, which further validated the effectiveness of the fault diagnosis model.

Detection of Incipient Faults in Power Transformers using Fuzzy Logic and Decision Tree Models Based on Dissolved Gas Analysis

This paper proposes an integrated approach utilizing Fuzzy Logic and Decision Tree algorithms to diagnose early-stage faults in power transformers based on Dissolved Gas Analysis (DGA) test results of transformer insulation oil. Overcoming limitations in conventional methods such as Duval Triangle, Key Gas Analysis, Rogers Ratio, IEC Ratio, and Doernenburg Ratio, our Fuzzy Logic and Decision Tree models address issues like inaccurate diagnosis, inconsistent diagnosis, lack of decisions or out-of-code results, and time-intensive manual calculations for large DGA datasets. The Decision Tree algorithm, a machine learning technique is applied to categorize faults into thermal and electrical types. Trained with over 300 DGA samples from transformers with known faults, the models exhibit robust performance during testing with different datasets. Notably, the Duval Triangle decision tree model attains the highest accuracy among the ten developed models, achieving a 98% accuracy rate when tested with 50 samples with known faults. Moreover, Decision Tree models for KGA, Doernenburg, Rogers, and IEC also demonstrate substantial prediction accuracy at 92%, 86%, 92%, and 90% respectively underscoring the efficacy of artificial intelligence methods over traditional approaches.

A novel approach for oil-based transformer fault identification in electrical secondary distribution networks

Oil-immersed transformers are mostly installed equipment in electrical power networks. Thus, ensuring the reliability of these equipment is paramount. Fault identification is one of the measures taken to ensure the reliability. Concerning fault identification mechanisms, several methods exist; one of which is the dissolved gas analysis (DGA) tool that has been widely used purposefully for oil-immersed transformers. The tool is used to capture and analyze the gases dissolved in the oil. Studies have proposed various fault identification methods based on this tool such as the Dornenburg, key gas, and International Electrotechnical Commission (IEC) standards methods using DGA data. However, the accuracy of these methods seems to be limited by the context, such as the type and number of parameters used. In this study, the novel hybrid fault identification method based on multilayer neural networks (MLANN) and expert knowledge is proposed to improve the accuracy fault identification process by considering the specific characteristic parameters extracted from the historical DGA dataset. The proposed hybrid method improves accuracy by 14.46% when compared to 12 benchmark methods. This improvement portrays that our method can be integrated into a scheduling platform for maintenance decision support in an electrical power network with accuracy. Furthermore, the promising accuracy in transformer fault identification is expected to increase safety and power reliability.

Partial Discharge Localization through k-NN and SVM

Power transformers are essential for the distribution and transmission of electricity, but they are prone to degradation due to faults early on. Partial Discharge (PD) is the most significant pointer of insulation breakdown in high-voltage apparatus. Dissolved Gas Analysis (DGA) is a commonly used technique for detecting and diagnosing PD. However, DGA data often contain missing values, which can significantly affect the accuracy of PD diagnosis. To mitigate the issues of missing values, this paper proposes using the k-Nearest Neighbors (kNN) technique to impute the missing values in the dataset. Further, it combines kNN with a Support Vector Machine (SVM) to detect the possibility of a PD source in the high-voltage apparatus. The approach was evaluated on a real-world DGA dataset and achieved high classification performance and discriminatory power for distinguishing between PD and non-PD instances. The effectiveness of the missing value imputation technique was evaluated, and the proposed approach demonstrated improved accuracy and precision compared to methods without imputation. The proposed approach offers a current solution for PD analysis in power transformers using DGA data with missing values.

Dissolved Gas Analysis of Transformer: An Approach Based on ML and MCDM

In the electric power business, power transformers are one of the most common and expensive components. The conventional diagnostic tool for understanding insulation incipient failures is the dissolved gas analysis (DGA) of transformers. Nonetheless, interpreting DGA fault gases remains a significant difficulty for engineers. Along with offline methods, a number of computational approaches have been created for DGA fault classification. Still there exist significant hurdles in applying those methods for DGA fault classification. To establish an effective fault classification system, very diversified and massive DGA data sets of in-service transformers were collected from various utilities for this study. The dataset consists of 3147 instances with four classes: No fault, Thermal fault, Low energy discharge and high energy discharge. Models were built using various machine learning approaches like Quadratic Discriminant Analysis (QDA), Gradient Boosting (GB), Extra Trees (ET), Light Gradient Boosting Machine(LGBM), Random Forest (RF), K Nearest Neighbors (KNN), Naive Bayes (NB), Decision Tree (DT), Ada Boost (AB), Logistic Regression (LR), Linear Discriminant Analysis (LDA), Ridge Classifier, and SVM - Linear Kernel. The proposed work adopts Analytic Hierarchy Process (AHP) technique for the estimation of weights of the criteria. Based on the generated weights, the performance of the various classifiers is assessed and ranked using Multi-Objective Optimization based on the Ratio Analysis (MOORA) approach. QDA is selected as the best classifier model by the proposed technique.

A Comparative Study of Machine Learning Classifiers in Oil-Immersed Power Transformer Fault Diagnosis

The most common fault diagnosis method for oil-immersed power transformers is dissolved gas analysis (DGA). Doernenburg ratios, Rogers ratios, IEC (International Electrotechnical Commission) ratios, and Duval's triangle are conventional DGA techniques for insulating oil in power transformers. In this study, Scikit-learn known as a popular open-source free machine learning tool for Python programming language has been used to develop different machine learning (ML) classifiers to effectively detect defects in oil-immersed power transformers. These classifiers are Decision Trees, Support Vector Machines, Gaussian Naive Bayes, k-Nearest Neighbours, Random Forests, and Multi-Layer Perceptron. The input vector of each classifier has been formed by Doernenburg ratios, Rogers ratios, IEC ratios, and CSUS (California State University Sacramento) method. After these classifiers are completely trained, unseen DGA data sets are then used to evaluate their performances. Based on a statistical analysis, the study can indicate the most effective type of the input vector and ML classifier for precisely detecting faults in power transformers.

Hybrid AI model for power transformer assessment using imbalanced DGA datasets

AbstractArtificial intelligence (AI) methods have been used widely in power transformer fault diagnosis with notable developments in solutions for big data problems. Training data is essential to accurately train AI models. The volume, scope and variety of data samples contribute significantly to the success and reliability of diagnostic outcomes. This paper provides a comprehensive review and comparison of different augmentation methods used to generate reliable data samples for minority and majority classes to balance the diversity and distribution of dissolved gas analysis (DGA) datasets. The augmentation method presented in this paper combines three common AI models—the Support Vector Machine (SVM), Decision Tree, and k‐Nearest Neighbour (KNN)—to assess performance for diagnostic fault determination and classification, with comparator assessment using no data augmentation. Comparative analysis of the hybrid models uses evaluation metrics including accuracy, precision, recall, specificity, F‐score, G‐mean, and the area under receiver operation characteristic (Auc). Experimental results presented in this paper confirm that the data augmentation applied to AI models can resolve difficulties in imbalanced data distribution and provide significant improvements for fault diagnosis, particularly for minority classes.

Accurate Identification of Transformer Faults From Dissolved Gas Data Using Recursive Feature Elimination Method

Dissolved gas analysis (DGA) of insulating oils is one of the most popular methods to detect incipient faults in power transformers. However, appropriate feature selection is crucial for accurately detecting incipient faults using DGA data. Another issue is the unavailability of a balanced DGA dataset, which can hamper the fault classification accuracy. Considering these two issues, this article proposes a novel and accurate fault classification framework using gas ratios as features obtained from the DGA data of power transformers. The obtained unbalanced DGA data was initially balanced using the synthetic minority oversampling technique (SMOTE) in the data pre-processing stage. Following this, an efficient feature selection algorithm, namely, recursive feature elimination (RFE) was used to select the best possible features prior to the fault classification using three benchmark machine learning (ML) classifiers, namely, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${k}$ </tex-math></inline-formula> -nearest neighbor (KNN), multiclass support vector machines (SVMs), and extreme gradient boost (XGBoost). The proposed classification model was tested on the DGA data obtained from the local power utility and on the benchmark IEC TC-10 database. Investigations revealed that the proposed classification model delivered detection accuracy of 98.84% and 97.43%, respectively. The proposed method may be reliably used to diagnose incipient faults in power transformers.

Sensing Incipient Faults in Power Transformers Using Bi-Directional Long Short-Term Memory Network

Dissolved gas analysis (DGA) is a standard technique for detecting incipient faults in oil-immersed power transformers. However, fault sensing accuracy depends on feature selection and the machine learning (ML) algorithm used for fault classification. To overcome these two issues, 50 features were extracted from a DGA dataset of 2242 samples obtained from local power utilities. Two state-of-the-art deep learning algorithms, i.e., long-short-term memory (LSTM) and bi-directional long-short-term memory (bi-LSTM), were used to classify different types of faults and normal conditions. In addition, the proposed method was further verified on IEC TC 10 database. Investigations revealed that both LSTM and bi-LSTM performed better than conventional ML classifiers.

Three‐step imputation of missing values in condition monitoring datasets

Missing values are a common occurrence in condition monitoring datasets. To effectively improve the integrity of data, many data imputation methods have been developed to replace the missing values with the estimated values. However, these methods do not always perform well in datasets containing different types of missing values. Three types of missing data are defined, namely isolated missing value, continuous missing variable, and continuous missing sample. A three-step data imputation method is proposed to sequentially impute these missing values following the principle from easy to difficult. The original time series data is first to split into different segments according to the positions of continuous missing samples. Then, interpolation and space-based methods are applied to sequentially estimate isolated missing values and continuous missing variables in each segment. Finally, a stepwise extrapolation prediction model based on the long short-term memory network is established to repair continuous missing samples between each segment. Two application examples are implemented on different dissolved gas analysis datasets and load datasets. Compared with state-of-the-art techniques, the proposed three-step data imputation method is general and can be applied to many scenarios because it establishes a rational data recovery sequence to accurately repair both stationary and non-stationary condition monitoring data.

Fault diagnosis of oil-immersed power transformers using common vector approach

This paper considers the problem of classifying power transformer faults in the incipient stage by using dissolved gas analysis (DGA) data. To solve this problem with high accuracy, we propose to use the common vector approach (CVA) that is a successful classifier when the number of data is insufficient. The feature vector required for the training and testing phases of the CVA is established by using both raw dissolved gas analysis data and some characteristics extracted from this data. The performance of the proposed method is evaluated over DGA data sets supplied from the Turkish Electricity Transmission Company and is compared with some conventional and intelligent methods in terms of classification accuracy and training/testing duration. The achieved results show that the proposed method exhibits superior performance than that of the other methods compared in the meaning of both diagnosis accuracy and computational time. Analysis performed on the physical faults, where the transformers fault types are verified with the electrical test methods, confirms the validity and reliability of the proposed method, as well. Being free from parameter settings is another advantage of this method for using it in online oil-gas analysis applications.

Improving diagnostic performance of a power transformer using an adaptive over-sampling method for imbalanced data

Dissolved gas analysis (DGA) of insulating oil in power transformers can offer valuable information related to faults. Due to the poor and unbalanced characteristics of typical DGA datasets, which threaten the generalization capability of artificial intelligent (AI)-based models, we propose the use of a new over-sampling technique called ASMOTE (adaptive synthetic minority over-sampling TEchnique) in the pre-processing step to enrich the dataset. ASMOTE can significantly improve the generalization performance of AI-based models by providing a sufficient synthetic dataset to train an AI classifier. To authenticate the effectiveness of the ASMOTE algorithm, we validate the transformer diagnostic accuracy of some typical classification algorithms such as multilayer perceptron (MLP), support vector machine (SVM), and k-nearest neighbor (k-NN) using synthetic datasets created by the SMOTE technique. In addition, the use of DGA ratios is also considered. By investigating the interactions between byproduct gases in insulating oil and transformer faults, the non-code ratios of the dissolved emissions are chosen as the characterizing input to the AI-based models. Moreover, with the ability to extract discriminate faulty information of a transformer from DGA data, MLP is used as a preferable classifier for diagnosing symptoms present in transformers. The empirical results of this study demonstrate that the proposed technique remarkably increases the diagnostic performance of power transformer faults.

Uncertainty-Aware Fusion of Probabilistic Classifiers for Improved Transformer Diagnostics

Transformers are critical assets for the reliable operation of the power grid. Transformers may fail in service if monitoring models do not identify degraded conditions in time. Dissolved gas analysis (DGA) focuses on the examination of dissolved gasses in transformer oil to diagnose the state of a transformer. Fusion of black-box (BB) classifiers, also known as an ensemble of diagnostics models, have been used to improve the accuracy of diagnostics models across many fields. When independent classifiers diagnose the same fault, this method can increase the veracity of the diagnostics. However, if these methods give conflicting results, it is not always clear which model is most accurate due to their BB nature. In this context, the use of white-box (WB) models can help resolve conflicted samples effectively by incorporating uncertainty information and improve the classification accuracy. This paper presents an uncertainty-aware fusion method to combine BB and WB diagnostics methods. The effectiveness of the proposed approach is validated using two publicly available DGA datasets.

Fault Diagnosis of Power Transformers With Membership Degree

Power transformers are important equipment for power systems, and a dissolved gas analysis (DGA) is widely used to detect incipient faults in oil-pregnant transformers. The conventional methods are prone to misinterpreting the gas data near the boundaries and the correct rate is low. Though a high correct rate is reported with intelligent methods as artificial neural network, support vector machine, and so on, these methods are usually too complicated to be implemented practically on a wide range. Based on clustering techniques, this paper proposes a new method for fault diagnosis of transformers with the DGA. A reference fault set is provided, and the fault diagnosis is implemented by calculating the membership of the DGA data to the reference fault set. Test with credible DGA dataset (201 field cases) shows that the correct rate of the new method is 89%, while the David triangle method is 79% and the IEC ratio method is 59%, which demonstrate the superiority of the proposed method to the conventional ones. The new method is simple and highly accurate, indicating a good application prospect in engineering practice.

FINNIM: Iterative Imputation of Missing Values in Dissolved Gas Analysis Dataset

Missing values are a common occurrence in a number of real world databases, and statistical methods have been developed to deal with this problem, referred to as missing data imputation. In the detection and prediction of incipient faults in power transformers using dissolved gas analysis (DGA), the problem of missing values is significant and has resulted in inconclusive decision-making. This study proposes an efficient nonparametric iterative imputation method named FINNIM, which comprises of three components: 1) the imputation ordering; 2) the imputation estimator; and 3) the iterative imputation. The relationship between gases and faults, and the percentage of missing values in an instance are used as a basis for the imputation ordering; whereas the plausible values for the missing values are estimated from k-nearest neighbor instances in the imputation estimator, and the iterative imputation allows complete and incomplete instances in a DGA dataset to be utilized iteratively for imputing all the missing values. Experimental results on both artificially inserted and actual missing values found in a few DGA datasets demonstrate that the proposed method outperforms the existing methods in imputation accuracy, classification performance, and convergence criteria at different missing percentages.

Support Vector Machine-Based Fault Diagnosis of Power Transformer Using k Nearest-Neighbor Imputed DGA Dataset

Missing values are prevalent in real-world datasets and they may reduce predictive performance of a learning algorithm. Dissolved Gas Analysis (DGA), one of the most deployable methods for detecting and predicting incipient faults in power transformers is one of the casualties. Thus, this paper proposes filling-in the missing values found in a DGA dataset using the k-nearest neighbor imputation method with two different distance metrics: Euclidean and Cityblock. Thereafter, using these imputed datasets as inputs, this study applies Support Vector Machine (SVM) to built models which are used to classify transformer faults. Experimental results are provided to show the effectiveness of the proposed approach.

Association Rule Mining-Based Dissolved Gas Analysis for Fault Diagnosis of Power Transformers

This paper presents a novel association rule mining (ARM)-based dissolved gas analysis (DGA) approach to fault diagnosis (FD) of power transformers. In the development of the ARM-based DGA approach, an attribute selection method and a continuous datum attribute discretization method are used for choosing user-interested ARM attributes from a DGA data set, i.e. the items that are employed to extract association rules. The given DGA data set is composed of two parts, i.e. training and test DGA data sets. An ARM algorithm namely Apriori-Total From Partial is proposed for generating an association rule set (ARS) from the training DGA data set. Afterwards, an ARS simplification method and a rule fitness evaluation method are utilized to select useful rules from the ARS and assign a fitness value to each of the useful rules, respectively. Based upon the useful association rules, a transformer FD classifier is developed, in which an optimal rule selection method is employed for selecting the most accurate rule from the classifier for diagnosing a test DGA record. For comparison purposes, five widely used FD methods are also tested with the same training and test data sets in experiments. Results show that the proposed ARM-based DGA approach is capable of generating a number of meaningful association rules, which can also cover the empirical rules defined in industry standards. Moreover, a higher FD accuracy can be achieved with the association rule-based FD classifier, compared with that derived by the other methods.