Search

Very limited studies have forecast the gas concentration and faults of high voltage direct current (HVDC) converter transformers. In this study, a hybrid algorithm consisting seasonal autoregressive integrated moving average-convolutional neural network-gated recurrent unit (SARIMA-CNN-GRU) was proposed to forecast the seven gas concentrations, i.e., hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), acetylene (C2H2), carbon-dioxide (CO2), and carbon monoxide (CO). Moreover, extreme gradient boost (XGBoost), adaptive boosting (AdaBoost), and light gradient boosting machine (LightGBM) were utilized to predict the fault inside the HVDC converter transformer. Two types of datasets collected from different sources were used to train the regression and classification models. The performance of the regression model was evaluated by root mean squared error and mean absolute error, whereas the performance of the classification model was evaluated by accuracy, precision, and recall. Finally, a comparative analysis was conducted to showcase the superiority of the proposed methodology.

In this article, the lightning impulse breakdown (LIBV) behavior of two low pour point synthetic ester liquids is presented in comparison to a typical synthetic ester (TSE) at both positive and negative polarities. Traditional mineral insulating oil has also been considered for reference purposes. A detailed breakdown behavior analysis of the four test liquids under a nonuniform field (medium gap, point-plane electrode system) and quasi-uniform field (smaller gap, U-plane electrode system) is envisaged. The LIBV measurements based on the source voltage waveforms and light activity during the discharge process are presented. The Weibull breakdown failure rates and streamer velocity during the breakdown of different liquids for all the cases (± polarities and both electrode configurations) are reported in support of the discussions. In the case of nonuniform fields, the lightning breakdown voltage of the low pour point liquids is found to be higher than TSE s and is comparable to mineral oil (MO) under both polarities. While in the case of quasi-uniform field, the lightning breakdown voltage of the low pour point liquids is found to be lower than MO and comparable to the TSE under both polarities. These findings add to limited knowledge on the application of esters in cold countries and allow insulation designers to estimate the behavior of the low pour point synthetic ester liquids under lightning conditions.

The furan prediction in the transformer oil can provide information about the oil degradation rate and overall health of the oil-filled transformer. For this study, the furan content in mineral oil of a 300-MVA converter transformer installed at ±500-kV high-voltage direct current (HVDC) inverter station is taken. Six features are taken as an input to the machine learning (ML) model to predict the furan content for predictive maintenance of the converter transformer, i.e., carbon monoxide, carbon-di-oxide, acidity, breakdown voltage (BDV), total combustible gases, and water content. A hybrid ML model is developed consisting of a convolutional neural network-long short-term memory network (CNN-LSTM) and extreme gradient boosting (XGBoost). Then, the mean absolute percentage error with recursive weighting (MAPE-RW) technique is employed to combine the predictions of two models to achieve better outcomes than a single model. For this purpose, a total of 350 data are collected from the local utilities for training and testing. Moreover, a total of 41 numbers of data are collected from the converter transformer for validation of the proposed model. The performance of the proposed algorithm is evaluated by mean absolute error (MAE), root means squared error, and coefficient of determination. The hybrid model is also compared with other models to show its efficacy and robustness. Depending upon the furan prediction, converter transformer insulation health is classified into three classes for preventive maintenance.

One of the major issues with natural ester fluid as transformer insulation is encountered in the oxidation process, which forms a thin gel layer due to polymerization. This latter can hinder the electrical and heat transfer properties. In order to understand the process, methodical thermal aging of natural esters is carried out for a 500-h duration under atmospheric conditions at 160 °C. The characteristic variation of fluid-forming gel is investigated by nuclear magnetic resonance (NMR) with thermal aging, through which molecular weight degradation and iodine index of thermally aged oil are estimated for transformer applications. The molecular weight and iodine index of thermally aged oil are reduced to 8.31% and 43.016%, respectively, after 500 h of aging. Thermogravimetric analysis (TGA) is carried out to further comprehend this, and heating calorific value is estimated through bomb calorimetric studies. Heat calorific value (HV) for 500-h aged oil and gel is nearly similar. The electrical discharge performance of the oil is estimated through corona inception voltage (CIV) studies under nonuniform electric field configuration adopting ultrahigh frequency (UHF) sensor techniques. The iodine index is in direct correlation with the CIV measurement.

In the present study, rheological studies dealing with the flow behavior of thermally aged natural ester fluid under laboratory-controlled conditions are performed. Visual observation of gelling is witnessed due to thermal aging. An enhancement in viscosity of about 188% is observed with the 500-h aged fluid compared to the nonaged fluid. It is estimated that the flow behavior altered by gel follows non-Newtonian behavior with a shear-thinning effect on applied shear rate conditions. The dependency of storage and loss modulus to amplitude and frequency sweep under oscillatory shear flow was investigated. The viscous state of gel dominates over its initial elastic state, beyond the gel point. The flow models that govern the behavior of the gel were applied, adopting nonlinear least square methods to investigate the rheological parameters in terms of flow behavior index and yield stress. The results of the thermally aged natural ester fluid demonstrate a positive correlation with Bingham’s model and Mizrahi–Berk’s model based on the yield stress and flow behavior index, respectively. The phase transition of gel investigated in the present study can be used as a time-based preventive measure, by the transformer manufacturer, in the heat transfer performance of natural ester-filled transformers.

In this article, some investigations on the prebreakdown and breakdown phenomena of low pour point insulating liquids under negative lightning impulse (LI) voltage are reported. The tested liquids include mineral oil (MO), a typical synthetic ester (TSE), and two low-pour point synthetic esters. These liquids underwent accelerated thermal aging. The nonaged and aged samples were subjected to LIs using a point-plane electrode arrangement. The discussions are focused on the initiation of partial discharges, propagation of streamers, and breakdown behavior in the nonaged and aged liquids. The investigated parameters include inception voltage, LI breakdown voltage, streamer acceleration voltage, and streamer velocity. The results are supported by the oscillographs of the light activity that is recorded during the discharge process. The prebreakdown phenomenon noticed in the TSE vis-à-vis mineral insulating oil is in line with the existing literature. Importantly, it is noticed that the inception and breakdown voltages of the nonaged low pour point synthetic esters are similar to nonaged MO. In addition, the inception and breakdown voltages of the aged low pour point synthetic esters are noticed to be higher than that of the aged MO. These results add to the arguments in favor of replacing MOs in power transformers.

Various techniques of fault diagnosis in oil immersed power transformers have been utilized for the last two decades and are well reported in the literature. The traditional and widely accepted dissolved gas analysis (DGA) methods include IEC gas ratio code, IEEE Key gas, Doernenburg ratios, and the Duval triangle. Amongst these methods, Duval triangle method (DTM) is reported to be the most prominent. A new fault gas interpretation approach has been proposed in the present work. This method combines three Duval triangles with triangular membership functions (MFs) evaluated empirically for diagnosing the incipient faults. Like the existing Duval triangles, five gases, namely, hydrogen, methane, ethylene, ethane, and acetylene, are considered for the present study. Initially, the relevant gas percentages for the three Duval triangles are calculated and later converted into triangular fuzzy MFs. Finally, an expert rule-based method synthesizes the outcomes into a specific main fault type or subfault type. Test data of 139 transformers are used to compare the proposed method with the conventional DTM. The proposed approach is simple and can potentially determine the main faults and subfaults simultaneously based on numerical indices in a single fuzzy logic (FL) system. In addition, the proposed method has overcome the ambiguity in overlapping faults in the traditional methods.