Decision making on transformer insulation condition based on the evaluated incipient faults and aging stresses has been the norm for many asset managers. Despite being the extensively applied methodology in power transformer incipient fault detection, solely dissolved gas analysis (DGA) techniques cannot quantify the detected fault severity. Fault severity is the core property in transformer maintenance rankings. This paper presents a fuzzy logic methodology in determining transformer faults and severity through use of energy of fault formation of the evolved gasses during transformer faulting event. Additionally, the energy of fault formation is a temperature-dependent factor for all the associated evolved gases. Instead of using the energy-weighted DGA, the calculated total energy of related incipient fault is used for severity determination. Severity of faults detected by fuzzy logic-based key gas method is evaluated through the use of collected data from several in-service and faulty transformers. DGA results of oil samples drawn from transformers of different specifications and age are used to validate the model. Model results show that correctly detecting fault type and its severity determination based on total energy released during faults can enhance decision-making in prioritizing maintenance of faulty transformers.
Power transformers are crucial equipment for viable and dependable performance of a power system. Hence, their continuity of operation is the daily business of power utilities. Accordingly, their opportune upkeep based on detected incipient faults and or deterioration state is essential in realizing this objective. When faults manifestation is evidenced within a transformer, ensuing maintenance planning is of utmost importance; otherwise, operational malfunctions emanate that may affect in system failure.
A long in-service transformer generates gases even at normal working conditions. However, as time passes, it is regularly subjected to electric, mechanical, chemical, and thermal stresses that causes high rate of gases being evolved in the transformer insulation system [
Analysis of the levels and ratios of dissolved combustible gasses in transformer insulating fluids through nonintrusive in-service DGA have grown into one of the most dominant techniques available to diagnose probable transformer incipient faults. Since DGA is a process, several techniques highlighted in [
In this paper, a fuzzy logic fault detection model is developed based on the seven key gases (DGA) interlinked with total energy involved in the faulting process. The fault detection model is centered on seven key gases as the inputs, and the fault type is the output of the fuzzy logic model. In addition, the output also signifies the criticality of the fault stress. Instead of using the energy correction relative factor or the energy weighted ratio of gases, this paper proposes the use of total energy involved in the formation of the fault to determine the fault severity. Although, in DGA-based diagnostics, there are methods which can diagnose faults accurately with fewer number of gases, like three gases in Duval triangles or pentagons, this study adopts the seven key gases approach mainly to impact on quantifying accurately the severity of the detected faults which involves these characteristic key gases. The use of total fault energy helps in quantifying the seriousness of the fault especially in the event that more than one transformer suffers from the same type of fault. Accordingly, insulation deterioration and damage are influenced by the extent to which the fault has occurred. The transformer insulation subjected to the fault with high energy incurs more stresses triggering accelerated insulation deterioration. Additionally, energy of fault approach can also help in determining the severity of multiple incipient faults happening simultaneously in the transformer. For example, the proposed fuzzy-DGA model can detect high arcing fault energy, but at the same time, the insulation is experiencing thermal fault and its severity is shown by significant amount of its oil thermal faulting energy, and it signifies that the transformer is experiencing multiple faults in which its severity can be quantified well by energy of fault approach. Therefore, the asset manager decision will not be biased towards arcing fault only but also on the severity of the thermal-related fault. For that reason, fault type and total energy of the fault indicated by the magnitude of the evolved gases should be taken into consideration in transformer condition monitoring systems. Depending on the amount of energy involved in the faulting process, the asset managers can judge whether to maintain the transformer on-line or off-line depending on the criticality of the faults.
Crude oil is the source of the commonly used liquid insulation in oil immersed transformers which is the mineral oil. This insulating oil mainly comprises of alkanes, aromatic, and hydrocarbons products in different magnitudes. In [
The nonintrusive DGA approach consisting of the seven gases is used in transformer fault diagnosis. From the eicosane (C20H42) molecule, five reactions are used to represent how the decomposition of mineral oil results in evolving of H2, CH4, C2H6, C2H4, and C2H2 gases inside the transformer [
Thermal decomposition of paper insulation consisting mainly of cellulosic material leads to the formation of carbon oxides. As the glycosidic bonds in the cellulose break down, carbon monoxide is one of the byproducts as highlighted by the reaction in the following equation [
Oxidation of carbon monoxide in the presence of oxygen produces carbon dioxide as the byproduct as highlighted in the following equation:
The severity of a fault, electrical and/or thermal, in a transformer can be determined by taking hold of the increasing concentration of the gas responsible for fault [
The amount of energy needed to release the fault gases from the crude oil during the faulting process is considered in determining the transformer fault severity. This faulting energy is calculated using the enthalpy change of reaction (ΔHo reaction) as in the following expression [
Under the standard state, the change in energy that excites the generation of one mole of a molecule from its principle composites is characterized as the standard enthalpy of formation (
Standard enthalpy of formation [
Molecule |
|
Molecule |
|
---|---|---|---|
|
−455.8 |
|
−74.8 |
|
−357.9 |
|
0 |
|
−345.9 |
|
226.8 |
|
−414.6 |
|
52.3 |
|
−314.1 |
|
−84.7 |
|
1273.4 |
|
−110.5 |
|
0 |
|
−393.5 |
The standard enthalpy change of a chemical reaction
Similarly, the enthalpy change of reaction for the remaining dissolved gases can be computed. The calculated enthalpies change of reactions for the dissolved gases using the eicosane decomposition reactions are highlighted in Table
Calculated enthalpies of the fault gases.
Reaction in equation | Gas |
|
---|---|---|
1 |
|
35.1 |
2 |
|
97.9 |
3 |
|
268 |
4 |
|
93.5 |
5 |
|
57 |
6 |
|
101.7 |
7 |
|
283 |
As articulated in [
Therefore, the total fault energy (T.F.E) in kJ/kL evaluated using the dissolved gas analysis is given by equation (
Power transformers faulting usually manifests when the electrical and thermal insulation withstand limits are being exceeded. The dissolved gas analysis (DGA) is commonly used to diagnose these incipient faults within oil immersed transformer. Since transformer incipient faults are categorized into electrical and thermal driven, each fault category evolves certain distinctive gases. In this paper, fault determination is achieved through inputting the key gases concentration determined by DGA in to a fuzzy logic diagnostic tool. The fuzzy logic diagnostic tool is developed upon data driven from Figure
Faults and generated gases.
Fault categories [
No. | Characteristic | Fault type |
---|---|---|
1 | F0 | No fault |
2 | F1 | Partial discharge (PD) |
3 | F2 | Thermal fault of low temperature range |
4 | F3 | Thermal fault of medium temperature range 300°C < |
5 | F4 | Discharge (arc) of low energy (LED) |
6 | F5 | Thermal fault of high temperature range |
7 | F6 | Discharge (arc) of high energy (HED) |
IEEE DGA limits [
Evolved gas | Normal limits (acceptable) (ppm) | Extreme limits (unacceptable) (ppm) |
---|---|---|
Hydrogen | <100 | >1800 |
Methane | <120 | >1000 |
Acetylene | <1 | >35 |
Ethylene | <50 | >200 |
Ethane | <65 | >150 |
Carbon monoxide | <350 | >1400 |
Carbon dioxide | <2500 | >10 000 |
An aspect of artificial intelligence (AI) in the form of fuzzy logic demonstrates the nature of human rational and excise decision-making based on linguistic elucidation in problem solving situations. A fuzzy logic model is established based on the transformer fault assessment diagram in Figure
A set of intuitive rules that define the input-output mapping were developed. Contrasting to mathematical models, the rules are developed in the linguistic form of IF-THEN statements. In this paper, experts experience aided by subjective reasoning was adopted in assigning of weights to different attributes during fuzzy rule formulation of the developed fuzzy models. Fuzzy rule formulation criteria can differ depending on the experts’ experience and the weights assigned to different variables. Typical examples of the formulated rules are read as follows: IF (ethylene is normal) and (ethane is safe), THEN (oil thermal faulting level is safe) IF (acetylene is safe) and (hydrogen is high), THEN (arcing level is safe) IF (paper thermal is high) and (oil thermal is moderate), THEN (paper-oil faulting level is high) IF (thermal is critical) and (electrical is high), THEN (fault stress is high and fault type is F5)
As the transformer mineral oil is thermally overstressed, it breaks down and evolves C2H4 and C2H6 as the chief gases. These gases are soluble in oil; their magnitude determines the extent to which the fault stress reaches. An oil thermal faulting level submodel is established based on these two key gases. The two inputs (C2H4) and (C2H6) universe of discourse were measured on a scale of 0–200 ppm and 0–150 ppm, respectively. The oil thermal faulting level is drawn on a scale of 0 to 1 and is marked severe when reaching to 1.
Overheating of transformer solid insulation leads to paper related faults which manifest in the transformer by evolving high concentrations of carbon monoxide and carbon dioxide. The magnitude and rate of increment of these gases determine the degree of paper faulting within the transformer. CO and CO2 are the primary input variables to the paper thermal fault level fuzzy logic model. The linguistic labels for CO and CO2 are partitioned on a scale of 0 to 1800 ppm and 0 to 12000 ppm respectively, whilst the output of the fault level spans from 0 to 1.
When a transformer is experiencing arcing fault, it is recommended not to retain the transformer in-service until proper maintenance is done. Arcing is a high-energy electrical discharge activity that results in the evolving of C2H2 and H2 as fundamental gases in the transformer insulation system. Accordingly, C2H2 and H2 are the two inputs whose linguistic labels are partitioned on a range of 0–50 ppm and 0–1800 ppm, respectively. From the thermodynamic decomposition of mineral oil, acetylene formation requires more energy relative to that of hydrogen; thus, weighting factors of 0.8 and 0.2 during fuzzy rule formulation were assigned to acetylene and hydrogen, respectively. The output of the model with membership functions between 0 and 1 characterizes the growth of electrical arcing with increase in magnitude of the input variables concentration.
Electrical discharge within the transformer in the form of low energy is quantified as partial discharge (PD). This activity within the transformer results in formation of CH4 and H2 as the principal gases in the transformer insulation system. The magnitude of these gases connotes the amount of buildup of partial discharges in the transformer. Input variables of H2 and CH4 are drawn on a scale of 0–1800 ppm and 0–1200 ppm respectively, whilst the output variable universe of discourse for the PD level is measured on a scale of 0 to 1. Electrical discharge (PD) criticality is deemed serious when reaching to 1.
Transformer thermal faults can be evidenced through paper-oil overheating or in worse scenarios of conductor melting and or transformer explosion. Long-term emerging thermal faulting can be assessed by considering the key gases dissolved in oil which portrays overheating of paper (CO and CO2) and oil (C2H6 and C2H4). Thus, by merging thermal stress in paper and oil, the resulting thermal fault stress enforced to the transformer can be estimated through the fuzzy model. The inputs to the total thermal faulting model are oil thermal stress and paper thermal stress (the outputs of thermal stress level of oil and paper fuzzy models and are drawn on a scale of 0 to 1). Through subjective reasoning and knowledge gained from power utility experts, the assigned weights of 0.6 and 0.4 to thermal level in paper and oil, respectively, are used during fuzzy rule formulation. The paper thermal stress level was assigned more weight since it was deemed dangerous as the paper insulation is in direct contact with live conductors. Thus, failure of paper insulation can lead to catastrophic faulting of the transformer. The thermal fault level output also spans from 0 to 1. The transformer thermal fault stress level for oil-paper for different set of input variables can also be deduced from the surface graph, Figure
Surface graph-overall thermal fault level.
The electrical fault stresses are caused by localized excessive field leading to faulting manifested by partial discharge, tacking, treeing, arcing, flashovers, and short-circuiting [
Surface graph-overall electrical fault level.
Transformer in-service can suffer from electrical and or thermal faults. The overall fault stress and severity determination were arrived at after incorporating the thermal and electrical fault level fuzzy models. The inputs and thermal and electrical membership functions are established on a scale of 0 to 1 as shown in Figures
Input variable MF-thermal fault.
Input variable MF-electrical fault.
Output variable MF-fault stress.
Output variable MF-fault type.
The overall transformer fault stress level for different sets of inputs can also be interpreted from a fuzzy rule surface viewer as shown in Figure
Overall fault stress surface graph.
Based on the concentration of the dissolved key gases evolved in the transformer insulation system, the fault and severity determination model was established. The fault model depends on fuzzy logic-DGA diagnostic tool, whilst severity determination was upon the calculated energy of formation during faulting activity. The output of the fuzzy model shows the estimated fault type and the stress level of the fault (state of the insulation system). The formulated equation for fault severity is as in equation (
The overall proposed model developed upon MATLAB/Simulink platform is depicted in Figure
Proposed fault and severity model.
To evaluate the validity of the proposed methodology for fault identification and severity determination, several oil samples from transformers of different magnitudes and various services spans have been presented as primary data. Dissolved gas analysis was performed in all the oil samples from which the evolved gas concentration was quantified. Since it was difficult to deduce the inaccuracies of distinct instruments and human errors of each data set from different sources, a 95% confidence level was assumed to cater for these uncertainties. The outcome of the developed fuzzy-DGA model was based on the IEEE key gases acceptable limits as denoted in Table
Test data and model results.
Test data | Model results | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Tx. no. | H2 | CH4 | C2H2 | C2H6 | C2H4 | CO | CO2 | Temp. | A.FAULT | E.FAULT | A.F.E.L.L (kJ/kL) | A.F.ENERGY (kJ/kL) | T.F.E (kJ/kL) | Severity |
1 | 234 | 300 | 52 | 56 | 29 | 193 | 713 | 50 | F4 | F4 | 1.3 | 3.7 | 15.7 | Low |
2 | 607 | 119 | 0 | 192 | 5 | 161 | 2320 | 62 | F4 | F4 | 1.3 | 6.7 | 44.0 | Low |
3 | 11 | 4 | 1 | 4 | 4 | 558 | 7624 | 48 | F2 | F2 | 39.4 | 116.3 | 116 | Medium |
4 | 74 | 65 | 35 | 194 | 172 | 1272 | 10529 | 55 | F5 | F5 | 40.3 | 136.2 | 137.6 | High |
5 | 71 | 347 | 18 | 127 | 79 | 504 | 2481 | 52 | F2 | F2 | 38.2 | 40.8 | 42.5 | Low |
6 | 16 | 34 | 10 | 14 | 16 | 188 | 1656 | 44 | F0 | F0 | — | — | 24.8 | — |
7 | 979 | 236 | 112 | 183 | 180 | 1843 | 8492 | 63 | F5 | F5 & F6 | 41.3 & 1.3 | 143.8 & 12.6 | 156.5 | High |
8 | 113 | 24 | 61 | 20 | 57 | 32 | 871 | 36 | F4 | F4 | 0.5 | 1.4 | 14.9 | Low |
9 | 294 | 748 | 6 | 212 | 1348 | 242 | 2286 | 41 | F3 | F4 | 1.2 | 4.4 | 46.0 | Low |
10 | 163 | 106 | 9 | 298 | 1517 | 213 | 1303 | 58 | F3 | F3 | 0.5 | 8.6 | 31.8 | High |
11 | 151 | 8 | 8 | 151 | 10 | 86 | 1538 | 48 | F0 | F0 | — | — | 25.3 | — |
12 | 678 | 368 | 163 | 92 | 108 | 216 | 2211 | 52 | F6 | F6 | 0.5 | 5.8 | 43.5 | High |
13 | 893 | 724 | 1 | 6 | 18 | 350 | 2207 | 55 | F1 | F1 | 0.8 | 5.5 | 22.8 | High |
14 | 195 | 660 | 22 | 127 | 79 | 607 | 3674 | 50 | F4 | F4 | 40 | 100.9 | 102 | High |
15 | 440 | 522 | 183 | 31 | 62 | 428 | 1232 | 48 | F6 | F6 | 1.3 | 12.0 | 33.0 | High |
16 | 15 | 8 | 0 | 9 | 5 | 168 | 1549 | 53 | F0 | F0 | — | — | 24.3 | — |
17 | 1176 | 3426 | 0 | 1178 | 2931 | 299 | 3400 | 51 | F2 | F1 & F2 | 39.8 & 1.3 | 70.7 & 18.6 | 89.3 | High |
18 | 441 | 678 | 0 | 73 | 62 | 302 | 492 | 47 | F1 | F1 | 0.7 | 3.5 | 13.9 | Low |
19 | 358 | 260 | 5 | 66 | 55 | 4288 | 11492 | 38 | F3 | F3 | 37.8 | 187.5 | 192.1 | High |
20 | 1498 | 395 | 92 | 323 | 395 | 487 | 3176 | 42 | F6 | F5 & F6 | 0.4 & 1.2 | 2.8 & 17.1 | 68.8 | Low/high |
In Table
As an illustration, the model’s results for transformer 7 are also shown in Figure
Table
Comparison of different fault detecting diagnostic approaches with respect to fault type.
Fault type | Total faulty cases | Fuzzy Duval-EWR method [ |
Fuzzy IEC-EWR method [ |
Proposed fuzzy key gas-TFE method | |||
---|---|---|---|---|---|---|---|
Correct diagnosis | % accuracy | Correct diagnosis | % accuracy | Correct diagnosis | % accuracy | ||
Partial discharges | 7 | 7 | 100 | 7 | 100 | 7 | 100 |
Discharges of low energy | 11 | 10 | 91.9 | 10 | 91.9 | 9 | 81.8 |
Discharges of high energy | 22 | 21 | 95.9 | 21 | 95.9 | 20 | 90.9 |
Low-level thermal faults | 13 | 12 | 92.3 | 13 | 100 | 13 | 100 |
Medium-level thermal faults | 19 | 16 | 84.2 | 16 | 84.2 | 15 | 78.9 |
High-level thermal faults | 12 | 10 | 83.3 | 10 | 83.3 | 9 | 75 |
Comparison of severity accuracies of fault diagnosis models.
Method | Fuzzy Duval-EWR method [ |
Fuzzy IEC-EWR method [ |
Proposed fuzzy key gas-TFE method |
---|---|---|---|
Total fault cases tested | 87 | 87 | 87 |
No. of correct diagnosis of faulty cases | 79 | 78 | 75 |
Fault detecting % accuracy | 90.8 | 89.7 | 86.2 |
Fault severity % accuracy | 88 | 90.7 | 94.7 |
The benefits of considering energy involved during faults are that severity of individual fault from multifaulting can be easily noted. In addition, the overall faulting energy can help asset managers to quantify the overall severity of the faulting transformers in order to rank them for maintenances. Faulting resulting in high-energy intensity initiates intense harm on the insulation system.
Early detection of power transformers internal faulting is vital and effective in minimizing asset damages, economic loss, and effects on reliability of the overall power system. From the nature and concentration of the evolved gas, the fault type can be determined. In this paper, a fuzzy-DGA-based diagnostic tool was developed to detect faults and condition of the transformer, whilst energy of fault of involved fault gases was used in fault severity determination. The seven keys gases paired according to their cross-correlation in signifying the nature of faults were used as inputs to the developed fuzzy logic model. For severity determination, enthalpy energy of change of reaction of fault gases was established upon the eicosane (C20H42) as the starting decomposition material. Simulation results shows that the model managed to correctly detect the faults encountered by the different transformers. In addition, energy involved during fault proved to be an effective method in determining fault severity as it can reflect the extent of insulation damage. Since asset managers’ primary business is to enable reliable performance of its assets, such that in the event of faults, it is recommended that fusion of fault type and fault energy can be an effective method of classifying maintainable faulty transformers.
The dissolved gas analysis (DGA) data used to support the findings of this study are included within the article (Table
The authors declare that they have no conflicts of interest.
This research was supported by the Pan African University Institute for Basic Sciences, Technology and Innovation in the form of a postgraduate student research funding.