Reliable Estimation for Health Index of Transformer Oil Based on Novel Combined Predictive Maintenance Techniques

Transformer oil insulation level may be deteriorated due to electrical and thermal faults, which may lead to transformer failure and system outage. In this regard, the first part of this paper presents comprehensive maintenance for power transformers aiming to diagnose transformer faults more accurately. Specifically, it aims to identify incipient faults in power transformers using what is known as dissolved gas analysis (DGA) with a new proposed integrated method. This proposed method for DGA is implemented based on the integration among the results of five different DGA techniques; 1) conditional probability, 2) clustering, 3) Duval triangle, 4) Roger’s four ratios refined, and 5) artificial neural network. Accordingly, this proposed integrated DGA method could improve the overall accuracy by 93.6 % compared to the existing DGA techniques. In addition, the second part used for predictive maintenance is based on determining the health index for a new and an aged power transformer (66/11 kV, 40 MVA) filled with NYTRO 10XN oil by evaluating the breakdown voltage, DGA, moisture content, and acidity for the oil itself. In the breakdown voltage test, two practical types of transformer oil; transformer oil type (NYTRO 10 XN) and another type HyVolt III alongside their mixtures are evaluated and compared, where this mixture has not been reported in the literature. In addition, aged oil samples extracted from a real case study in-service transformer during operation with different aged durations; 9, 10, 11, 12, and 13 years, are tested for breakdown voltage and then compared with fresh oil samples. In the DGA task, the temperature rise test is carried out firstly on the case study, and hot dc resistance of winding is measured under different cooling systems. Considering five transformers of the same rating and oil type, the DGA was performed on oil samples extracted before and after the temperature rise test. The DGA results predicted that there was a possibility exists of local overheating. While the complementation of oil breakdown voltage technique for dielectric severity indication and temperature rise technique for thermal severity indication in addition to moisture and acidity measurements can lead to an effective maintenance plan for power transformers into the field. Hence, by using the proposed multi-criteria analysis approach (i.e., breakdown, acidity, moisture, and DGA results), the various factors are combined into a condition-based Health Index, yielding reliable estimation results compared to individual approaches.

Power transformers are an essential apparatus in the electrical power network. It is an expensive item in electrical networks. The failure of a power transformer leads to substantial costs either due to system outage or due to increased maintenance costs [1]. Oil is used in power transformers as an insulating medium and a heat transfer medium. The transformer oil insulation level shall be deteriorated by electrical and thermal possible faults. These possible faults can be classified into electrical and thermal types. Electrical fault types include partial discharge (PD), low and high arching discharge (D1 and D2), while thermal fault types include low, medium, and high thermal faults (T1, T2, and T3) [2,3]. Hence, the detection of oil insulation levels, with continuous monitoring of oil insulation levels, plays an important role in increasing the power system reliability and reducing the risk of power transformer failure. Electrical and chemical test techniques are used as diagnostic techniques for detecting the oil insulation level. Dissolved Gas Analysis (DGA) is one of the most common chemical methods used for detecting the insulation fault type [4][5][6]. During the normal transformer operation, some low concentrated gases are released such as Hydrogen (H2), Methane (CH 4 ), Acetylene (C 2 H 2 ), Ethylene (C 2 H 4 ), Ethane (C 2 H 6 ), Carbon monoxide (CO), and Carbon Dioxide (CO 2 ). While during fault conditions, these gases are produced with high concentration levels, and hence DGA techniques depend on detecting the released concentration level in case of thermal and electrical stresses [7,8].
Many conventional techniques were developed and used for interpretation of transformer incipient faults based on DGA such as key gas ratio [9], Rogers' ratio [9,10], IEC gas ratio [10], Dornenburg ratio technique [9], and Duval Techniques [11,12]. The conventional approaches do not often yield accurate analysis while missing important incipient faults, thereby leading to the 'no decision' issue [7]. Most of the existing DGA techniques have limited diagnostic accuracy and may fail to interpret the oil faults in transformers [13]. More studies have been presented to improve the traditional interpretation techniques using artificial intelligence (AI) and computational techniques [14][15][16]. However, most AI techniques are complicated and difficult for practical engineers to apply. Hence, this research topic is still open to improving the diagnostic accuracy of transformer faults. Recently, new approaches have been presented including a combination and integration between the diagnosis results of multiple techniques. Four techniques have been presented with an accuracy of 69.96%, 84.96%, 83.27%, and 85.3% based on using 532 samples [17]. Two Graphical Shapes are presented to improve the diagnosis accuracy of the transformer fault types using the DGA approach [18]. In addition, an approach is presented based on the limits of new gas concentration percentages to improve the faults diagnostic accuracy [19]. In [20], a model of convolutional neural network (CNN) is studied to accurately predict the types of faults under different levels of noise in measurements. In addition, another proposed DGA technique based on extreme machine learning is presented as in [21].
For power transformer, a new maintenance decision-making model is proposed based on economy and reliability assessment [22]. According to support vector machine (SVM), a novel technique is proposed to improve the diagnostic accuracy of transformer fault types [16]. DGA lacks two main issues. Firstly, its accuracy needs to be increased for reliable fault diagnosis. Secondly, it cannot provide sufficient information about the condition of the transformer oil itself. On the other side, different approaches of diagnosis and monitoring are used, based on a wide variety of thermal, mechanical, electrical, and optical effects. These approaches provide information on aging, allowing the state assessment, and recommend measures to improve the quality of the insulation and lifetime assessment [23,24]. One of these approaches is the health index, whose evaluation is based on different factors [23,25]. These factors include dielectric strength measurement, DGA, turns ratio, leakage reactance, winding resistance, and degree of polymerization for paper insulation.
For breakdown voltage test, it is an indicator for dielectric severity in power transformers [26]. In [27], dielectric characteristics of transformer mineral oil with different compositions of corn oil are studied. In addition, the effect of the natural fiber on the dielectric strength of natural ester as a liquid insulating material is presented as in [28]. In addition, in [29], a comparative study of the positive and negative DC breakdown voltages of different transformer oil is proposed. The aging characteristics of mineral oil are studied [30]. Further, a temperature rise test is an indicator of thermal severity in power transformers [31]. According to the literature review, most existing DGA techniques can have poor diagnostic accuracy for certain fault types, which in turn can fail to interpret the diverse transformer oil faults. Based on this notice, a combined DGA method could distinct the conflict among the various transformer oil faults, yielding higher accuracy rates. Another gap in the literature is that the health condition index of power transformer oil is estimated by individual approaches, which can lead to unreliable estimation results, which is covered in this work.
To cover the abovementioned gap in the literature, this paper presents comprehensive predictive maintenance for power transformers based on a new DGA technique alongside additional healthy index evaluation that is based simultaneously on the breakdown voltage test, acidity and moisture measurements, and DGA before and after temperature rise test. First, the new DGA technique is presented and developed using 360 samples of known actual faults collected from the Egyptian Electricity network. Additional 36 validation samples (10 % from all datasets) are tested to check the accuracy, then the proposed combined technique accuracy is compared with its individual techniques as well as the recently published techniques. Next, a health index calculation is presented, based on the test results of a transformer insulation system (breakdown strength, dissolved gas analysis, moisture, and acidity) for new and aged transformers. Transformer case study of rating (66/11) kV of NYTRO 10 XN oil type is collected from ELMACO Company, Egypt. For calculation of the health index transformer insulation system, breakdown voltage tests for two practical types of transformer oil; NYTRO 10 XN and HyVolt III alongside their mixtures are estimated and compared. The oil mixture has not been reported before in the literature. It is used to take advantage in advance in the absence of any of the two oils. Finally, a temperature rise test is performed on five new transformers of the same rating with a comparison between dissolved gases before and after the temperature rise. In addition, winding resistance is measured in this analysis. Also, acidity and moisture are measured for the same oil type during its first year in service. The same tests are carried out for a real case study in-service transformer for studying its health index.
The results show that by using more than one diagnosis criteria (i.e., breakdown, moisture, acidity, and DGA results), various factors are joined into a condition-based Health Index, providing reliable estimation results compared to individual DGA approaches. To prove this unique feature of the proposed approach, the health index of the transformer insulation system was examined using more techniques: DGA alone and DGA plus breakdown voltage (BDV), moisture, and acidity. As noticed, we can predict the risk of local overheating using DGA criteria before and after the temperature rise test. While the health index was produced based on the breakdown, acidity, moisture, and DGA data, it revealed that the insulation state was quite excellent, with a projected lifetime of more than 15 years with routine maintenance needs for the new transformer. While, for 13 years in-service aged transformer, the transformer lifetime is 3 years with poor condition. When the status of transformers in service was studied, the health index result utilizing more than one diagnosis criteria (i.e., breakdown, moisture, acidity, and DGA results), was closer to reality than using DGA alone.

II. PROPOSED DGA BASED FAULT DIAGNOSTIC TECHNIQUE
The existing interpretation techniques have not provided properly accepted results for detecting transformer faults. Hence, it is an open research area for new proposed techniques to overcome the problems of existing DGA techniques. The accuracy limit of conventional DGA techniques is still a major problem for diagnosing transformer faults due to interference between electrical and thermal faults, which in turn can fail interpretation of the transformer oil fault. A newly proposed technique for fault diagnosis is prepared and presented. This proposed technique is developed based on the integration among the results of conventional and non-conventional techniques. These techniques are five techniques; conditional probability [32], clustering [33], Duval, Roger's four ratios refined [7,34], and artificial neural network (ANN) [35] techniques. A total of 360 samples of known actual faults are used in developing the proposed technique, and also are used as a dataset for testing the technique accuracy. A total 120 samples have been collected from the Central Laboratory of the Egyptian Ministry of Electricity and 240 samples are extracted from literature [33]. Table I summarizes the distribution of all these samples for various fault types.
The selected samples have a convergence between the numbers of samples for each fault type avoiding a noticeable discrepancy between number of samples for each fault used in the literature [17]. The used datasets have been analyzed based on the IEEE Std. C57.104-2008 [9]. Fig. 1 presents the gas concentration percentage (GCP) for each fault type. GCP is the parentage of each gas divided by the sum of the five main gases (H 2 , CH 4 , C 2 H 2 , C 2 H 4, and C 2 H 6 ) as shown in the following equations from (1) to (5).   (2) Clustering method [33] (3) Duval method [11] (4) Rogers refined method [34] (5) Rogers, IEC [10] and Duval Methods using ANN [35] ₂ ₂ = C₂H₂ H₂+CH₄+C₂H₄+C₂H₆+C₂H₂ % (5) Fig. 1(a) shows that H 2 is the most generated gas due to the PD fault type. Hence, H 2 is the main key gas in partial discharge. Fig. 1(b) shows that Hydrogen (H 2 ) is the dominant key gas in the case of D1. Acetylene (C 2 H 2 ) is the dominant gas in case of high discharge Energy (D2), see  [9]. Hence, the accuracy and reliability of the laboratory data set are confirmed as shown in Table II.
The new technique for the fault diagnosis is prepared based on the integration among the results of conventional and non-conventional techniques. Hence, we tried to get the best techniques having good accuracy. These techniques are five techniques; conditional probability, clustering, Duval, Roger's four ratios refined, and ANN as shown in Table III.
Techniques (1, 2, 3, 4, and 5) are based on the five main gases (H 2 , CH 4 , C 2 H 6 , C 2 H 4 , and C 2 H 2 ). Techniques (1 and 2) are based on the main five gases with respect to their summation [32,33] as shown in equations from (1) to (5). The technique (3) is based on the percentages of the three gases ratio (R₁, R₂, and R₃) as shown in equations from (6) to (8). The four ratios (R₄, R₅, R₆, and R₇) are used for fault diagnosis using Technique (4) [34] as shown in equations from (9) to (12). While, in [35] we find the technique (5) is based on techniques (3) and (4), in addition to the IEC DGA technique of ratio as in equations (9, 11, and 12) with the application of ANN.
Using the 360 samples, the five techniques diagnosed the fault type, and these diagnoses are compared with the actual fault.   proposed technique can be accomplished via the flowchart in Fig. 2. The procedures of the proposed technique based on the five techniques diagnosis are as follows: Stage 1: If the detection result of conditional probability and clustering is partial discharge, the fault type is diagnosed as partial discharge. Stage 2: The same condition as before but, for high discharge energy, the fault type is diagnosed as high discharge energy. Stage 3: If Roger's result is a low thermal fault type, the diagnosis is a low thermal fault type. Stage 4: If the diagnosis Duval is low discharge and conditional probability result is low or high discharge energy, the low discharge energy diagnosis is diagnosed. Stage 5: If Rogers and conditional probability methods diagnosed high thermal fault type, the proposed technique result is high thermal fault type. Stage 6: Medium thermal fault type is detected if the result of conditional probability is medium thermal fault and results of clustering, Duval and Rogers are low or medium or high thermal fault types. If no condition is satisfied, the result is a diagnosis of conditional probability.  The accuracy of the proposed technique is compared with the five techniques, conditional probability, clustering, Duval, Rogers refined, and ANN, in addition, it is compared with the previous techniques (a, b, c and d) in [17]. These previous techniques were constructed by combining the results of previous studies. The technique (a) was formed based on the outputs of three methods, Roger's four ratios refined, Duval, IEC refined techniques, while technique (b) was constructed based on three (DGA) methods (Duval triangle, clustering, and conditional probability). The technique (c) was preceded depending on the outputs of two (DGA) methods (Roger's refined method and the artificial neural network (ANN) technique). Finally, technique (d) was based on the combined outputs of techniques (b) and (c) [17]. The improved accuracy of the proposed technique is proven using a large number of datasets (360) of known actual faults. For these samples, the convergence between each fault number is taken into account. The accuracy of central laboratory results is proved by analyzing the 360 samples and comparing the key gas for each fault with it in the standard [9].  [17,19,[32][33][34] and laboratory, as shown in Table VII. These datasets of known incipient fault have been used for checking the accuracy of the proposed technique. The proposed technique accuracy, for these datasets, was compared with its five construction techniques as well as the previously published techniques (a, b, c, and d). The overall accuracy of the proposed technique keeps up its improvement compared to these techniques (a, b, c, and d), as shown in Table VIII and Fig. 5. The proposed technique is considered one of the techniques that achieved an accuracy of more than 91.5 % for the validation samples. As given in this comparison in Fig. 5, the robustness of the accuracy of the proposed method is proven to be higher than the other techniques.

III. TRANSFORMER CASE STUDY
This part introduces the transformer real case study of rating (40 MVA, 66/11 kV) with NYTRO 10XN oil type. The first section of this part presents the breakdown voltage test for the oil type of the study case and its comparison with another type; (HyVolt III) and their mixtures. The second section studies the temperature rise test effect on winding resistance and DGA. Then acidity and moisture are measured for the same case and are presented in the third section. Finally, the healthy index is calculated using the breakdown voltage measurement, dissolved gases concentrations, and the acidity and moisture for the new transformer and an aged transformer.
where α and β are defined as the scale and the shape parameters respectively. These parameters are calculated by the following equation (14) [39].
In the BDV test, the variable x referred to the breakdown voltage, and the probabilities of different breakdown voltages were estimated using Weibull distribution. Hence, the BDV at the probability of 50% refers to the mean value.
According to the breakdown test for NYTRO 10 XN, and HyVolt III and their mixture by 50% oil types, the mean values at the probability of 50% was 71.1 kV, 48.9 kV, and   a and 6.b. Figure 6 shows that by mixing 50% of the two oil types; as NYTRO 10 XN and HyVolt III, the dielectric strength was slightly improved compared to pure HyVolt III. However, the type of NYTRO 10 XN oil remains the best in terms of dielectric strength.
For Nytro 10 XN oil type, oil samples extracted from a real in-service aged transformer of rating (40 MVA, 66/11 kV) with different operation durations; 9, 10, 11, 12, and 13-years. These samples have been tested for a breakdown voltage test. Twelve measurements were taken for each sample, and their average breakdown values are 77, 75, 27, 65, and 38 kV, respectively, as shown in Figures 7.a and  7.b. Comparing the BDV of the aged samples and graded fresh samples, the average BDV was good after 9-and 10years in service. Further, after 11 operation years, the BDV was reduced effectively till reaching 27 kV, which is below the acceptable limit, due to aging occurrence. So that, the oil was treated and filtered, then the transformer returned to the service again. Moreover, after the oil treatment and with one more year in operation, the breakdown voltage was 65 kV after 12 years. Finally, after 13 years in operation, the BDV decreased to 38 kV.
The breakdown voltage of the fresh and aged samples (13 years in-service) shall be used as an indication for the health index of oil quality for a new transformer and another aged one [41,42].

B. TEMPERATURE RISE TEST
Temperature is a quantitative measure of thermal energy contained in an object, the higher the thermal energy, the higher the temperature. The temperature rise is always defined as the difference between the temperature of an object and the temperature of the corresponding cooling medium (e.g., oil). The thermal run test or overheat test is one of the type tests prepared on the power transformer. The mean purpose of this test is to check whether the oil and winding temperatures of the transformer meet the values specified in the standard and technical projects. Temperature rise test is performed according to IEC 60076-2 standard [43], low voltage winding is a short circuit, the voltage is applied to high voltage winding, and temperature rise test for top oil of transformer should be continued until the winding and oil in the tank have reached a stable temperature [43]. This test was carried out on a transformer of rating (40 MVA-66/11 kV) with oil type (Nytro 10 XN). The temperature rise test objective is to measure the winding resistance and to predict the possibility fault based on the rate of change in dissolved gases concentrations before and after the test and compare its health index calculated with health index measurement using DGA plus acidity, moisture, and breakdown tests. Hence, this test is used for checking the transformer health index with a slight effect on the deterioration of the insulating paper or production of Methane, Ethan, and Ethylene gases with acceptable limits. In addition, these gases that were produced after the temperature rise test help us in the fault prediction that can occur compared with their values before performing this test.

1) MEASUREMENT OF THE WINDING RESISTANCE
The measurement values should be carried out loading ONAN, ONAF (100% load), and ONAF (110% load for two hours) according to IEC 60076-2 [43]. The measurement of the winding resistance is started after the shutdown of the test power and connection to the winding of the DC measuring device. The winding temperature at the top of the winding does not correspond to the real winding hot spot. This is due to the influence of the increased stray loss caused by the horizontal component of stray flux. The winding oil gradient g at the top of the winding has to be multiplied by the factor H. this H factor may vary from 1.1 to 1.5 depending on transformer size, short circuit impedance, and winding design, in case study 1.3, according to equation (15).
The instrumentation used for the measurement for automatic recording, a considerable number of discrete readings are being made over a period of time, and these have to be evaluated for extrapolation backward in time to the instant of the shutdown. The average winding temperature variation after shutdown for secondary winding with cooling secondary winding resistance 250.1 mΩ at oil temperature 28 ⁰C, and primary winding with cooling primary winding resistance 29.25 mΩ at temperature 28 ⁰C, shown in Figs. 8 and 9.

2) DISSOLVED GAS ANALYSIS BEFORE AND AFTER THE TEMPERATURE RISE TEST
On the other hand, DGA approach for transformer oil, the first sample was taken before starting temperature rise and for oil forced cooling systems; the second sample was taken approximately 24 hours after the temperature rise test, according to IEC Std [43]. This study was performed on five new transformers of the same rating and oil type before and after the test. For a 13 years in-service transformer, DGA has been performed. These studying cases were to confirm the results obtained. For a new five cases, a study of these relationships reveals that gases are generated H 2 , CO, CO 2 , CH 4 , C 2 H 6 , C 2 H 4 , and C 2 H 2 with an increase in temperature. Oil samples are analyzed in the Central Chemical Laboratories of the Egyptian Ministry of Electricity. The oil samples were taken by manual collection according to IEEE Std. C57.104 and IEC 60567 [9,44]. These samples were taken from the located point in the transformer tank, where the oil is circulating; this is to increase the chance to catch gases generated in another location in the transformer than the sampling point. The oil syringe used for oil sample transporting to the Chemical laboratories from the field is in the 50 ml range. For taking the oil sample, wipe the clean sampling valve. then, flush valve and rinse 2 to 3 times to remove water and debris, and sufficient amount sample fills without causing aeration/bubble and correct level (small headspace -95%) and records sampling oil temperature close/seal and store for transport. Fig. 10 shows the position and use tools for taking the oil samples. For each case of a new five transformers, two samples of oil are analyzed using DGA, one of these samples before and the other after the temperature rise test. The first sample for each case was taken two hours before the temperature rise test, while the second sample was taken after 24 hours of the temperature rise test, as recommended in standard (IEC 60076-2). The results of dissolved gases according to the central laboratories for a new five cases of transformers before and after the temperature rise and the aged case are as presented in Table X. According to the laboratory result, the diagnosis of new transformer cases before and after temperature rise is normal because the gases concentration is not exceeding the limits according to IEEE Std. C57.104 [9]. While by  Resistance (mΩ)

C. ACIDITY AND MOISTURE MEASUREMENT
By the chemical reactions in the oil, Carboxylic acids are produced. Increased acidity of the oil affects the electrical properties of the oil, it may also cause abnormal deterioration of paper used for insulation in the transformer windings [44,45]. As well as the moisture presence can deteriorate transformer oil. Water content can cause the bubble formation risk in transformer oil when water desorption increases the local concentration of gases in the transformer oil. Hence, this section presents acidity and moisture measurements for oils extracted from the new five transformers (66/11kV) of oil type (Nytro 10 XN). These were collected from ELMACO Company -Egypt during their first-year in-service. Another sample extracted from a 13-year in-service transformer was collected and tested. Using volumetric syringes, the samples were taken, and their moisture and acidity were measured in the central laboratory of the ministry of electricity-Egypt. Table XI and Fig. 12 show the measured acidity and moisture in mg KOH/g and (ppm), respectively for a new five cases and aged case study. The acidity measurement showed that mg KOH/g was between 0.01 and 0.06 during the first year of transformer in-service which is less than the maximum limit [45][46][47]. In addition, the moisture content for the same new five transformers were 7.5, 6.4, 7.1, 6.8, and 6.2 (ppm) which were not exceeded the limit value [48]. For the 13 years in-service transformer, the acidity, and moisture were 0.19 (KOH/g) and 12.5 (ppm), respectively. These values exceeded the maximum limits due to the diagnosed thermal fault, as shown in Table XI and Fig. 12. These acidity and moisture measurements can be used to evaluate the health index of the transformer as a diagnosis factor.

D. HEALTHY INDEX OF TRANSFORMER INSULATION SYSTEM BASED
The transformer Health Indices (HI) represent a practical method to quantify the results of operating observations, field inspections, and site and laboratory testing into an objective and quantitative index, providing the overall health of the assets. This section presented two techniques of predictive maintenance and a healthy index of transformer oil. These techniques were by using DGA only and by using DGA with the BDV and typical chemical properties. The healthy index of transformer oil can be predicted using the individual DGA approach before and after the temperature rise test. By analyzing DGA results for five cases according to IEC 60076-2 [43], the time interval (t) during which the top-oil temperature rise is at least 80% of the final value in the steady-state condition is 3 hours. According to this time interval, the gases rate increases (GRI) can be calculated for each gas according to the below equation (16): Where 1 represent the gas concentration before the temperature rise, as well as 2 represent the concentration of gases after rising the temperature. After calculating GRI for each gas separately for gases (i = 1, 2, 3 …7), GRI's values for each case study are compared with Admissible limits for gas rate increases in standard IEC 60076-2 [43], see Table XII.  4) and (5), the summation of (GRI)'s of the main five gases H₂, CH₄, C₂H₆, C₂H₄, and C₂H₂ exceeded the limit of the first series of the standard [43]. Hence, for the three cases, the possibility exists of local overheating. Consequently, it is recommended that in service, oil   (2), the values of gases concentration reduced after the temperature rise test. This may be because of the exposure of this sample to atmospheric leaks. In general, most cases predicted the possibility exists of local overheating. These results proved when reviewing the transformers of the same rating and oil type but in the service that a healthy index of transformer oil using DGA after temperature rise gave unrealistic results.

2) HEALTHY INDEX OF A NEW AND AGING TRANSFORMERS INSULATION SYSTEM BASED ON BDV, ACIDITY, MOISTURE, AND DGA
By using the results of breakdown, DGA, moisture content, and acidity, the healthy index was calculated for the new five transformer cases and the aged one. The description factor for breakdown, DGA, moisture, and acidity can be computed according to the following equation (17) [25].
Description factor (DF) = Where the score (S i ) is an assessment of the testing results of each parameter. According to the standards of the testing for these parameters, it can be represented in the form of numbers with ranges. The weight of a parameter (W i ) is a factor of the test parameter influencing a transformer; it can take the integer value for 1 to 5 [47]. According to Table  XIII, the score (Sᵢ) and weight factor (Wᵢ) of each gas after the temperature rise test can be determined for new cases and for an aged one. Hence, the DGA description factor was calculated. As well, calculation the description factor of breakdown, moisture content, and acidity was according to Table XIV. By calculating the description factor for DGA, acidity, moisture, and dielectric strength according to their scoring and weight factor, the rating code and condition can be known according to Table XV. For the new transformer study cases, the average DGA Description factor for the new five cases after the temperature rise test can be calculated according to equation (17). Therefore, the (a) (b)   rating code was determined. Table XVI shows that for (66/11 kV) transformer of NYTRO 10XN oil type, the DF for DGA were 1, 1.6, 1, 1.11, and 1 for a new five cases, respectively, of rating code (A), and DF for dielectric strength was 1 and the rating code was (A) and also for all moisture and acidity measurements, except acidity of case No. 4, its rating code was (C). On the other hand, using equation (17), the DF for the 13 years in-service transformer based on DGA, dielectric strength, acidity, moisture content was 2.88, 2, 3, and 1 of rating code D, D, E, and A, respectively, as showing in Table XVI. The rating (A, B, C, D, E) is converted to a factor between 4 and 0, respectively, called the health index factor (HIF) in Table XVII. The health index percentage was calculated according to the following equation (18).
By using the DGA, breakdown voltage test, moisture content, and acidity measurements of the oil, HI % for new cases was approximately 100%. According to Table XVIII [25], the condition was very good; the expected lifetime was more than 15 years, and normal maintenance requirement. For the aged transformer case with 13 years in-service, HI % was approximately 36%. Hence, the aged transformer lifetime is 3 years with poor condition. With studying the state of transformers in the service, the health index results using DGA, breakdown, and the measurement of the chemical property was closer to reality compared to using DGA only.

IV. DISCUSSIONS
In this research, the discussions can be summarized as: -For improving the diagnostic accuracy of transformer faults using the DGA approach, a proposed technique was presented. It is based on the integration of five of the previous DGA methods but with novelty procedures. The proposed technique proved its high accuracy in faults diagnosis. The results of this part are based on a number of datasets that are collected from the Egyptian Electricity Transmission Company (EETC) and literature datasets. The diagnostic accuracy using the proposed  techniques was improved compared to some of the previous methods. -For studying a transformer of rating (66/11 kV). A healthy index of its insulation system was evaluated using firstly DGA alone then DGA plus breakdown voltage test, moisture content, and acidity measurement. For this objective, Firstly, the breakdown test was carried out on study case oil type and another oil type wants to use it as well as a mixture of them by 50/50%. Mixing 50% of the two oil types; NYTRO 10 XN and HyVolt III, was slightly higher than pure HyVolt III. But the type of NYTRO 10 XN oil remains the best in terms of dielectric strength. -The aim of this section was to clarify the effect of mixing the two types of oils on the dielectric strength to take advantage of that in advance in the absence of one of them, which has not been reported in the literature. In addition, this average breakdown voltage was used to measure the health index of oil quality of the case study. A secondary, the dissolved gases in oil before and after temperature rise test were analyzed and compared with each other. The objective of the temperature rise test was to determine the oil health index and the predicted fault based on the rate of change in dissolved gases concentrations before and after the test. A thirdly, moisture and acidity were measured for five new cases and aged transformers of the same rating, the acidity results were ranged between 0.01 and 0.06 mg KOH/g, and moisture measurements were not exceeded 10 ppm. Accordingly, a healthy index was measured using DGA alone, then using DGA plus breakdown voltage, acidity, and water content. The dissolved gas analysis before and   after the temperature rise test predicted the possibility exists of local overheating. While according to the breakdown and DGA results plus chemical properties, the health index was calculated and resulted that the insulation condition was good, expected lifetime was more than 15 years for the new transformers with normal maintenance requirements. For the aged transformer case with 13 years in-service, the health index results that the aged transformer lifetime is 3 years with poor condition.
With studying the state of transformers in the service, the health index results using DGA, breakdown, and the measurement of the chemical property was closer to reality compared to using DGA only.

V. CONCLUSIONS
A proposed combined DGA technique has been presented based on the integration between results of existing techniques used in literature with higher accuracies. The new DGA technique has accuracy higher than the recently published techniques as well as the techniques used in constructing the proposed technique. Where, the diagnosis accuracy results of the proposed technique are 93.3% for PD diagnosis, 86.3 % for D1, 98.2 % for D2, 96 % for T1, 83.7% for T2, and 98.2 % for T2. Hence, the overall accuracy of the proposed DGA technique improved significantly, reaching 93.6%. While the techniques used in constructing this proposed one are 90%, 53.6%, 47.2%, 59.1%, and 70.8%, respectively. In addition, the recently published four techniques are 63.6%, 89.7%, 76.9%, and 85.8%, respectively. This means that the newly proposed one presents higher accuracy of more than 90%. In addition, new validation datasets (36 samples, 10% from all datasets) have been used for checking the proposed technique accuracy, which found 91.5 %. This is a higher value than the construction techniques and the previously published techniques.
For transformer study case (11/66) kV, NYTRO 10XN, the breakdown voltage test for transformer oil concluded that; by mixing 50% for two oil types as NYTRO 10 XN and HyVolt III, the dielectric strength was slightly improved compared to HyVolt III. But the type of NYTRO 10 XN oil remains the best in terms of dielectric strength. The study case oil type (NYTRO 10 XN) remains the best in terms of dielectric strength. In addition, the temperature rise test was carried out on the case study; winding dc resistance was measured; in addition, the dissolved gases before and after the test were analyzed and compared. The most effective gases for the temperature rise test were (CO 2 , CO, and H₂). Using DGA, acidity, and BDV results, the health index of the transformer insulation system was evaluated using DGA only, DGA plus BDV moisture and acidity. By using DGA before and after the temperature rise test, it can expect the probability exists of local overheating. Though according to the breakdown, DGA, moisture content acidity results, the health index was computed and followed that the insulation condition was superior in the prediction process and its expected lifetime was more than 15 years with normal maintenance needs for a new case. The health index for the aged transformer case was that the lifetime is 3 years with poor condition. Hence, the health index of the transformer oil quality using more than one diagnosis factor is more accurate than using one factor.