Keywords

1 Introduction

The power transformer is the most valuable equipment in the power generation, transmission, and distribution sector [1]. Nowadays, the demand for electricity in commercial, domestic, industrial, and agricultural sectors is increasing rapidly, the consequence of which is surge in transformer, which increases chances of fault occurrence. So for reliable operation of the transformer, it has become mandatory to continuously monitor and diagnose the insulation condition so that it can detect the incipient fault if any [2,3,4,5,6,7,8]. Dissolved gas analysis (DGA) is the most prominent technique used for detecting the incipient fault occurring in the transformer.

As the temperature in the transformer reaches beyond a particular limit in the presence of oxygen (O2) and moisture, thermal and electrical stresses are introduced degrading the physical, chemical, and electrical properties of oil. This oil degradation is the major cause of transformer failure [9, 10]. Due to these stresses, the carbon chain of oil breaks, thereby introducing hydrocarbon gases in oil as by-products, namely methane (CH4), hydrogen(H2), ethane(C2H6), ethylene(C2H4), acetylene(C2H2), and degradation of insulation paper introduces carbon monoxide(CO) and carbon dioxide (CO2) in oil. These gases introduce further electrical and thermal faults in transformer which are categorized as per IEC 60599 and IEEE C.57.104 standards as shown in Fig. 1. The electrical faults are categorized into partial discharge (PD), discharge of low energy (D1), and discharge of high energy (D2), whereas thermal faults are categorized based on the temperature range as thermal faults of temperature T < 300 °C (T1), 300 < T< 700 °C (T2), and T > 700 °C (T3) [11].

Fig. 1
figure 1

Categorization of fault as per IEC 60599 and IEEE C.57.104 standards

Full size image

There are several methods for detection of incipient faults in power transformer such as ratio method, key gas method, Duval triangle, artificial intelligence-based methods. [12]. In all these methods, Duval triangle came out to be the most accurate method but the problem with it is, it also does not yield 100% accuracy.

From this point of view, a new technique is required, which can detect faults with better performance than Duval triangle. So a new DGA interpretation scheme has been employed in this paper. The new technique is based on the certain rules programmed in MATLAB which can directly provide the fault present in the transformer. This paper evaluates and compares the results obtained from Duval triangle and the new DGA interpretation scheme.

2 Basic Knowledge

2.1 Graphical Method: Duval Triangle

The Duval triangle fault diagnostic method is a graphical approach first introduced by Michel Duval in 1974 [13]. This method uses a different set of hydrocarbon gases, placed on the three sides of an equilateral triangle representing relative proportion of gases [14]. Till date, a total of seven Duval triangles have been introduced by Duval on the basis of oil used in the equipment. The triangles 1, 2, 3 uses CH4, C2H4, and C2H2 gases, and triangles 4 and 6 use H2, C2H6, and CH4 gases, whereas triangles 5 and 7 uses CH4, C2H6, C2H4 gases [15]. In this paper, we are concerned with triangle 1 of mineral oil to study the type of fault present in the transformer.

The triangle is subdivided into seven zones, each zone signifying a type of fault as mentioned in Fig. 1 with one intermediate fault DT signifying mixture of electrical and thermal fault in the power transformer. The Duval triangle is shown in Fig. 2.

Fig. 2
figure 2

Transformer fault diagnosis using Duval triangle

Full size image

The relative percentage of each gas is given by

$$\% {\text{CH}}_{4} = \frac{{100({\text{CH}}_{4} )}}{{{\text{CH}}_{4} + {\text{C}}_{2} {\text{H}}_{4} + {\text{C}}_{ 2} {\text{H}}_{2} }}$$
(1)
$$\% {\text{C}}_{ 2} {\text{H}}_{4} = \frac{{ 1 0 0 ( {\text{C}}_{ 2} {\text{H}}_{ 4} )}}{{{\text{CH}}_{ 4} + {\text{C}}_{ 2} {\text{H}}_{ 4} + {\text{C}}_{ 2} {\text{H}}_{2} }}$$
(2)
$$\% {\text{C}}_{ 2} {\text{H}}_{2} = \frac{{100({\text{C}}_{ 2} {\text{H}}_{2} )}}{{{\text{CH}}_{4} + {\text{C}}_{ 2} {\text{H}}_{4} + {\text{C}}_{ 2} {\text{H}}_{2} }}$$
(3)

where CH4, C2H4, C2H2 concentrations all are in ppm.

Plotting the relative percentage of each gas in the triangle provides only one point in the triangle. The location of point at a particular zone decides the type of fault present in the transformer.

2.2 New DGA Interpretation Technique

The techniques use relative proportion of four combustible gases (%H2, %C2H2, %CH4, %C2H4) out of seven gases present in the oil sample to detect the fault type [16]. For this, certain rules are employed derived from several case studies. Table 1 shows the rules employed in the technique, and Table 2 shows the ratios associated with the rules. The percentage relative proportion of CH4, C2H2, C2H4 is as shown in Eqs. (1)–(3), respectively, and the percentage relative proportion of H2 gas is given by

Table 1 Rules employed
Full size table
Table 2 Ratios involved
Full size table
$$\% {\text{H}}_{2} = \frac{{100({\text{H}}_{2} )}}{{{\text{H}}_{2} + {\text{CO}} + {\text{CO}}_{2} + {\text{C}}_{ 2} {\text{H}}_{6} }}$$
(4)

where H2, C2H6, CO, CO2 concentrations all are in ppm.

The faults in this technique are categorized mainly into five types usually found in transformers in service as partial discharge (PD), discharge of low energy (D1), discharge of high energy (D2), thermal faults of T < 700 °C (T1/T2), and thermal faults of T > 700 °C (T3) [17].

Initially, step 1 is used to detect the PD fault type. If rule 1 is true, the fault is PD. If not, step 2 is employed in which rule 2 is checked for identifying whether fault is electrical fault (D1 or D2) or thermal fault ((T1/T2), T3). If rule 2 is true implies the fault is electrical fault type; further in step 3, rule 3 is checked. If rule 3 is true, the fault is D1 else D2. Also, if rule 2 is false, the fault lies in the thermal fault category; further in step 4, rule 4 is checked. If rule 4 is true, the fault is T3 else T1 or T2. Figure 3 shows the flowchart associated with the technique [16].

Fig. 3
figure 3

Flowchart for fault diagnosis [16]

Full size image

The fault diagnosis using the proposed technique is carried out by developing and running a MATLAB program based on the rules and the flowchart as shown in Table 1 and Fig. 3, respectively.

3 Results and Discussion

The proposed new technique is compared with the conventional Duval triangle, the most efficient technique till date. A total of 109 DGA samples have been analyzed to predict and compare the percentage accuracy of the two techniques. Out of 109 DGA samples, 29 samples and 10 samples are collected from reference [4] and reference [18], respectively. Remaining 70 samples has been collected from TIFAC-Core Laboratory, NIT Hamirpur.

The 29 DGA samples from reference [4] include several different types of oil-filled electrical equipment such as power transformers with or without on load tap changers, instrument transformer, bushing, reactor, and cable. The other 70 DGA samples are of various oil-filled working transformers currently operating in Himachal Pradesh, India.

For the new technique, a total of 7 DGA samples became a challenging task due to unavailability of one or more gases introducing (0/0) or (a/0) condition. For this, the ratio (0/0) is set to 0 and ratio (a/0) where ‘a’ is not zero is set to 25 in the MATLAB programming.

Table 3 shows some of the DGA samples and the fault obtained by the two techniques. Table 4 shows the number of correct diagnoses, wrong diagnosis, and the percentage accuracy of the two techniques for all the DGA samples collected.

Table 3 Some DGA samples and their fault diagnosis using (a) Duval triangle and (b) new proposed method [4, 18]
Full size table
Table 4 Diagnosis result (in percentage)
Full size table

From Table 4, it is clear that the proposed new technique has better performance than the conventional Duval triangle with 88.99% accuracy and the method can be universally accepted as a new technique for fault diagnosis.

4 Conclusion

The new proposed method uses gas ratios and proportion as their main tool to detect the type of fault in the oil-filled transformers. The method is simple and easy to use. Several steps and the rules associated with the method can easily be programmed in MATLAB for fault diagnosis. The method calculates higher accuracy compared to conventional Duval triangle which is considered as the most accurate method till date. The reason for higher accuracy is that the proposed method uses all the seven gases concentration for fault diagnosis whereas the Duval triangle uses only three gases.