1998

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

Reliability Assessment of Insulation System for Dry Type Transformers Minmin Wen, Jiancheng Song, Yuan Song, Yuan Liu, Chuanyang Li and Peng Wang Shanxi Key Laboratory of Coal Mining Equipment and Safety Control Taiyuan University of Technology, No.79 Yingze West Street, Taiyuan, 030024, China

ABSTRACT The insulation system of a dry-type transformer under the combined thermal, electrical, mechanical, and environmental stresses undergoes gradual deterioration, which ultimately leads to transformer failure. It is hence important to test specimen insulation of transformer coil so as to explore the aging law and the associated characteristic parameters related to the insulation condition. This paper proposes a method to evaluate reliability of the insulation system of a dry-type transformer based on the three models of degradation, regression and expert judgment. The three models are developed based on the quality parameters (QPs) from the lab tests. The method is illustrated by estimating the reliability of the insulation system of a mining flameproof dry-type transformer. The accelerated aging tests on the Nomex paper (i.e. the turn-to-turn insulation of the transformer) are conducted. The QPs of the Nomex samples of degree of polymerization, partial discharges, and photomicrograph were measured periodically during aging test. The effectiveness of the proposed method was proved through the lab experiments. The results show that the parameters from the aging test can be used to represent the condition of the insulation system. Index Terms — dry-type transformer insulation, reliability assessment, degree of polymerization, partial discharge, photomicrograph.

1 INTRODUCTION TRANSFORMERs play an important role in the electrical power transmission and distribution networks. The first transformer in the world invented in the 1880s was of dry type [1]. Oil as insulation material of transformers was started by Elihu Thomson and the first commercial oil-immersed transformer was introduced by Westinghouse in 1886. Compared with dry-type transformers, the oil-immersed transformers are not suitable in some particular places for its relatively poor fireproof performance and ability of environmental protection. Therefore, since 1990s, oil-immersed transformers have been replaced by dry-type transformers in many industrial and commercial installations. Recently, the failures caused by the decline of the insulation properties of dry-type transformers are frequently reported due to non-sinusoidal loads. The lifetime of a dry-type transformer mainly depends on the condition of insulation material. The insulation system under combined thermal, electrical, mechanical, and environmental stresses undergoes gradual deterioration, which may lead to serious accidents and great losses for utilities and customers. The insulation system of a dry-type transformer mainly refers to the winding insulation. The rated voltage and loading of dryManuscript received on 30 January 2013, in final form 12 June 2013.

type transformers used in mining industry, China have been increased in recent years. Therefore winding insulation failure due to high voltage and loading operation might become the dominant cause of transformer failure in future [2-3]. It is essential to develop the reliability assessment methods to estimate the operating condition of the insulation system and to predict earlier failures of the transformer. Considerable research works [4-6] on aging evaluation of the insulation of dry-type transformers have provided theoretical basis for further research. The influence of temperature on insulation properties of dry-type transformers has been mainly studied [7-8]. This paper proposes a method to evaluate reliability of the insulation system of dry-type transformers based on the three models of degradation, regression and expert judgment. The three models are developed based on QPs. The method is illustrated by estimating the insulation system reliability of a mining flameproof dry-type transformer. In particular, its turnto-turn insulation, i.e. Nomex paper, is taken as a component example to illustrate the method. The accelerated aging tests on the Nomex paper specimens are conducted. The QPs of the Nomex paper such as degree of polymerization (DP), partial discharges (PDs) and photomicrograph (PG) were measured periodically during the aging test. The parameters obtained from the aging test are analyzed to represent the condition of insulation specimens. The effectiveness of the proposed method was proved through experiments.

1070-9878/13/$25.00 © 2013 IEEE

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

1999

2 RELIABILITY MODELING 2.1 QUALITY PARAMETERS The reliability of the insulation system of a dry-type transformer is evaluated based on operating conditions of system components represented by QPs. In [9], the QPs are classified into the directly linked measurable QP (D-M-QP) which is measurable and directly related to a degradation process, the indirectly linked measurable QP (I-M-QP) which is externally measurable and indirectly related to the degradation process, and the indirectly linked comparative QP (I-C-QP) which is not measurable and is indirectly related to the degradation process. The QPs are divided into two categories in this paper, which are measurable QPs and un-measurable QPs. Both the D-M-QP and I-M-QP belong to measurable QPs. Un-measurable QPs include the I-C-QP, the observable information (OI) from measure instruments and artificial expertise (AE) from the historical data of transformers and the experience of experts, etc. The QPs are utilized to establish the reliability model of the insulation system of dry-type transformers in next section. 2.2 MODELING Degradation model (DM), regression model (RM) and expert judgment model (EJM) are proposed to estimate the component reliability of the insulation system based on QPs. The procedure of component reliability assessment based on the measurable and un-measurable QPs is shown in Figure 1. The turn-to-turn insulation material (Nomex paper) of a mining flameproof drytype transformer is used as an example.

Figure 2. The condition change process due to degradation.

The decline of DP is a result of the scission of the polymeric chains caused by cascade chemical reaction. The relationship between temperature and DP is given by Arrhenius through the reaction rate k [11-13]. DP as the function of time t is given as DP (t )

DP (t 0 ) t

1 DP (t 0 ) k ( )d t0

where DP(t) is the DP at time t, DP(t0) is the DP at an initial time t0 and k(t) is reaction rate at time t, which can be obtained from Arrhenius equation: k (t ) A exp( E RT (t ))

C ' C ( HS TA Tk ) ( C TAT Tk )N

The D-M-QP include the degree of polymerization (DP) which indicates the degradation of the Nomex paper; The I-M-QP include the partial discharge (PD) which does not directly contribute to a failure mechanism in [10]; OI include the photomicrograph (PG) of the Nomex paper; The AE is based on the historical failure records from the condition monitoring of the mining flameproof dry-type transformers and the experience of experts. 2.2.1 DEGRADATION MODEL The degradation model as shown in Figure 2 is proposed to track the degradation process of the insulation. The DP is changing with the insulation degradation. The degradation can be determined by the decline of the DP.

(2)

where A is a frequency factor which depends on water content, oxygen or acidity, E is the molar activation energy of the degradation reactions (hydrolysis and thermo-oxidation), R is the universal gas constant, and T(t) is the temperature at time t during the reaction in Kelvin calculated by equation (3) and equation (4) in IEEE Std C57.134-2000 [14]. It is important to note that T is the experiment temperature in thermogravimetric analysis (TGA).

HS C ' 1 ( E C ' )

Figure 1. Flow chart of the reliability estimation method.

(1)

1

N

(3)

N

(4)

where C is the temperature rise of hottest-spot during current-only test, C’ is corrected winding hottest-spot temperature rise due to current-only test, E is winding hottest-spot temperature rise during excitation voltage-only test, HS is winding hottest-spot temperature rise for rated conditions, TA is rated ambient temperature which is usually 30℃, TAT is actual ambient temperature during test, Tk is 234.5 ℃ for copper and 225 ℃ for aluminum, N is 0.8 for self cooled and 0.9 for forced air. The reliability and failure probability due to DP is described in [3]. The failure probability is the probability that the DP-value was lower than its threshold value when insulation fails. It involves two probability distributions. The first is the probability distribution of DP-value at time t. The probability of DP with value between x and x+dx at time t is denoted as pdp(x,t)dx. And the second distribution is DPvalue of which below the threshold value. The probability of the threshold having a value between x and x+dx is denoted as pth ( x)dx . The probability that the threshold being above a certain DP-value x is given by equation (5) and the failure probability at time t is calculated by equation (6) [3].

Pth ( x) pth ( x' )dx' x

(5)

2000

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

Pf (t ) Pth ( x) pdp ( x, t )dx

(6)

0

The time dependent reliability based on DP can be calculated using equation (7). Rdp (t ) 1 Pf (t )

(7)

In this paper, the threshold value of DP is determined to be 30% of the initial value of DP. The degradation of the Nomex paper follows the reaction kinetics equation (8).

d dt k (t )(1 ) n

(8)

where is the reaction extent which is given in equation (9), n is the reaction order. w0 wt w0 w

(9)

where w0 is the initial mass of the sample in the TGA measurement, wt is the mass at time t, w is the mass at the end of the measurement. The parameters A and E are determined using the following procedure. It is assumed that the heating rate of the TGA measurement is

, dT (t ) dt . Combining (2) and (8) gives d (1 ) n A exp( E RT (t )) dT (t )

Integrating

(10)

0 (1 )

n

d

exp( E T

RT ) dT

T

The inputs of RBF can be the features of PD signals such as statistical characteristics from wavelet packet transformation [16] and time domain characteristics, since the pattern of PD changes with insulation degradation. In this paper, six statistical characteristics of PDs are extracted and serve as the inputs of the RBF, i.e. +Skewness, +Kurtosis, -Skewness, -Kurtosis, Crosscorrelation and Discharge factor [17]. In order to improve the accuracy, the least mean square (LMS) algorithm was used for updating the network parameters (centers, widths, and weights of Gaussian neurons). The output of the RBF is the component reliability Rr. 2.2.3 EXPERT JUDGMENT MODEL The expert judgment model is based on the AE and OI. Component reliability is determined by the experts through grading the component according to the experience and the photomicrographs of the insulation. Assuming that n experts participate in judgment, component reliability Re can be calculated as

Re the

A

2.2.2 REGRESSION MODEL The radial basis function neural network (RBF) [15] with three layers is utilized as the regression model to evaluate reliability.

(10)

with

0 and T (0) T0 , gives

The A and E can be obtained from (15) through functionfitting for given the heating rate and the reaction order n. The reaction order n of Nomex is about 1.

initial

conditions

1 (r1 r2 ... rn ) n

where ri is the reliability judgment by the ith expert. Table 1. Relationship between reliability and grades

(11)

0

Defining p( x)

x

e x dx and x E RT , equation (11) can x2

be represented by

0

(1 ) n d

AE p( x) R

(12)

The approximate expression of Doyle approximate function p(x) is as ln p ( x ) 5.331 1.052 x

(13)

Substituting equation (13) and x E RT into equation (12), gives ln[

0

AE E (1 ) d ] ln( ) 5.331 1.052 R RT n

(14)

Equation (14) also can be transformed into equation (15). lg lg

AE

R (1 ) d 0

n

2.315 0.4567

E RT

(15)

Reliability (Rr)

grade

98%＜ Re≤100%

normal state

95%＜ Re ≤98%

attention state

85%＜Re≤95%

abnormal state

0＜ Re≤85%

severe state

The lifespan of the insulation is divided into four states corresponding to the insulation grades of normal, attention, abnormal and severe state. The relationship between the reliability and states are shown in Table 1. 2.2.4 COMPONENT RELIABILITY The weighted mean method is utilized in this paper to calculate the reliability of component i of a insulation system as n

(16)

Ri (t ) w j R j (t )

(17)

1

where Rj is the reliability obtained from the jth model of the component with weight wj. The weights are determined according to the degree of correlation between insulation model and aging process. The accuracy of the model is also considered.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

2.2.5 INSULATION SYSTEM RELIABILITY The insulation system of a mining flameproof dry-type transformer can be divided into four components of turn-toturn insulation, layer-to-layer insulation, windings-to-ground insulation and winding-to-winding insulation. The main insulation materials of the four components are Nomex paper, polyimide film, combination of Nomex paper and air, panel made of imide and glass fiber, respectively. From the insulation structure characteristics of mining flameproof drytype transformers, it is found that the impacts of components on insulation system reliability are independent. The insulation system cannot operate normally if any of the components fails. Therefore, the reliability of the insulation system is modeled as a series system as n

Rs (t ) Ri (t )

(18)

1

where Rs is the reliability of the insulation system, Ri is the reliability of the ith component of the insulation system.

3 MEASUREMENTS AND RESULTS The turn-to-turn insulation specimens of a mining flameproof dry-type transformer are tested in this section. The QPs were obtained from the experiments using the proposed method. 3.1 ACCELERATED AGING TEST 3.1.1 INSULATION MODEL

(a)Copper electrode (b)Turn-to-turn insulation model Figure 3. Copper electrode and turn-to-turn insulation model.

The model shown in Figure 3 is designed according to the characteristics of turn-to-turn insulation. The model includes two copper electrodes of the high-field electrode and the lowfield electrode. The Nomex paper from the DuPont Corporation of America is cut into strips and fixed between the two electrodes to simulate the turn-to-turn insulation. There are three layers of paper strips (one group) and each layer is 0.8mm thickness in the model. 3.1.2 ELECTRICAL-THERMAL AGING TEST The aging tests were performed according to the IEEE Standard C57.12.56 [18]. Before the aging test, the specimens of Nomex paper were dried under 90 ℃ for 12 hours. The specimens were tested under the voltage and temperature used in [19] and [13] respectively. 1.6 kV AC, which is about 1.6 times of the inception voltage of partial discharge for the specimens, is applied to the specimens. The aging temperature is kept at 130℃. The specimens were aged for 10 hours per day. The related QPs were measured each week during the aging process periodically.

2001

At least 5 groups of specimens are taken out per week for the measurement so as to guarantee the accuracy of the QPs.

(a) (b) Figure 4. Macroscopic photos of the specimens before and after the aging test

The two groups of aged specimens are shown in Figure 4a and 4b respectively. The left specimen in Figure 4a and 4b is intact specimen. The two groups of specimens (the right three specimens in Figure 4a and 4b) are specimens aged for 1 week and 4 weeks respectively. It can be found that the specimen color is becoming more and more yellow with increase of the aging time. 3.2 QUALITY PARAMETERS 3.2.1 DP MEASUREMENT DP of the Nomex paper is generally measured indirectly by viscometry. The viscosity of the Nomex in special solvent is measured in this paper. It is usually very difficult to dissolve aromatic polyamides with traditional solvents. The literature [20] indicate that aromatic polyamides can be dissolved in polar solvents, such as N,N-dimethylacetamide (often only in the presence of lithium salt), or strong acid, such as concentrated sulfuric acid or chlorosulphonic acid. Therefore the sulfuric acid is selected as the solvent of the Nomex paper specimens. The measurement system is shown in Figure 5. The viscosity of the Nomex was measured in 96% sulfuric acid. The initial concentration of the dilute solution made of the Nomex and sulfuric acid was 0.005 g/ml. In order to ensure the accuracy of measurement, the concentration of the sulfuric acid solution was controlled within 95%~97%. At the same time, the temperature was maintained within 30 0.3℃. The efflux time of the solution was measured by high precision timer of 1/100 second. The intrinsic viscosities were calculated by extrapolating the reduced viscosity measured by the capillary viscosimeter (Ostwald viscosimeter,=0.8 mm) until the concentration is zero.

Figure 5. The viscosity measurement device

2002

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

The measurements of the aged specimens were performed per week. Five groups of specimens are measured simultaneously to calculate the average intrinsic viscosity. The relationship between DP-value and intrinsic viscosity is derived as follow: Mark-Houwink-Sakurada [19] defined the relation between the intrinsic viscosity [ and the absolute weight-average molar mass M as

[ ] aM b

(19)

signals in specimens were measured during different aging periods in this paper. Then the PD patterns were established according to the insulation conditions of specimens. The PD measurement system was designed according to the IEC 60270 standard and IEEE Standard C57.124 [22], which consists of high voltage generator, capacitive voltage divider and coupling device, and the experiments were conducted in the electromagnetic shielded room as shown in Figure 6.

where a and b are constant parameters of the polymer/solvent system. The intrinsic viscosity is calculated by extrapolating the measured reduced viscosity.

[] limsp c limlnr c 100liminh c0

c0

c0

(20)

where c is the concentration of the measured solution, is the relative viscosity of the measured solution, inh is the inherent viscosity of the measured solution, sp is the specific viscosity of the measured solution. The equations of Huggins (21) and Kraemer (22) are developed from equation (20).

sp c [ ] [ ]2 c

(21)

ln r c [ ] [ ] 2 c

(22)

Figure 6. The system for aging and PD tests

3.2.3 PHOTOMICROGRAPHS OF NOMEX PAPERS

where and are constants. The relationship between the absolute weight-average molar mass and the viscometric DP-value P is depicted by M P M0

(23) Figure 7. Leica DM2500M microscope.

where M0 is molecular weight of the monomeric unit. Combining equation (23) with (19), gives [ ] a ( PM 0 ) b

(24)

Taking logarithm on both sides of equation (24), gives

ln P

1 1 ln[ ] ln M 0 ln a b b

(25)

1 1 Introducing b and a ' (ln M 0 ln a ) , gives b b '

ln P b ' ln[ ] a ' (26) Equation (26) shows the linear relationship between ln[] and lnP. As a result, P can be replaced by [] to characterize aging status of the insulation and the same replacement should be done in the equation (1), (5) and (6). 3.2.2 PD MEASUREMENT Partial discharges (PD) might happen when there are defects in the insulation of electrical equipment. The PD signals can be used to identify the nature of failure and aging status of the insulation system, which should make a breakdown distribution on fault diagnosis and failure warning. The PD

It has been verified that the local photomicrographs of the insulation material can reflect its aging degree in terms of the microscopic characteristics. Leica DM2500M microscope was used to observe the microscopic structure of the aged specimens as shown in Figure 7. 3.3 MEASURED RESULTS 3.3.1 DP MEASUREMENTS The DP was measured per week during the aging test. The viscosity data of the intact samples and the aged samples of 28 days are shown in Table 2 and Table 3 respectively. Table 2. Viscosity of the intact sample. No.

0

1

2

3

4

5

c t0 t r sp sp/c lnr/c

/ 965 / / / / /

0.005 / 1904 1.973 0.973 194.6 135.91

0.0025 / 1409 1.460 0.460 184.0 151.38

0.00125 / 1172 1.215 0.215 172.0 155.80

0.000625

0.0003725 / 1016 1.053 0.053 169.6 165.26

1070 1.109 0.109 174.4 165.53

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

Table 3. Viscosity of the aged sample of 28 days No.

0

1

2

3

4

5

c t0 t r sp sp/c lnr/c

/ 965 / / / / /

0.005 / 1703 1.765 0.765 153.0 113.60

0.0025 / 1307 1.354 0.354 141.8 121.35

0.00125 / 1119 1.160 0.160 127.7 118.45

0.000625

0.0003725 / 984 1.020 0.020 63.0 62.39

1030 1.067 0.067 107.8 104.30

where t0 is the efflux time of the solvent, t is the efflux time of the solution. The intrinsic viscosities of the Nomex/sulfuric acid solutions in different aging periods were calculated and the results are shown in Table 4.

(9b) Partial discharge in the second aging period

Table 4. Viscosity parameters in different aging periods Time(day)

0

7

14

21

28

[ ln[

167.4 5.12

145.5 4.98

113.5 4.73

82.2 4.41

62.7 4.14

(9c) Partial discharge in the third aging period

Figure 8. Viscosity parameters versus aging time.

Table 4 shows that the degradation of the solid insulation is of accumulative effect, and the intrinsic viscosity [] and ln[] declines with aging time as depicted in Figure 8. Generally, the inherent viscosities of Nomex paper samples varied in the range of 0.5 to 8 dl/g. 3.3.2 PD MEASUREMENTS The PD signals for different aging periods as shown in Figure 9 are used to train RBF model. Figure 9e shows the PD when the insulation is close to breakdown. Figures 9a-9d clearly indicate the PD Patterns during degradation process. Those PD patterns are utilized to estimate the reliability of the turn-to-turn insulation.

(9d) Partial discharge in the fourth aging period

(9e) Partial discharge when the insulation is close to breakdown (9a) Partial discharge in the first aging period

Figure 9. Partial discharges in different aging period.

2003

2004

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

3.3.3 PHOTOMICROGRAPHS OF THE NOMEX PAPER

(a)

In this paper, the exponential model by Fallou [26] shown in equation (27) is chose to describe the relationship between the time to failure and the applied stresses on the insulation since it is much more suitable for the thin insulation materials and also considers the influence of the electric and thermal stress together. The expected life of the insulation system and its components can be extrapolated from equation (27).

(b)

t exp( A1 A2U

(c)

(d)

Figure 10. The observed aging process of Nomex paper

B1 B2U ) ,U 0 T

where U and T are the applied voltage and temperature, A1, A2, B1 and B2 are constants which can be determined by the aging tests of different voltage and temperature levels, t is the time to failure. The reliability of the insulation system and its components can be estimated by the loading guide model as shown in [3] based on the remaining life. However it needs huge workload to get the constants A1, A2, B1 and B2, hence the method will not be given in detail in the paper. It is assumed that the expected life of dry-type transformers is 20 years, and the loss of life (LOL) is defined as: LOL 20t1 t 2

(a)

(b)

Figure 11. Distribution of the defects caused by electrical aging.

The Photomicrographs of Nomex paper during different aging periods were obtained using Leica DM2500M microscope and shown in Figure 10 and Figure 11. It can be seen from the photomicrographs that the conditions for aging periods of 1 week, 2 weeks, 3 weeks and 4 weeks as shown in Figure 10. Figure 10a shows that Nomex fibers are getting transparent. The small black spots can be observed around fibers in Figure 10b, which suggests that the defects were generated from these regions. In Figure 10c, a lot of bigger defective zones like chrysophoron distribute around the surface positions where fibers interlace together. It has been proved that if the fibers are oriented in one direction, extremely high strength characteristics of fibrous materials can be obtained, which is the same with the results in [23]. The color of defective zones changes from yellow to green, and the morphology from “chrysophoron” to “butterfly wing” as shown in Figure 10c and Figure 10d. Figure 10c, Figure 10d and Figure 11 show that the defects of Nomex start from the positions of fibers interlacing and develop along the fiber direction, which are consistent with those from [24] and [25]. The results also show that each site of a broken chain may lead to a successive breakage of its neighbor units by the local electric field concentration according to thermally activated bond breakdown mechanism [25].

(28)

where t1 is the aging time of the specimens, t2 is the lifetime of the specimens under the condition of aging test. The life end of specimens is defined at the point when the DPvalue is below 30% of its initial value. The lifetime of specimens under the aging test can be accordingly determined. The LOL and the remaining life RL of the ten groups of specimens can be calculated by (28) and the result is shown in Table 6. The reliability of the ten groups of specimens was estimated with the proposed method based on the proposed models. The estimated reliability results show the good consistency with the remaining life. The process of the reliability estimation and results are presented as follow. 4.1 COMPONENT RELIABILITY 4.1.1 RELIABILITY FROM DM Based on NETZSCH TG 209 F3, the TGA measurements for Nomex paper under argon were performed to obtain the parameters of DM. The TG curves for different heating rate are show in Figure 12.

4 RELIABILITY ASSESSMENT OF THE INSULATION SYSTEM It is not realistic to test many mining flameproof dry-type transformers to check the accuracy of the proposed method. In this paper, only the turn-to-turn insulation, i.e. Nomex specimens are tested and estimated. Ten groups of insulation specimens are tested.

(27)

Figure 12. TG curves for Nomex paper under argon.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

The Ozawa method based on equation (15) was utilized to fit the data. The temperature T in equation (15) is the characteristic temperature from the maximum slope of the TG curve. The results are shown in Figure 13.

Figure 13. The lg vs 1/T curve of Nomex paper.

The frequency factor A and the molar activation energy E were obtained and shown in Table 5. Table 5. The parameters of the thermal decomposition kinetics parameters

Values

E(kJ/mol) A(1/s)

429 9.81013 1

n

2005

Table 7. Statistical characteristics of partial discharge S 1 1 1 2 2 2 3 3 3 4 4 4

+ k

S

0.046 0.037 -0.022 0.129 0.865 0.327 14.222 18.597 1.63 17.222 18.597 15.532

Ku+

Sk-

KU-

CC

QF

Rr

0.015 -0.354 -0.058 0.008 -0.865 0.058 0.314 6.75 -1.871 7.34 6.75 0.451

-2.816 -2.951 -2.886 -2.543 -0.16 1.078 63.37 90.25 0.002 73.37 90.25 57.09

-2.987 -1.881 -2.708 -2.991 -0.16 -2.735 -1.361 27.375 0.298 15.361 27.375 -1.452

2.08 1.276 1.361 1.338 1.011 1.333 2.027 0.693 0.754 1.027 0.693 1.753

0.536 0.639 0.779 0.043 0.035 0.021 0.105 0.094 0.552 0.105 0.094 0.098

0.98 0.98 0.98 0.95 0.95 0.95 0.85 0.85 0.85 0.65 0.65 0.65

4.1.3 RELIABILITY FROM EJM The reliability parameters of the ten groups of specimens estimated based on the photomicrographs by three experts are shown in Table 8. 4.1.4 RELIABILITY OF THE TURN-TO-TURN INSULATION The reliabilities of the turn-to-turn insulation specimens calculated using equation (29) are also shown in Table 8. The weights were determined according to the degree of the correlation between the models and the aging process of the Nomex. Rt t (t ) 0.5 Rdp 0.3Rr 0.2 Re

The threshold value of the intrinsic viscosity was defined to be 30% of its initial value according to the threshold value of DP. The reliability estimated using the degradation model from the ten groups of specimens is shown in Table 6. Table 6. Reliability estimated by the Degradation model No

LOL

RL

[

Ln[

Rdp

1 2 3 4 5 6 7 8 9 10

1.2 2.5 5.1 5.3 7.4 8.0 8.2 13.1 15.0 17.0

18.8 17.5 14.9 14.7 12.6 12.0 1.8 6.9 5.0 3.0

164.5 155.2 163.6 159.6 149.1 158.6 148.2 121.3 83.1 95.4

5.10 5.04 5.10 5.07 5.00 5.07 5.00 4.80 4.42 4.56

0.99 0.97 0.99 0.98 0.97 0.98 0.97 0.92 0.78 0.86

4.1.2 RELIABILITY FROM RM The statistical characteristics +Skewness (Sk+), +Kurtosis (Ku+), -Skewness (Sk-), -Kurtosis (Ku-), Cross-correlation (cc) and Discharge factor (Qf) of the PD signals extracted by the corresponding formulas are shown in Table 7 where S represents the insulation state of the specimens corresponding to the four PD patterns shown in Figure 9a to 9d respectively and Rr is the target of the RBF. The RBF was trained with data of Table 7 and then used to identify the insulation state of the ten groups of specimens.

(29)

Table 8. Viscosity parameters in different aging time No

RL

Rdp

Rr

Re

Rt-t

1 2 3 4 5 6 7 8 9 10

18.8 17.5 14.9 14.7 12.6 12.0 1.8 6.9 5.0 3.0

0.99 0.97 0.99 0.98 0.97 0.98 0.97 0.92 0.78 0.86

0.95 0.95 0.98 0.98 0.98 0.98 0.95 0.85 0.66 0.85

0.97 0.97 0.99 0.97 0.97 0.97 0.97 0.93 0.80 0.85

0.974 0.964 0.987 0.978 0.973 0.978 0.964 0.901 0.748 0.855

4.2 RELIABILITY OF THE INSULATION SYSTEM The reliability of the insulation system of flameproof drytype transformers can be calculated by equation (18) based on reliability of the components.

5 DISCUSSION The service life of a mining flameproof dry-type transformer depends largely on the properties of the insulation system. The characteristics of the insulation paper will affect the degradation rate of the insulation system. The insulation paper with low quality will lead to premature insulation degradation under the action of multiple stress factors, which in turn can result in transformer failures [11]. It is hence necessary to estimate the reliability of the insulation system regularly.

2006

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

The insulation system is divided into four components to prove the effectiveness of the proposed method. The turn-toturn insulation made of Nomex paper is utilized as an example to describe the reliability evaluation process of components. The main chemical composition of Nomex (meta-aramid) paper is polyisophthaloyl metaphenylene diamine (PMIA). The chemical structure of the repeating unit of the Nomex is shown in Figure 14. The polymeric chains associated with crystalline and amorphous regions, and the aging process of Nomex in [27] has been verified in Figure 10a and 10b. Figure 8 clearly shows that the DPvalue decreases with aging time, which is because the interunit linkages in the polymeric chains has cleaved and the mechanical strength of the Nomex decreased. This result is similar to that in [24] which indicated that the polymer degradation process generally starts from chain scission. It can be concluded that the insulation failure of the Nomex paper is coupled with its DP-value and the Degradation model is effective in the reliability estimation.

during the aging periods. The microscopic characteristics of Nomex paper in different aging time shown in Figure 10 and Figure 11 indicate that the defects of the Nomex paper begin from the positions of the fibers interlacing and grow along the fiber direction, which suggests that extremely high strength characteristics of fibrous materials can occur if the fibers are oriented in one direction. Additionally, it can be found from Figure 10 that the morphology of Nomex paper changes from “transparent” to “small black spots” in the first place, then to “chrysophoron” and finally to “butterfly wing” with the aging time. These laws indicate that the degradation of Nomex paper follows a special rule, which confirms that the photomicrographs of the Nomex paper can be used as a QP in the reliability model. It can be seen from the analysis that the results of reliability estimation for the turn-to-turn insulation are relative accurate. Therefore the proposed method is effective. It can be deduced that the method can be utilized to estimate the reliability of insulation system for other dry-type transformers.

Figure 14. Chemical structure of repeating unit of meta-aramid fiber

Generally speaking, the chemical decomposition of metaaramid insulation can attribute to three different processes, namely oxidation, hydrolysis and pyrolysis [28]. Forms and products of the decomposition of the Nomex are determined by aging conditions and the aging time. In the aging test, the hydrolysis is the major way of the decomposition because the aging temperature is lower. There are three hydrolytic ways of Nomex as depicted in Figure 15. Hence, the excess of moisture or water in solid dielectrics of electrical equipment is always considered to be a important reason that can cause the insulation degradation and will lead to insulation faults under high electrical stress. The electrical properties of the solid polymer, which are influenced by a number of different factors, such as temperature, humidity, test duration and applied voltage type etc., decline with the degradation of insulation materials. In fact, the PD signal of the insulation can characterize the performance of the insulation system. The PD signals of the turn-to-turn insulation specimens obtained in this paper show that the PD is a highly stochastic phenomenon and the aging state of insulation cannot be simply determined by its amplitude and times. As a result, it is necessary to extract the statistical characteristics of PD signals so as to provide the more information for the reliability estimation. In addition, it has been found by the aging test that the PD signals have four patterns during the aging time which are corresponding to four states of the insulation as shown in Figure 9. Since the local photomicrographs of the insulation material can reflect its aging degree in terms of the microscopic characteristics, the photomicrographs of the turn-to-turn insulation specimens were periodically obtained

(a)

(b)

(c) Figure 15. Ways and products of fiber decomposition by hydrolysis

6 CONCLUSION This paper proposes a method to evaluate reliability of the insulation system of a dry-type transformer based on DM, RM and EJM. The three models have been developed based on quality parameters (QPs). The method is illustrated by estimating the reliability of the insulation system of a mining flameproof dry-type transformer. The accelerated aging tests on the turn-to-turn insulation, i.e. Nomex paper specimens have been conducted. The DPs, PDs and PGs of the Nomex samples were measured periodically during aging test. The condition of insulation specimens can be observed by the parameters obtained from the aging test.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

The DP of the aged specimens dramatically decline with time, which can be also found in the cellulose paper. Hence the decomposition of the Nomex should have the similar characteristics to the cellulose, which confirms that the DP can reflect the degradation of the Nomex paper and provides theoretical support for the degradation model. Four PD patterns corresponding to the insulation conditions were established from the PD signals. It was found that there is not a regular relationship between the amplitude of PD signals and the insulation condition, the number of PD signals as well. Hence the statistical characteristics were extracted and selected as the inputs of the Regression model. The photomicrographs of specimens in different aging time indicate that the microscopic characteristic of the Nomex follows a special law. The degradation of the Nomex is generated from the regions where fibers interlace together and develop along the fibers. The parameters of the proposed models were calculated using the QPs. Ten groups of Nomex paper specimens were tested and estimated to simulate the turn-to-turn insulation of flameproof dry-type transformers. The estimation results demonstrate that the reliability of the specimens is closely related to the remaining life of the transformers, which verified the effectiveness of the proposed methods and models. The different QPs measured during the aging test can be used as the benchmarks to predict life of the same insulation type of transformers. The real time parameters of the online transformers from the monitoring system and obtained during the maintenance are required to be compared with the standard parameters from the aging test to predict the remaining life.

[6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

[17] [18] [19] [20]

ACKNOWLEDGMENT The authors would like to thank the Major science and technology projects in Shanxi Province and the State international technology cooperation projects of the Ministry of Science and Technology of the People’s Republic of China for the financial support.

[21]

[22] [23]

REFERENCES [1] [2] [3]

[4] [5]

L. W. Pierce, “Thermal considerations in specifying dry-type transformers”, IEEE Trans. Ind. Appl., Vol. 30, No. 4, pp. 1090-1098, 1994. P. A. A. F. Wouters, A. van Schijndel, and J. M. Wetzer, “Remaining lifetime modeling of power transformers: individual assets and fleets”, IEEE Electr. Insul. Mag., Vol. 27, No. 3, pp. 45-51, 2011. B. Gorgan, P. V. Notingher, J. M. Wetzer, H. F. A. Verhaart, and P. A. A. F. Wouters, “Calculation of the remaining lifetime of power transformers paper insulation”, IEEE 13th Int’l. Conf. Optimization of Electrical and Electronic Equipment (OPTIM), pp. 293-300, 2012. H. C. Stewart, L. C. Whitman, and A. L. Scheideler. “Aging evaluation of dry-type transformer insulation system”, IEEE Trans. Power App. Syst., Vol. 72, No. 2, pp. 267-277, 1953. T. R. Walters and A. L. Scheideler. “A study of models for use in evaluating dry-type transformer insulation systems”, IEEE Trans. Power App. Syst., Vol. 75, no. 3, pp. 520-527, 1956.

[24]

[25] [26] [27] [28]

2007

H. C. Stewart and L. C. Whitman. “Aging characteristics of dry-type transformer insulation at high temperature”, Trans. Amer. Inst. Electr. Eng., Trans., Vol. 67, No. 2, pp. 1600-1607, 1948. H. C. Stewart and L. C. Whitman. “Hot-spot temperatures in dry-type transformer windings”, Trans. Amer. Inst. Electr. Eng., Trans., Vol. 63, No. 10, pp. 763-768, 1944. L. W. Pierce. “Prediction hottest spot temperatures in ventilated dry type transformer windings”, IEEE Trans. Power Del., Vol.9, No. 2, pp. 11601172, 1994. A. Van Schijndel, J. M. Wetzer, and P. A. A. F. Wouters, “Reliability estimation of paper insulated components”, IEEE Conf. Electr. Insul. Dielectr. Phenomena (CEIDP), pp.17-20, 2007. G. C. Stone, “A perspective on online partial discharge monitoring for assessment of the condition of rotating machine stator winding insulation”, IEEE Electr. Insul. Mag., Vol. 28, No. 5, pp.8-13, 2012. A. Van Schijndel, Power Transformer Reliability Modeling, Ph.D. Dissertation, Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven, the Netherlands, 2010. A. Van Schijndel, P. A. A. F. Wouters, E. F. Steennis, and J. M. Wetzer, “Approach for an integral power transformer reliability model”, Euro. Trans. Electr. Power, Vol. 22, pp.491-503, 2012. M. Farahani, E. Gockenbach, and H. Borsi, “Behavior of machine insulation systems subjected to accelerated thermal aging test”, IEEE Trans. Dielectr. Electr. Insul., Vol. 17, No. 5, pp.1364-1372, 2010. IEEE Guide for determination of hottest-spot temperature in dry-type transformers, pp. 6, 2000. G. Huang, P. Saratchandran, and N. Sundararajan, “A generalized growing and pruning RBF (GGAP-RBF) neural network for function approximation”, IEEE Trans. Neural Networks, Vol. 16, No. 1, pp. 1929-1934, 2005. D. Evagorou, A. Kyprianou, P. L. Lewin, A. Stavrou, V. Efthymiou, A. C. Metaxas, and G. E. Georghiou “Feature extraction of partial discharge signals using the wavelet packet transform and classification with a probabilistic neural network”, IET Sci. Measurement Technology (UK), Vol. 4, No. 3, pp. 177-192, 2010. C.-C. Kuo, “Artificial identification system for transformer insulation aging”, Expert Systems with Applications, Vol. 37, pp. 4190-4197, 2010. IEEE Standard Test procedure for thermal evaluation of insulation systems for ventilated dry-type power and distribution transformers, 1986. R. Morin, R. Bartnikas, and P. Ménard, “A three-phase multi-stress accelerated electrical aging test facility for stator bars”, IEEE Trans. Enegry Conversion, Vol. 15, No. 2, pp. 149-156, 2000. L. Yao, C. Lee, and J. Kim, “Fabrication of electrospun meta-aramid nanofibers in different solvent systems”, Fibers and Polymers, Vol. 11, No. 7, pp. 1032-1040, 2010. D. Harwood, H. Aoki, Y. D. Lee, J. F. Fellers, and J. L. White. “Solution and Rheological Properties of Poly(m-phenyleneisophthalamide) in Dimethylacetamide/LiCl”, J. Appl. Polymer Sci., Vol. 23, pp. 2155-2168, 1979. IEEE Recommended practice for the detection of partial discharge and the measurement of apparent charge in dry-type transformers, 1991. H.-Z. Ding and B. R. Varlow, “Raman spectroscopy—a technique to assess the residual stress in fiber-reinforced polymeric insulation materials”, IEEE Electr. Insul. Mag., Vol. 20, No. 1, pp. 5-13, 2004. H.-Z. Ding and B. R. Varlow, “Thermodynamic model for electrical tree propagation kinetics in combined electrical and mechanical Stresses” IEEE Trans. Dielectr. Electr. Insul., Vol. 12, No. 1, pp. 8189, 2005. H.-Z. Ding, X. -S. Xing and H.-S. Zhu, “A kinetic model of timedependent dielectric breakdown for polymers”, J. Phys. D: Appl. Phys., Vol. 27, pp. 591-595, 1994. P. Cygan and J. R. Laghari. “Models for Insulation Aging under Electrical and Thermal Multistress”, IEEE Trans. Electr. Insul. Vol.25. No. 5, 1990, pp. 923-934. L. E. Lundgaard, W. Hansen, D. Linhjell, and T. J. Painter, “Aging of oil-impregnated paper in power transformers”, IEEE Trans. Power Del., Vol. 19, No. 1, pp. 230-239, 2004. S. Villar-Rodil, A. Martı́nez-Alonso, and J. M. D. Tascón, “Studies on pyrolysis of Nomex polyaramid fibers”, J. Anal. Appl. Pyrolysis, Vol. 58 –59, pp. 105–115, 2001.

2008

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers Minmin Wen obtained the B.Sc. degree in Taiyuan University of Technology, China, in 2008. She has been studying at Taiyuan University of Technology for the M.Sc. and Ph.D. degrees since 2008. She works on condition monitoring and diagnoses of high voltage insulation for power apparatus.

Jiancheng Song received the B.Sc. degree from Taiyuan University of Technology, China, in 1982, the M.Sc. degree from Newcastle University, England, in 1987, respectively and the Ph.D. degree from Xian Jiaotong University, China, in 1999. Currently, he is a professor of the College of Electrical and Power Engineering at Taiyuan University of Technology. He has experience in the field of condition assessment, remaining life assessment and intellectual automation technology. He has performed a number of electrical failure investigations about coal mine. He has presented a number of technical and scientific papers at international conferences and seminars. Yuan Song obtained the B.Sc. and M. Sc. Degrees in Taiyuan University of Technology, China, in 2008 and 2011, respectively. He has been studying at Taiyuan University of Technology for the Ph.D. degree since 2011. He works on intellectual electric apparatus and process automation technology.

Yuan Liu obtained the B.Sc. and M. Sc. degrees both in Taiyuan University of Technology, China, in 2008 and 2011 respectively. She has been studying at Taiyuan University of Technology for the Ph.D. degree since 2011. She works on Power electronic transformation and automatic control technology.

Chuanyang Li received the B.S. degree from Taiyuan University of Technology, China, in 2011. He has been studying in Taiyuan University of Technology for the M.S. degree since 2011. His main research interest is condition monitoring and PD pattern recognition for HV motors.

Peng Wang (M’00) received the B.Sc. degree from Xian Jiaotong University, China, in 1978, the M.Sc. degree from Taiyuan University of Technology, China, in 1987, and the M.Sc. and Ph.D. degrees from the University of Saskatchewan, Canada, in 1995 and 1998, respectively. Currently, he is a professor at Taiyuan University of Technology, China and associate professor at Nanyang Technological University, Singapore.

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

Reliability Assessment of Insulation System for Dry Type Transformers Minmin Wen, Jiancheng Song, Yuan Song, Yuan Liu, Chuanyang Li and Peng Wang Shanxi Key Laboratory of Coal Mining Equipment and Safety Control Taiyuan University of Technology, No.79 Yingze West Street, Taiyuan, 030024, China

ABSTRACT The insulation system of a dry-type transformer under the combined thermal, electrical, mechanical, and environmental stresses undergoes gradual deterioration, which ultimately leads to transformer failure. It is hence important to test specimen insulation of transformer coil so as to explore the aging law and the associated characteristic parameters related to the insulation condition. This paper proposes a method to evaluate reliability of the insulation system of a dry-type transformer based on the three models of degradation, regression and expert judgment. The three models are developed based on the quality parameters (QPs) from the lab tests. The method is illustrated by estimating the reliability of the insulation system of a mining flameproof dry-type transformer. The accelerated aging tests on the Nomex paper (i.e. the turn-to-turn insulation of the transformer) are conducted. The QPs of the Nomex samples of degree of polymerization, partial discharges, and photomicrograph were measured periodically during aging test. The effectiveness of the proposed method was proved through the lab experiments. The results show that the parameters from the aging test can be used to represent the condition of the insulation system. Index Terms — dry-type transformer insulation, reliability assessment, degree of polymerization, partial discharge, photomicrograph.

1 INTRODUCTION TRANSFORMERs play an important role in the electrical power transmission and distribution networks. The first transformer in the world invented in the 1880s was of dry type [1]. Oil as insulation material of transformers was started by Elihu Thomson and the first commercial oil-immersed transformer was introduced by Westinghouse in 1886. Compared with dry-type transformers, the oil-immersed transformers are not suitable in some particular places for its relatively poor fireproof performance and ability of environmental protection. Therefore, since 1990s, oil-immersed transformers have been replaced by dry-type transformers in many industrial and commercial installations. Recently, the failures caused by the decline of the insulation properties of dry-type transformers are frequently reported due to non-sinusoidal loads. The lifetime of a dry-type transformer mainly depends on the condition of insulation material. The insulation system under combined thermal, electrical, mechanical, and environmental stresses undergoes gradual deterioration, which may lead to serious accidents and great losses for utilities and customers. The insulation system of a dry-type transformer mainly refers to the winding insulation. The rated voltage and loading of dryManuscript received on 30 January 2013, in final form 12 June 2013.

type transformers used in mining industry, China have been increased in recent years. Therefore winding insulation failure due to high voltage and loading operation might become the dominant cause of transformer failure in future [2-3]. It is essential to develop the reliability assessment methods to estimate the operating condition of the insulation system and to predict earlier failures of the transformer. Considerable research works [4-6] on aging evaluation of the insulation of dry-type transformers have provided theoretical basis for further research. The influence of temperature on insulation properties of dry-type transformers has been mainly studied [7-8]. This paper proposes a method to evaluate reliability of the insulation system of dry-type transformers based on the three models of degradation, regression and expert judgment. The three models are developed based on QPs. The method is illustrated by estimating the insulation system reliability of a mining flameproof dry-type transformer. In particular, its turnto-turn insulation, i.e. Nomex paper, is taken as a component example to illustrate the method. The accelerated aging tests on the Nomex paper specimens are conducted. The QPs of the Nomex paper such as degree of polymerization (DP), partial discharges (PDs) and photomicrograph (PG) were measured periodically during the aging test. The parameters obtained from the aging test are analyzed to represent the condition of insulation specimens. The effectiveness of the proposed method was proved through experiments.

1070-9878/13/$25.00 © 2013 IEEE

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

1999

2 RELIABILITY MODELING 2.1 QUALITY PARAMETERS The reliability of the insulation system of a dry-type transformer is evaluated based on operating conditions of system components represented by QPs. In [9], the QPs are classified into the directly linked measurable QP (D-M-QP) which is measurable and directly related to a degradation process, the indirectly linked measurable QP (I-M-QP) which is externally measurable and indirectly related to the degradation process, and the indirectly linked comparative QP (I-C-QP) which is not measurable and is indirectly related to the degradation process. The QPs are divided into two categories in this paper, which are measurable QPs and un-measurable QPs. Both the D-M-QP and I-M-QP belong to measurable QPs. Un-measurable QPs include the I-C-QP, the observable information (OI) from measure instruments and artificial expertise (AE) from the historical data of transformers and the experience of experts, etc. The QPs are utilized to establish the reliability model of the insulation system of dry-type transformers in next section. 2.2 MODELING Degradation model (DM), regression model (RM) and expert judgment model (EJM) are proposed to estimate the component reliability of the insulation system based on QPs. The procedure of component reliability assessment based on the measurable and un-measurable QPs is shown in Figure 1. The turn-to-turn insulation material (Nomex paper) of a mining flameproof drytype transformer is used as an example.

Figure 2. The condition change process due to degradation.

The decline of DP is a result of the scission of the polymeric chains caused by cascade chemical reaction. The relationship between temperature and DP is given by Arrhenius through the reaction rate k [11-13]. DP as the function of time t is given as DP (t )

DP (t 0 ) t

1 DP (t 0 ) k ( )d t0

where DP(t) is the DP at time t, DP(t0) is the DP at an initial time t0 and k(t) is reaction rate at time t, which can be obtained from Arrhenius equation: k (t ) A exp( E RT (t ))

C ' C ( HS TA Tk ) ( C TAT Tk )N

The D-M-QP include the degree of polymerization (DP) which indicates the degradation of the Nomex paper; The I-M-QP include the partial discharge (PD) which does not directly contribute to a failure mechanism in [10]; OI include the photomicrograph (PG) of the Nomex paper; The AE is based on the historical failure records from the condition monitoring of the mining flameproof dry-type transformers and the experience of experts. 2.2.1 DEGRADATION MODEL The degradation model as shown in Figure 2 is proposed to track the degradation process of the insulation. The DP is changing with the insulation degradation. The degradation can be determined by the decline of the DP.

(2)

where A is a frequency factor which depends on water content, oxygen or acidity, E is the molar activation energy of the degradation reactions (hydrolysis and thermo-oxidation), R is the universal gas constant, and T(t) is the temperature at time t during the reaction in Kelvin calculated by equation (3) and equation (4) in IEEE Std C57.134-2000 [14]. It is important to note that T is the experiment temperature in thermogravimetric analysis (TGA).

HS C ' 1 ( E C ' )

Figure 1. Flow chart of the reliability estimation method.

(1)

1

N

(3)

N

(4)

where C is the temperature rise of hottest-spot during current-only test, C’ is corrected winding hottest-spot temperature rise due to current-only test, E is winding hottest-spot temperature rise during excitation voltage-only test, HS is winding hottest-spot temperature rise for rated conditions, TA is rated ambient temperature which is usually 30℃, TAT is actual ambient temperature during test, Tk is 234.5 ℃ for copper and 225 ℃ for aluminum, N is 0.8 for self cooled and 0.9 for forced air. The reliability and failure probability due to DP is described in [3]. The failure probability is the probability that the DP-value was lower than its threshold value when insulation fails. It involves two probability distributions. The first is the probability distribution of DP-value at time t. The probability of DP with value between x and x+dx at time t is denoted as pdp(x,t)dx. And the second distribution is DPvalue of which below the threshold value. The probability of the threshold having a value between x and x+dx is denoted as pth ( x)dx . The probability that the threshold being above a certain DP-value x is given by equation (5) and the failure probability at time t is calculated by equation (6) [3].

Pth ( x) pth ( x' )dx' x

(5)

2000

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

Pf (t ) Pth ( x) pdp ( x, t )dx

(6)

0

The time dependent reliability based on DP can be calculated using equation (7). Rdp (t ) 1 Pf (t )

(7)

In this paper, the threshold value of DP is determined to be 30% of the initial value of DP. The degradation of the Nomex paper follows the reaction kinetics equation (8).

d dt k (t )(1 ) n

(8)

where is the reaction extent which is given in equation (9), n is the reaction order. w0 wt w0 w

(9)

where w0 is the initial mass of the sample in the TGA measurement, wt is the mass at time t, w is the mass at the end of the measurement. The parameters A and E are determined using the following procedure. It is assumed that the heating rate of the TGA measurement is

, dT (t ) dt . Combining (2) and (8) gives d (1 ) n A exp( E RT (t )) dT (t )

Integrating

(10)

0 (1 )

n

d

exp( E T

RT ) dT

T

The inputs of RBF can be the features of PD signals such as statistical characteristics from wavelet packet transformation [16] and time domain characteristics, since the pattern of PD changes with insulation degradation. In this paper, six statistical characteristics of PDs are extracted and serve as the inputs of the RBF, i.e. +Skewness, +Kurtosis, -Skewness, -Kurtosis, Crosscorrelation and Discharge factor [17]. In order to improve the accuracy, the least mean square (LMS) algorithm was used for updating the network parameters (centers, widths, and weights of Gaussian neurons). The output of the RBF is the component reliability Rr. 2.2.3 EXPERT JUDGMENT MODEL The expert judgment model is based on the AE and OI. Component reliability is determined by the experts through grading the component according to the experience and the photomicrographs of the insulation. Assuming that n experts participate in judgment, component reliability Re can be calculated as

Re the

A

2.2.2 REGRESSION MODEL The radial basis function neural network (RBF) [15] with three layers is utilized as the regression model to evaluate reliability.

(10)

with

0 and T (0) T0 , gives

The A and E can be obtained from (15) through functionfitting for given the heating rate and the reaction order n. The reaction order n of Nomex is about 1.

initial

conditions

1 (r1 r2 ... rn ) n

where ri is the reliability judgment by the ith expert. Table 1. Relationship between reliability and grades

(11)

0

Defining p( x)

x

e x dx and x E RT , equation (11) can x2

be represented by

0

(1 ) n d

AE p( x) R

(12)

The approximate expression of Doyle approximate function p(x) is as ln p ( x ) 5.331 1.052 x

(13)

Substituting equation (13) and x E RT into equation (12), gives ln[

0

AE E (1 ) d ] ln( ) 5.331 1.052 R RT n

(14)

Equation (14) also can be transformed into equation (15). lg lg

AE

R (1 ) d 0

n

2.315 0.4567

E RT

(15)

Reliability (Rr)

grade

98%＜ Re≤100%

normal state

95%＜ Re ≤98%

attention state

85%＜Re≤95%

abnormal state

0＜ Re≤85%

severe state

The lifespan of the insulation is divided into four states corresponding to the insulation grades of normal, attention, abnormal and severe state. The relationship between the reliability and states are shown in Table 1. 2.2.4 COMPONENT RELIABILITY The weighted mean method is utilized in this paper to calculate the reliability of component i of a insulation system as n

(16)

Ri (t ) w j R j (t )

(17)

1

where Rj is the reliability obtained from the jth model of the component with weight wj. The weights are determined according to the degree of correlation between insulation model and aging process. The accuracy of the model is also considered.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

2.2.5 INSULATION SYSTEM RELIABILITY The insulation system of a mining flameproof dry-type transformer can be divided into four components of turn-toturn insulation, layer-to-layer insulation, windings-to-ground insulation and winding-to-winding insulation. The main insulation materials of the four components are Nomex paper, polyimide film, combination of Nomex paper and air, panel made of imide and glass fiber, respectively. From the insulation structure characteristics of mining flameproof drytype transformers, it is found that the impacts of components on insulation system reliability are independent. The insulation system cannot operate normally if any of the components fails. Therefore, the reliability of the insulation system is modeled as a series system as n

Rs (t ) Ri (t )

(18)

1

where Rs is the reliability of the insulation system, Ri is the reliability of the ith component of the insulation system.

3 MEASUREMENTS AND RESULTS The turn-to-turn insulation specimens of a mining flameproof dry-type transformer are tested in this section. The QPs were obtained from the experiments using the proposed method. 3.1 ACCELERATED AGING TEST 3.1.1 INSULATION MODEL

(a)Copper electrode (b)Turn-to-turn insulation model Figure 3. Copper electrode and turn-to-turn insulation model.

The model shown in Figure 3 is designed according to the characteristics of turn-to-turn insulation. The model includes two copper electrodes of the high-field electrode and the lowfield electrode. The Nomex paper from the DuPont Corporation of America is cut into strips and fixed between the two electrodes to simulate the turn-to-turn insulation. There are three layers of paper strips (one group) and each layer is 0.8mm thickness in the model. 3.1.2 ELECTRICAL-THERMAL AGING TEST The aging tests were performed according to the IEEE Standard C57.12.56 [18]. Before the aging test, the specimens of Nomex paper were dried under 90 ℃ for 12 hours. The specimens were tested under the voltage and temperature used in [19] and [13] respectively. 1.6 kV AC, which is about 1.6 times of the inception voltage of partial discharge for the specimens, is applied to the specimens. The aging temperature is kept at 130℃. The specimens were aged for 10 hours per day. The related QPs were measured each week during the aging process periodically.

2001

At least 5 groups of specimens are taken out per week for the measurement so as to guarantee the accuracy of the QPs.

(a) (b) Figure 4. Macroscopic photos of the specimens before and after the aging test

The two groups of aged specimens are shown in Figure 4a and 4b respectively. The left specimen in Figure 4a and 4b is intact specimen. The two groups of specimens (the right three specimens in Figure 4a and 4b) are specimens aged for 1 week and 4 weeks respectively. It can be found that the specimen color is becoming more and more yellow with increase of the aging time. 3.2 QUALITY PARAMETERS 3.2.1 DP MEASUREMENT DP of the Nomex paper is generally measured indirectly by viscometry. The viscosity of the Nomex in special solvent is measured in this paper. It is usually very difficult to dissolve aromatic polyamides with traditional solvents. The literature [20] indicate that aromatic polyamides can be dissolved in polar solvents, such as N,N-dimethylacetamide (often only in the presence of lithium salt), or strong acid, such as concentrated sulfuric acid or chlorosulphonic acid. Therefore the sulfuric acid is selected as the solvent of the Nomex paper specimens. The measurement system is shown in Figure 5. The viscosity of the Nomex was measured in 96% sulfuric acid. The initial concentration of the dilute solution made of the Nomex and sulfuric acid was 0.005 g/ml. In order to ensure the accuracy of measurement, the concentration of the sulfuric acid solution was controlled within 95%~97%. At the same time, the temperature was maintained within 30 0.3℃. The efflux time of the solution was measured by high precision timer of 1/100 second. The intrinsic viscosities were calculated by extrapolating the reduced viscosity measured by the capillary viscosimeter (Ostwald viscosimeter,=0.8 mm) until the concentration is zero.

Figure 5. The viscosity measurement device

2002

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

The measurements of the aged specimens were performed per week. Five groups of specimens are measured simultaneously to calculate the average intrinsic viscosity. The relationship between DP-value and intrinsic viscosity is derived as follow: Mark-Houwink-Sakurada [19] defined the relation between the intrinsic viscosity [ and the absolute weight-average molar mass M as

[ ] aM b

(19)

signals in specimens were measured during different aging periods in this paper. Then the PD patterns were established according to the insulation conditions of specimens. The PD measurement system was designed according to the IEC 60270 standard and IEEE Standard C57.124 [22], which consists of high voltage generator, capacitive voltage divider and coupling device, and the experiments were conducted in the electromagnetic shielded room as shown in Figure 6.

where a and b are constant parameters of the polymer/solvent system. The intrinsic viscosity is calculated by extrapolating the measured reduced viscosity.

[] limsp c limlnr c 100liminh c0

c0

c0

(20)

where c is the concentration of the measured solution, is the relative viscosity of the measured solution, inh is the inherent viscosity of the measured solution, sp is the specific viscosity of the measured solution. The equations of Huggins (21) and Kraemer (22) are developed from equation (20).

sp c [ ] [ ]2 c

(21)

ln r c [ ] [ ] 2 c

(22)

Figure 6. The system for aging and PD tests

3.2.3 PHOTOMICROGRAPHS OF NOMEX PAPERS

where and are constants. The relationship between the absolute weight-average molar mass and the viscometric DP-value P is depicted by M P M0

(23) Figure 7. Leica DM2500M microscope.

where M0 is molecular weight of the monomeric unit. Combining equation (23) with (19), gives [ ] a ( PM 0 ) b

(24)

Taking logarithm on both sides of equation (24), gives

ln P

1 1 ln[ ] ln M 0 ln a b b

(25)

1 1 Introducing b and a ' (ln M 0 ln a ) , gives b b '

ln P b ' ln[ ] a ' (26) Equation (26) shows the linear relationship between ln[] and lnP. As a result, P can be replaced by [] to characterize aging status of the insulation and the same replacement should be done in the equation (1), (5) and (6). 3.2.2 PD MEASUREMENT Partial discharges (PD) might happen when there are defects in the insulation of electrical equipment. The PD signals can be used to identify the nature of failure and aging status of the insulation system, which should make a breakdown distribution on fault diagnosis and failure warning. The PD

It has been verified that the local photomicrographs of the insulation material can reflect its aging degree in terms of the microscopic characteristics. Leica DM2500M microscope was used to observe the microscopic structure of the aged specimens as shown in Figure 7. 3.3 MEASURED RESULTS 3.3.1 DP MEASUREMENTS The DP was measured per week during the aging test. The viscosity data of the intact samples and the aged samples of 28 days are shown in Table 2 and Table 3 respectively. Table 2. Viscosity of the intact sample. No.

0

1

2

3

4

5

c t0 t r sp sp/c lnr/c

/ 965 / / / / /

0.005 / 1904 1.973 0.973 194.6 135.91

0.0025 / 1409 1.460 0.460 184.0 151.38

0.00125 / 1172 1.215 0.215 172.0 155.80

0.000625

0.0003725 / 1016 1.053 0.053 169.6 165.26

1070 1.109 0.109 174.4 165.53

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

Table 3. Viscosity of the aged sample of 28 days No.

0

1

2

3

4

5

c t0 t r sp sp/c lnr/c

/ 965 / / / / /

0.005 / 1703 1.765 0.765 153.0 113.60

0.0025 / 1307 1.354 0.354 141.8 121.35

0.00125 / 1119 1.160 0.160 127.7 118.45

0.000625

0.0003725 / 984 1.020 0.020 63.0 62.39

1030 1.067 0.067 107.8 104.30

where t0 is the efflux time of the solvent, t is the efflux time of the solution. The intrinsic viscosities of the Nomex/sulfuric acid solutions in different aging periods were calculated and the results are shown in Table 4.

(9b) Partial discharge in the second aging period

Table 4. Viscosity parameters in different aging periods Time(day)

0

7

14

21

28

[ ln[

167.4 5.12

145.5 4.98

113.5 4.73

82.2 4.41

62.7 4.14

(9c) Partial discharge in the third aging period

Figure 8. Viscosity parameters versus aging time.

Table 4 shows that the degradation of the solid insulation is of accumulative effect, and the intrinsic viscosity [] and ln[] declines with aging time as depicted in Figure 8. Generally, the inherent viscosities of Nomex paper samples varied in the range of 0.5 to 8 dl/g. 3.3.2 PD MEASUREMENTS The PD signals for different aging periods as shown in Figure 9 are used to train RBF model. Figure 9e shows the PD when the insulation is close to breakdown. Figures 9a-9d clearly indicate the PD Patterns during degradation process. Those PD patterns are utilized to estimate the reliability of the turn-to-turn insulation.

(9d) Partial discharge in the fourth aging period

(9e) Partial discharge when the insulation is close to breakdown (9a) Partial discharge in the first aging period

Figure 9. Partial discharges in different aging period.

2003

2004

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

3.3.3 PHOTOMICROGRAPHS OF THE NOMEX PAPER

(a)

In this paper, the exponential model by Fallou [26] shown in equation (27) is chose to describe the relationship between the time to failure and the applied stresses on the insulation since it is much more suitable for the thin insulation materials and also considers the influence of the electric and thermal stress together. The expected life of the insulation system and its components can be extrapolated from equation (27).

(b)

t exp( A1 A2U

(c)

(d)

Figure 10. The observed aging process of Nomex paper

B1 B2U ) ,U 0 T

where U and T are the applied voltage and temperature, A1, A2, B1 and B2 are constants which can be determined by the aging tests of different voltage and temperature levels, t is the time to failure. The reliability of the insulation system and its components can be estimated by the loading guide model as shown in [3] based on the remaining life. However it needs huge workload to get the constants A1, A2, B1 and B2, hence the method will not be given in detail in the paper. It is assumed that the expected life of dry-type transformers is 20 years, and the loss of life (LOL) is defined as: LOL 20t1 t 2

(a)

(b)

Figure 11. Distribution of the defects caused by electrical aging.

The Photomicrographs of Nomex paper during different aging periods were obtained using Leica DM2500M microscope and shown in Figure 10 and Figure 11. It can be seen from the photomicrographs that the conditions for aging periods of 1 week, 2 weeks, 3 weeks and 4 weeks as shown in Figure 10. Figure 10a shows that Nomex fibers are getting transparent. The small black spots can be observed around fibers in Figure 10b, which suggests that the defects were generated from these regions. In Figure 10c, a lot of bigger defective zones like chrysophoron distribute around the surface positions where fibers interlace together. It has been proved that if the fibers are oriented in one direction, extremely high strength characteristics of fibrous materials can be obtained, which is the same with the results in [23]. The color of defective zones changes from yellow to green, and the morphology from “chrysophoron” to “butterfly wing” as shown in Figure 10c and Figure 10d. Figure 10c, Figure 10d and Figure 11 show that the defects of Nomex start from the positions of fibers interlacing and develop along the fiber direction, which are consistent with those from [24] and [25]. The results also show that each site of a broken chain may lead to a successive breakage of its neighbor units by the local electric field concentration according to thermally activated bond breakdown mechanism [25].

(28)

where t1 is the aging time of the specimens, t2 is the lifetime of the specimens under the condition of aging test. The life end of specimens is defined at the point when the DPvalue is below 30% of its initial value. The lifetime of specimens under the aging test can be accordingly determined. The LOL and the remaining life RL of the ten groups of specimens can be calculated by (28) and the result is shown in Table 6. The reliability of the ten groups of specimens was estimated with the proposed method based on the proposed models. The estimated reliability results show the good consistency with the remaining life. The process of the reliability estimation and results are presented as follow. 4.1 COMPONENT RELIABILITY 4.1.1 RELIABILITY FROM DM Based on NETZSCH TG 209 F3, the TGA measurements for Nomex paper under argon were performed to obtain the parameters of DM. The TG curves for different heating rate are show in Figure 12.

4 RELIABILITY ASSESSMENT OF THE INSULATION SYSTEM It is not realistic to test many mining flameproof dry-type transformers to check the accuracy of the proposed method. In this paper, only the turn-to-turn insulation, i.e. Nomex specimens are tested and estimated. Ten groups of insulation specimens are tested.

(27)

Figure 12. TG curves for Nomex paper under argon.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

The Ozawa method based on equation (15) was utilized to fit the data. The temperature T in equation (15) is the characteristic temperature from the maximum slope of the TG curve. The results are shown in Figure 13.

Figure 13. The lg vs 1/T curve of Nomex paper.

The frequency factor A and the molar activation energy E were obtained and shown in Table 5. Table 5. The parameters of the thermal decomposition kinetics parameters

Values

E(kJ/mol) A(1/s)

429 9.81013 1

n

2005

Table 7. Statistical characteristics of partial discharge S 1 1 1 2 2 2 3 3 3 4 4 4

+ k

S

0.046 0.037 -0.022 0.129 0.865 0.327 14.222 18.597 1.63 17.222 18.597 15.532

Ku+

Sk-

KU-

CC

QF

Rr

0.015 -0.354 -0.058 0.008 -0.865 0.058 0.314 6.75 -1.871 7.34 6.75 0.451

-2.816 -2.951 -2.886 -2.543 -0.16 1.078 63.37 90.25 0.002 73.37 90.25 57.09

-2.987 -1.881 -2.708 -2.991 -0.16 -2.735 -1.361 27.375 0.298 15.361 27.375 -1.452

2.08 1.276 1.361 1.338 1.011 1.333 2.027 0.693 0.754 1.027 0.693 1.753

0.536 0.639 0.779 0.043 0.035 0.021 0.105 0.094 0.552 0.105 0.094 0.098

0.98 0.98 0.98 0.95 0.95 0.95 0.85 0.85 0.85 0.65 0.65 0.65

4.1.3 RELIABILITY FROM EJM The reliability parameters of the ten groups of specimens estimated based on the photomicrographs by three experts are shown in Table 8. 4.1.4 RELIABILITY OF THE TURN-TO-TURN INSULATION The reliabilities of the turn-to-turn insulation specimens calculated using equation (29) are also shown in Table 8. The weights were determined according to the degree of the correlation between the models and the aging process of the Nomex. Rt t (t ) 0.5 Rdp 0.3Rr 0.2 Re

The threshold value of the intrinsic viscosity was defined to be 30% of its initial value according to the threshold value of DP. The reliability estimated using the degradation model from the ten groups of specimens is shown in Table 6. Table 6. Reliability estimated by the Degradation model No

LOL

RL

[

Ln[

Rdp

1 2 3 4 5 6 7 8 9 10

1.2 2.5 5.1 5.3 7.4 8.0 8.2 13.1 15.0 17.0

18.8 17.5 14.9 14.7 12.6 12.0 1.8 6.9 5.0 3.0

164.5 155.2 163.6 159.6 149.1 158.6 148.2 121.3 83.1 95.4

5.10 5.04 5.10 5.07 5.00 5.07 5.00 4.80 4.42 4.56

0.99 0.97 0.99 0.98 0.97 0.98 0.97 0.92 0.78 0.86

4.1.2 RELIABILITY FROM RM The statistical characteristics +Skewness (Sk+), +Kurtosis (Ku+), -Skewness (Sk-), -Kurtosis (Ku-), Cross-correlation (cc) and Discharge factor (Qf) of the PD signals extracted by the corresponding formulas are shown in Table 7 where S represents the insulation state of the specimens corresponding to the four PD patterns shown in Figure 9a to 9d respectively and Rr is the target of the RBF. The RBF was trained with data of Table 7 and then used to identify the insulation state of the ten groups of specimens.

(29)

Table 8. Viscosity parameters in different aging time No

RL

Rdp

Rr

Re

Rt-t

1 2 3 4 5 6 7 8 9 10

18.8 17.5 14.9 14.7 12.6 12.0 1.8 6.9 5.0 3.0

0.99 0.97 0.99 0.98 0.97 0.98 0.97 0.92 0.78 0.86

0.95 0.95 0.98 0.98 0.98 0.98 0.95 0.85 0.66 0.85

0.97 0.97 0.99 0.97 0.97 0.97 0.97 0.93 0.80 0.85

0.974 0.964 0.987 0.978 0.973 0.978 0.964 0.901 0.748 0.855

4.2 RELIABILITY OF THE INSULATION SYSTEM The reliability of the insulation system of flameproof drytype transformers can be calculated by equation (18) based on reliability of the components.

5 DISCUSSION The service life of a mining flameproof dry-type transformer depends largely on the properties of the insulation system. The characteristics of the insulation paper will affect the degradation rate of the insulation system. The insulation paper with low quality will lead to premature insulation degradation under the action of multiple stress factors, which in turn can result in transformer failures [11]. It is hence necessary to estimate the reliability of the insulation system regularly.

2006

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers

The insulation system is divided into four components to prove the effectiveness of the proposed method. The turn-toturn insulation made of Nomex paper is utilized as an example to describe the reliability evaluation process of components. The main chemical composition of Nomex (meta-aramid) paper is polyisophthaloyl metaphenylene diamine (PMIA). The chemical structure of the repeating unit of the Nomex is shown in Figure 14. The polymeric chains associated with crystalline and amorphous regions, and the aging process of Nomex in [27] has been verified in Figure 10a and 10b. Figure 8 clearly shows that the DPvalue decreases with aging time, which is because the interunit linkages in the polymeric chains has cleaved and the mechanical strength of the Nomex decreased. This result is similar to that in [24] which indicated that the polymer degradation process generally starts from chain scission. It can be concluded that the insulation failure of the Nomex paper is coupled with its DP-value and the Degradation model is effective in the reliability estimation.

during the aging periods. The microscopic characteristics of Nomex paper in different aging time shown in Figure 10 and Figure 11 indicate that the defects of the Nomex paper begin from the positions of the fibers interlacing and grow along the fiber direction, which suggests that extremely high strength characteristics of fibrous materials can occur if the fibers are oriented in one direction. Additionally, it can be found from Figure 10 that the morphology of Nomex paper changes from “transparent” to “small black spots” in the first place, then to “chrysophoron” and finally to “butterfly wing” with the aging time. These laws indicate that the degradation of Nomex paper follows a special rule, which confirms that the photomicrographs of the Nomex paper can be used as a QP in the reliability model. It can be seen from the analysis that the results of reliability estimation for the turn-to-turn insulation are relative accurate. Therefore the proposed method is effective. It can be deduced that the method can be utilized to estimate the reliability of insulation system for other dry-type transformers.

Figure 14. Chemical structure of repeating unit of meta-aramid fiber

Generally speaking, the chemical decomposition of metaaramid insulation can attribute to three different processes, namely oxidation, hydrolysis and pyrolysis [28]. Forms and products of the decomposition of the Nomex are determined by aging conditions and the aging time. In the aging test, the hydrolysis is the major way of the decomposition because the aging temperature is lower. There are three hydrolytic ways of Nomex as depicted in Figure 15. Hence, the excess of moisture or water in solid dielectrics of electrical equipment is always considered to be a important reason that can cause the insulation degradation and will lead to insulation faults under high electrical stress. The electrical properties of the solid polymer, which are influenced by a number of different factors, such as temperature, humidity, test duration and applied voltage type etc., decline with the degradation of insulation materials. In fact, the PD signal of the insulation can characterize the performance of the insulation system. The PD signals of the turn-to-turn insulation specimens obtained in this paper show that the PD is a highly stochastic phenomenon and the aging state of insulation cannot be simply determined by its amplitude and times. As a result, it is necessary to extract the statistical characteristics of PD signals so as to provide the more information for the reliability estimation. In addition, it has been found by the aging test that the PD signals have four patterns during the aging time which are corresponding to four states of the insulation as shown in Figure 9. Since the local photomicrographs of the insulation material can reflect its aging degree in terms of the microscopic characteristics, the photomicrographs of the turn-to-turn insulation specimens were periodically obtained

(a)

(b)

(c) Figure 15. Ways and products of fiber decomposition by hydrolysis

6 CONCLUSION This paper proposes a method to evaluate reliability of the insulation system of a dry-type transformer based on DM, RM and EJM. The three models have been developed based on quality parameters (QPs). The method is illustrated by estimating the reliability of the insulation system of a mining flameproof dry-type transformer. The accelerated aging tests on the turn-to-turn insulation, i.e. Nomex paper specimens have been conducted. The DPs, PDs and PGs of the Nomex samples were measured periodically during aging test. The condition of insulation specimens can be observed by the parameters obtained from the aging test.

IEEE Transactions on Dielectrics and Electrical Insulation

Vol. 20, No. 6; December 2013

The DP of the aged specimens dramatically decline with time, which can be also found in the cellulose paper. Hence the decomposition of the Nomex should have the similar characteristics to the cellulose, which confirms that the DP can reflect the degradation of the Nomex paper and provides theoretical support for the degradation model. Four PD patterns corresponding to the insulation conditions were established from the PD signals. It was found that there is not a regular relationship between the amplitude of PD signals and the insulation condition, the number of PD signals as well. Hence the statistical characteristics were extracted and selected as the inputs of the Regression model. The photomicrographs of specimens in different aging time indicate that the microscopic characteristic of the Nomex follows a special law. The degradation of the Nomex is generated from the regions where fibers interlace together and develop along the fibers. The parameters of the proposed models were calculated using the QPs. Ten groups of Nomex paper specimens were tested and estimated to simulate the turn-to-turn insulation of flameproof dry-type transformers. The estimation results demonstrate that the reliability of the specimens is closely related to the remaining life of the transformers, which verified the effectiveness of the proposed methods and models. The different QPs measured during the aging test can be used as the benchmarks to predict life of the same insulation type of transformers. The real time parameters of the online transformers from the monitoring system and obtained during the maintenance are required to be compared with the standard parameters from the aging test to predict the remaining life.

[6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

[17] [18] [19] [20]

ACKNOWLEDGMENT The authors would like to thank the Major science and technology projects in Shanxi Province and the State international technology cooperation projects of the Ministry of Science and Technology of the People’s Republic of China for the financial support.

[21]

[22] [23]

REFERENCES [1] [2] [3]

[4] [5]

L. W. Pierce, “Thermal considerations in specifying dry-type transformers”, IEEE Trans. Ind. Appl., Vol. 30, No. 4, pp. 1090-1098, 1994. P. A. A. F. Wouters, A. van Schijndel, and J. M. Wetzer, “Remaining lifetime modeling of power transformers: individual assets and fleets”, IEEE Electr. Insul. Mag., Vol. 27, No. 3, pp. 45-51, 2011. B. Gorgan, P. V. Notingher, J. M. Wetzer, H. F. A. Verhaart, and P. A. A. F. Wouters, “Calculation of the remaining lifetime of power transformers paper insulation”, IEEE 13th Int’l. Conf. Optimization of Electrical and Electronic Equipment (OPTIM), pp. 293-300, 2012. H. C. Stewart, L. C. Whitman, and A. L. Scheideler. “Aging evaluation of dry-type transformer insulation system”, IEEE Trans. Power App. Syst., Vol. 72, No. 2, pp. 267-277, 1953. T. R. Walters and A. L. Scheideler. “A study of models for use in evaluating dry-type transformer insulation systems”, IEEE Trans. Power App. Syst., Vol. 75, no. 3, pp. 520-527, 1956.

[24]

[25] [26] [27] [28]

2007

H. C. Stewart and L. C. Whitman. “Aging characteristics of dry-type transformer insulation at high temperature”, Trans. Amer. Inst. Electr. Eng., Trans., Vol. 67, No. 2, pp. 1600-1607, 1948. H. C. Stewart and L. C. Whitman. “Hot-spot temperatures in dry-type transformer windings”, Trans. Amer. Inst. Electr. Eng., Trans., Vol. 63, No. 10, pp. 763-768, 1944. L. W. Pierce. “Prediction hottest spot temperatures in ventilated dry type transformer windings”, IEEE Trans. Power Del., Vol.9, No. 2, pp. 11601172, 1994. A. Van Schijndel, J. M. Wetzer, and P. A. A. F. Wouters, “Reliability estimation of paper insulated components”, IEEE Conf. Electr. Insul. Dielectr. Phenomena (CEIDP), pp.17-20, 2007. G. C. Stone, “A perspective on online partial discharge monitoring for assessment of the condition of rotating machine stator winding insulation”, IEEE Electr. Insul. Mag., Vol. 28, No. 5, pp.8-13, 2012. A. Van Schijndel, Power Transformer Reliability Modeling, Ph.D. Dissertation, Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven, the Netherlands, 2010. A. Van Schijndel, P. A. A. F. Wouters, E. F. Steennis, and J. M. Wetzer, “Approach for an integral power transformer reliability model”, Euro. Trans. Electr. Power, Vol. 22, pp.491-503, 2012. M. Farahani, E. Gockenbach, and H. Borsi, “Behavior of machine insulation systems subjected to accelerated thermal aging test”, IEEE Trans. Dielectr. Electr. Insul., Vol. 17, No. 5, pp.1364-1372, 2010. IEEE Guide for determination of hottest-spot temperature in dry-type transformers, pp. 6, 2000. G. Huang, P. Saratchandran, and N. Sundararajan, “A generalized growing and pruning RBF (GGAP-RBF) neural network for function approximation”, IEEE Trans. Neural Networks, Vol. 16, No. 1, pp. 1929-1934, 2005. D. Evagorou, A. Kyprianou, P. L. Lewin, A. Stavrou, V. Efthymiou, A. C. Metaxas, and G. E. Georghiou “Feature extraction of partial discharge signals using the wavelet packet transform and classification with a probabilistic neural network”, IET Sci. Measurement Technology (UK), Vol. 4, No. 3, pp. 177-192, 2010. C.-C. Kuo, “Artificial identification system for transformer insulation aging”, Expert Systems with Applications, Vol. 37, pp. 4190-4197, 2010. IEEE Standard Test procedure for thermal evaluation of insulation systems for ventilated dry-type power and distribution transformers, 1986. R. Morin, R. Bartnikas, and P. Ménard, “A three-phase multi-stress accelerated electrical aging test facility for stator bars”, IEEE Trans. Enegry Conversion, Vol. 15, No. 2, pp. 149-156, 2000. L. Yao, C. Lee, and J. Kim, “Fabrication of electrospun meta-aramid nanofibers in different solvent systems”, Fibers and Polymers, Vol. 11, No. 7, pp. 1032-1040, 2010. D. Harwood, H. Aoki, Y. D. Lee, J. F. Fellers, and J. L. White. “Solution and Rheological Properties of Poly(m-phenyleneisophthalamide) in Dimethylacetamide/LiCl”, J. Appl. Polymer Sci., Vol. 23, pp. 2155-2168, 1979. IEEE Recommended practice for the detection of partial discharge and the measurement of apparent charge in dry-type transformers, 1991. H.-Z. Ding and B. R. Varlow, “Raman spectroscopy—a technique to assess the residual stress in fiber-reinforced polymeric insulation materials”, IEEE Electr. Insul. Mag., Vol. 20, No. 1, pp. 5-13, 2004. H.-Z. Ding and B. R. Varlow, “Thermodynamic model for electrical tree propagation kinetics in combined electrical and mechanical Stresses” IEEE Trans. Dielectr. Electr. Insul., Vol. 12, No. 1, pp. 8189, 2005. H.-Z. Ding, X. -S. Xing and H.-S. Zhu, “A kinetic model of timedependent dielectric breakdown for polymers”, J. Phys. D: Appl. Phys., Vol. 27, pp. 591-595, 1994. P. Cygan and J. R. Laghari. “Models for Insulation Aging under Electrical and Thermal Multistress”, IEEE Trans. Electr. Insul. Vol.25. No. 5, 1990, pp. 923-934. L. E. Lundgaard, W. Hansen, D. Linhjell, and T. J. Painter, “Aging of oil-impregnated paper in power transformers”, IEEE Trans. Power Del., Vol. 19, No. 1, pp. 230-239, 2004. S. Villar-Rodil, A. Martı́nez-Alonso, and J. M. D. Tascón, “Studies on pyrolysis of Nomex polyaramid fibers”, J. Anal. Appl. Pyrolysis, Vol. 58 –59, pp. 105–115, 2001.

2008

M. Wen et al.: Reliability Assessment of Insulation System for Dry Type Transformers Minmin Wen obtained the B.Sc. degree in Taiyuan University of Technology, China, in 2008. She has been studying at Taiyuan University of Technology for the M.Sc. and Ph.D. degrees since 2008. She works on condition monitoring and diagnoses of high voltage insulation for power apparatus.

Jiancheng Song received the B.Sc. degree from Taiyuan University of Technology, China, in 1982, the M.Sc. degree from Newcastle University, England, in 1987, respectively and the Ph.D. degree from Xian Jiaotong University, China, in 1999. Currently, he is a professor of the College of Electrical and Power Engineering at Taiyuan University of Technology. He has experience in the field of condition assessment, remaining life assessment and intellectual automation technology. He has performed a number of electrical failure investigations about coal mine. He has presented a number of technical and scientific papers at international conferences and seminars. Yuan Song obtained the B.Sc. and M. Sc. Degrees in Taiyuan University of Technology, China, in 2008 and 2011, respectively. He has been studying at Taiyuan University of Technology for the Ph.D. degree since 2011. He works on intellectual electric apparatus and process automation technology.

Yuan Liu obtained the B.Sc. and M. Sc. degrees both in Taiyuan University of Technology, China, in 2008 and 2011 respectively. She has been studying at Taiyuan University of Technology for the Ph.D. degree since 2011. She works on Power electronic transformation and automatic control technology.

Chuanyang Li received the B.S. degree from Taiyuan University of Technology, China, in 2011. He has been studying in Taiyuan University of Technology for the M.S. degree since 2011. His main research interest is condition monitoring and PD pattern recognition for HV motors.

Peng Wang (M’00) received the B.Sc. degree from Xian Jiaotong University, China, in 1978, the M.Sc. degree from Taiyuan University of Technology, China, in 1987, and the M.Sc. and Ph.D. degrees from the University of Saskatchewan, Canada, in 1995 and 1998, respectively. Currently, he is a professor at Taiyuan University of Technology, China and associate professor at Nanyang Technological University, Singapore.