Cable_Ageing_Final
Cable_Ageing_Final
Implementation of Ageing Laws and Cable Models to Estimate Service Life for MV Cable
Designs using Laboratory Endurance Tests
Essay Wen SHU, Nigel HAMPTON, Josh PERKEL; NEETRAC, (USA), [email protected],
[email protected], [email protected]
Weibull Analysis. The Weibull distribution is widely used in shape parameter is assigned as 4 for PE based insulation
reliability and life data analysis due to its versatility and its and 2 for EPR insulation.
essentially graphical approach [16] and is recommended
by IEEE 1407 [2]. ANALYSIS
Examples of the Weibull analysis on the individual data set The approach used in this work was to collate the results
appear in Figure 2. This graph shows the analysis of 20 of many public domain ACLT tests to model the impact of
typical, out of the 124 collated, datasets that are tested at temperature, electrical stress and some elements of cable
different temperatures, voltages, length, insulation design. Algorithms were then constructed to scale test data
volumes, etc. The estimated correlation coefficients for the on short cores to long cables in service using Life
individual datasets indicate a good fit of the data to the Expansion and Reduction Factors.
Weibull distribution.
Volume and Stress Multipliers
ACLT is performed at discrete aging conditions. As an
example, ACLT testing at the average electrical stress of 2
kV/mm (Uo), 4 kV/mm (2Uo), 6 kV/mm (3 Uo) and 8 kV/mm
(4Uo) in the cable insulation for the standard cable
configuration. The average electrical stresses at field
operating conditions for many cable configurations are
different from any of the tested stresses. For example, the
average electrical stress in the cable insulation of a 15 kV
with 220 mil insulation thickness cable is 1.55 kV/mm. To
estimate cable life for any medium voltage class cable
configurations (i.e. cables operated at any given electrical
stress) operating at any conductor temperature between
ambient temperature to 90 °C or even 105 °C, a regression
model is needed to
Figure 2. Examples of Weibull analysis for the
individual dataset. • interpolate data at the region where test data is
available at discrete test conditions;
The scale parameter, α, gives an indication of the predicted • extrapolate data at the region beyond where test data
life and is the value of the measured variable at a is available at discrete test conditions (Figure 4).
probability of 63.2%. The expansion and reduction factors
will be applied to the scale parameters unless a change in
the failure mechanism is resulted from those factors.
Figure 3. Boxplot of the shape parameters by Figure 4. Illustration of using contour plot for data
insulation type – median given by central line within interpolation / extrapolation for XLPE data as.
box.
A regression model was constructed to predict cable life
The shape parameter, β, is the slope of the Weibull based on electrical stress and conductor temperature using
distribution curve. The value of the shape parameter the scale parameters calculated for XLPE insulation data.
indicates the characteristic failure mechanism. Figure 3 A statistically significant equation was developed,
shows the boxplot of the shape parameters of the collated
ACLT data by insulation type. Assuming that the failure 𝛼𝛼 𝑇𝑇 = 8445 − 988 ∙ 𝐸𝐸 − 91.4 ∙ 𝑇𝑇 + 10.85 ∙ 𝐸𝐸 ∙ 𝑇𝑇 [1]
modes for high molecular polyethylene (HMWPE), XLPE • E is the average electrical stress in the cable, kV/mm
and cross-linked polyethylene (TRXLPE) are the same, the • T is the conductor temperature, °C
shape parameters for these PE based insulation test data
are lumped together. It is noted that the shape para-meters The following equation is derived in order to start from given
for EPR insulation have a larger variation than those for PE ACLT data at a known testing condition (ET, TT).
based insulation. This is possibly due to the differences in 𝛼𝛼1 = 𝛼𝛼 𝑇𝑇 ∙ 𝑅𝑅(𝐸𝐸, 𝑇𝑇) [2]
chemical ingredient composition of EPR compound in its 𝛼𝛼(𝐸𝐸, 𝑇𝑇)
chemical ingredient compositions from manufacturer to 𝑅𝑅(𝐸𝐸, 𝑇𝑇) =
𝛼𝛼(𝐸𝐸𝑇𝑇 , 𝑇𝑇𝑇𝑇 )
manufacturer. Based on the results from Figure 3, the
• αT from ACLT data at a known testing condition (ET, TT) cable is plotted in Figure 6.
• α1 is the characteristic life after (E,T) transformation Jacket / Moisture Adjustment
Length Adjustment Upon research of public domain ACLT data, there is not
any external ACLT program that investigated the effects of
A typical ACLT test length is 15 ft. A significant portion of the cable jacket. A prior NEETRAC baseline project 97-409
the test length is submerged under the water and the rest focused on developing a new cable design aging protocol
above the water for termination purpose. This cable length and tested 20 cable design/test conditions with moisture-
is, however, not typical in field applications. A URD cable blocked conductor and an overall jacket [18]. These data
segment, which connects distribution transformers, is were analyzed using Weibull distribution and compared to
typically in the range of 100 ft. to 300 ft. A feeder cable, unjacketed data, as shown in Figure 7. A three-parameter
which emanates from substations to the distribution Weibull was used to fit the ACLT data of jacketed cables.
network, is typically in the range of 1000 ft. These lengths The adjustments for jacket/moisture effect were applied to
are significantly different from the tested cable length in the both the scale parameter and the shape parameter using
ACLT environment. The well-known enlargement law [17]
exercised in Equation 2 is thus applied to include cable 𝛼𝛼𝑆𝑆 = 𝛼𝛼3 ∙ 5.89 + 367 [4]
length effect on cable service.
𝛽𝛽𝑆𝑆 = 𝛽𝛽𝑇𝑇 + 2 [5]
𝜶𝜶𝟐𝟐 = 𝜶𝜶𝟏𝟏 ∙ 𝑅𝑅(𝐿𝐿) [3]
𝟏𝟏� • α3, βT are the Weibull parameters of unjacketed cable
𝒍𝒍 𝜷𝜷𝑻𝑻
with water-filled conductor strands
𝑅𝑅(𝐿𝐿) = � �
𝑳𝑳 • αS, βS are the Weibull parameter of jacketed cable with
• α2 is the characteristic life at the cable service length L blocked conductor strands
99
• α1 is the characteristic life at the tested cable length l
90
• R(L) is the cable length enlargement coefficient 80
70
• βT is the shape parameter of the tested cable core 60
50
40
(unjacketed cable with water-filled conductor strands) 30
20
for specific insulation type.
Percent
10
Since the standard ACLT test length is 15ft., the length 5
enlargement coefficient with reference to 15 ft. is commonly 3
2
used. As an example, the cable length enlargement
1
coefficient for a 250 ft. cable to a 15 ft. test length is 0.5 for
TRXLPE insulated cable and 0.25 for EPR insulated cable. Jacketed
N
Y
Insulation Volume Adjustment 0.1
0.01 0.1 1 10 1 00 1 000
Time to Failure (days) - Threshold
• αs, βs are the Weibull parameters derived from the 100 of these 300 ft. cables with this specific cable
acceleration/reduction adjustments configuration and operating conditions were installed in the
• p is tolerance of failure or unreliability level system at year 0,
• B(p) life is the age at which p% of the cables will
• 1 out of 100 would experience its first failure by year 10;
experience their first failure.
• 10 out of 100 would fail for their first time by year 16;
• 90 out of 100 would see first failure by year 25; and
CASE STUDY • The average life (B50, 50 out of 100 would fail by this
A few case studies show how the modeling may be used. age) of this cable is 21 years.
A summary of the tested case studies appears in Table 3.
These case studies also illustrate how the model may be Such a segment may well be repaired and remain in service
used. after the first failure. However, the reliability of this segment
will decline due to the continued ageing of the remaining
Table 3. A Summary of Tested Case Studies original cable, the introduction of one or more joints and
Case ACLT Service potentially deleterious effects of fault location. The jointing
process introduces potential week links and the joints
1 αT,βT 15kV, 1,000 kcmil themselves will age in service. Thus, the time in service is
@ ACLT (4,4) 100% XLPE wall, likely longer than estimated by this current modeling.
conditions from Jacketed Although it should be recognised that the segment is not
collated library 300 ft. (100 m) the same, after the repair, as the one that was installed.
Operated at 30 °C
Estimated service life – first failure:
B50 21 years B10 16 years
2 αT =260, βT =4 15kV, 1,000 kcmil
@ ACLT (4,4) 100% WTRXLPE wall,
conditions from Jacketed
user test data 300 ft. (100 m)
Operated at 30 °C
Estimated service life – first failure:
B50 35 years B10 26 years
3 αT,βT 35 kV, 1,000 kcmil
@ ACLT (4,3) 100% WTRXLPE wall
conditions from insulation, Jacketed
collated library 2,000 ft. (610 m)
operated at 60 °C
Estimated service life – first failure:
B50 7 years B10 5 years
Figure 8. Case 1 Weibull curves (adjusted scale and
4 αT,βT 15 kV, 350 kcmil
shape parameter) for life expansion factors. Curves
@ ACLT (4,3) Reduced wall (83%)
move right (longer life) as expansion factors applied.
conditions from EPR, Jacketed
collated library 300 ft. (100 m)
Operated at 75 °C
Estimated service life – first failure:
B50 4 years B10 3 years
5 αT=1019, βT=2 15 kV, 350 kcmil
@ ACLT (4,4) 100% EPR wall,
conditions from Jacketed
user test data 300 ft. (100 m)
Operated at 75 °C
Estimated service life – first failure:
B50 18 years B10 11 years
The first case study starts with ACLT test results for a
standard ACLT cable configuration at (4,4) test condition.
The interest is in using the composite ACLT core
experience to estimate the service life, for first failure, of a Figure 9. Case 1 Weibull curves (adjusted scale and
15 kV, 1,000 kcmil, 100% XLPE insulated jacketed cable shape parameter) for life reduction factors. Curves
operated at (1,1) condition with a service length of 300 ft. move left (shorter life) as reduction factors applied.
(experience shows that this is a typical length of cable
Thus the Life Statement from this case would be; ”10% of
between transformers in a residential sub division).
300 ft cable segments (15kV 1000kcmil jacketed XLPE
Weibull curves were generated based on the adjusted 100% wall) would be anticipated to experience their first
scale parameter and shape parameter and were displayed failure in service (operating at 30 °C) on or before age 16
separately by the expansion factors (Figure 8) and the years”.
reduction factors (Figure 9). The survival curve for the
service cable was also plotted in Figure 10. It shows that if