Documente Academic
Documente Profesional
Documente Cultură
29-31 0.50
32-35 0.60
36-40 0.80 )
MVA
(<2 5
k
ble r un
Three data pairs are required to solve for the parameters
Ca y T
ar
A, B, and C. The previous section has developed a condi-
im
Pr
tion ranking methodology that, by definition, results in
best, average, and worst condition scores of 0, ½, and 1, 0.001
respectively. Therefore, three natural data pairs correspond 0 0.2 0.4 0.6 0.8 1
to λ(0), λ(½), and λ(1). λ(½) can be approximated by tak- Condition (p.u.)
ing the average failure rate across many components or by Figure 1. Selected Equipment Failure Rate Functions
using average failure rates documented in relevant litera- Failure rate graphs for some of the equipment in Table 3
ture. λ(0) and λ(1) are more difficult to determine, but can are shown in Figure 1. These are simply plots of Eq. 2 us-
be derived through benchmarking, statistical analysis, or ing the stated A, B, and C parameters the displayed equip-
heuristics. Given these three values, function parameters ment. It is interesting to see that the range of failure rates
are determined as follows: of certain types of equipment is large, while other types
have a more moderate range. This reflects the ranges found
A=
[λ (½ ) − λ (0)]2 in a broad literature search which forms the basis of Table
3.
λ (1) − 2λ (½ ) + λ (0)
⎛ λ (½ ) + A − λ (0 ) ⎞ IV. MODEL CALIBRATION
B = 2 ln ⎜ ⎟ (3)
⎝ A ⎠ After creating a system reliability model, it is desirable
C = λ (0) − A to adjust component reliability data so that predicted sys-
tem reliability is equal to historical system reliability [11].
This process is called model calibration, and can be gener-
A detailed benchmarking of equipment failure rates is alized as the identification of a set of parameters that mini-
found in [18]. These results document low, typical, and mize an error function.
high failure rates corresponding to system averages across Traditionally, reliability parameters (such as equipment
a variety of systems. Assuming that (1) best-condition failure rates) either remained uncalibrated or were adjusted
equipment have failure rates half that of best system aver- based on average system reliability. For example, it may be
ages, (2) average-condition equipment have failure rates of known that an analysis area has an average of 1.2 interrup-
typical system averages, and (3) worst-condition equip- tions per customer per year. Based on this number, failure
ment have failure rates twice that of best system averages, rates can be adjusted until the predicted average number of
parameters for a variety of equipment are shown in Table customer interruptions is equal to this historical value. Af-
3. These parameters, based on historical failure studies ter failure rates are calibrated, switching and repair times
such as [14], are useful in the absence of system specific can be adjusted until predicted average interruption dura-
data, but should be viewed as initial conditions for calibra- tion is also equal to historical values.
tion, which is discussed in the next section. Calibrating based on system averages is useful, but does
not ensure that the predicted distribution of customer inter-
ruptions is equal to the historical distribution. That is, it
does not ensure that either the most or least reliable cus-
tomers are accurately represented – only that the average
across all customers reflects history. This is a subtle but
important point; since customer satisfaction is largely de-
termined by customers receiving below-average reliability,
calibration of reliability distribution is arguably more im-
portant than calibration of average reliability.
A system model with homogeneous failure rates will
produce a distribution of expected customer reliability lev-
els (e.g., customers close to the substation will generally
have better reliability than those at the end of the feeder).
If components on this same system are assigned random
failure rates such that average system reliability remains
the same, the variance of expected customer reliability will
tend to increase. That is, the best customers will tend to get
better, the worst customers will tend to get worse, and
fewer customers can expect average reliability.
The distribution of expected customer reliability is criti-
cal to customer satisfaction and should, if possible, be cali-
brated to historical data. A practical way to accomplish this
objective is to calibrate condition-mapping parameters so
that a distribution-based error function is minimized. Such
an error function can be based on one of three levels of
granularity: (1) individual customer reliability, (2) histo-
grams of customer reliability, or (3) statistical measures of
customer reliability.
An error function can be defined based on the difference
between each customer’s historical versus predicted reli-
ability. This approach calibrates reliability to the customer
level and utilizes historical data at the finest possible
granu-
Table 3. Representative Failure Rate Model Parameters (λ values in failures per year)
Description λ(0) λ(½) λ(1) A B C
Overhead Equipment
Overhead Lines
Primary Trunk* 0.0100 0.100 0.600 0.01976 3.4295969 -0.009756098
Lateral Tap* 0.0100 0.160 0.600 0.07759 2.1522789 -0.067586207
Secondary & Service Drop* 0.0100 0.088 0.600 0.01402 3.7632316 -0.004018433
Pole Mounted Transformer 0.0020 0.010 0.030 0.00533 1.8325815 -0.003333333
Disconnect Switch 0.0020 0.014 0.280 0.00057 6.1971793 0.001433071
Fuse Cutout 0.0020 0.009 0.060 0.00111 3.9718310 0.000886364
Line Recloser 0.0025 0.015 0.060 0.00481 2.5618677 -0.002307692
Shunt Capacitor 0.0055 0.020 0.170 0.00155 4.6729733 0.003948339
Voltage Regulator 0.0050 0.029 0.200 0.00392 3.9272195 0.001081633
Underground Equipment
Underground Cable
Primary Cable* 0.0015 0.070 1.174 0.00453 5.5597230 -0.003031386
Secondary Cable* 0.0025 0.100 0.300 0.09274 1.4369300 -0.090243902
Elbow Connectors 3.E-05 6.E-04 0.002 0.00039 1.7971823 -0.000361446
Cable Splices and Joints 3.E-05 0.030 0.318 0.00348 4.5255272 -0.003450994
Padmount Transformers 0.0005 0.010 0.100 0.00112 4.4970357 -0.000621118
Padmount Switches 0.0005 0.003 0.010 0.00139 2.0592388 -0.000888889
AIS Substation Equipment
Power Transformers
Less than 25 MVA 0.0075 0.040 0.140 0.01565 2.2478602 -0.008148148
Bigger than 25 MVA 0.0050 0.030 0.120 0.00962 2.5618677 -0.004615385
Circuit Breakers 0.0005 0.010 0.060 0.00223 3.3214624 -0.001728395
Disconnect Switches 0.0020 0.010 0.320 0.00021 7.3142615 0.001788079
Instrument Transformers 0.0000 0.010 0.060 0.00250 3.2188758 -0.002500000
Air Insulated Busbar 0.0005 0.010 0.076 0.00160 3.8767259 -0.001097345
GIS Substation Equipment
GIS Bay (before 1985) 0.0003 0.002 0.030 0.00011 5.6031525 0.000190114
GIS Bay (after 1985) 0.0002 0.001 0.018 0.00004 6.1127138 0.000160494
* Line and cable failure rates are per circuit mile
larity. However, historical customer reliability is stochastic Where n is the number of bins, h is the historical bin
in nature and will vary naturally from year to year. An er- value, and p is the predicted bin value. Using the chi-
ror function can be defined based on the difference be- squared error is attractive since it emphasized the distribu-
tween each customer’s historical versus predicted reliabil- tion of expected customer reliability which is strongly cor-
ity. This approach calibrates reliability to the customer related to customer satisfaction. Histograms will vary sto-
level and utilizes historical data at the finest possible chastically from year to year, but the large number of cus-
granularity. However, historical customer reliability is sto- tomers in typical calibration areas prevent this from be-
chastic in nature and will vary naturally from year to year. coming a major concern.
This is especially problematic with frequency measures. Last, an error function can be based on statistical meas-
Although customers on average may experience 1 interrup- ures such as mean value (µ) and standard deviation (σ).
tion per year, a large number of customers will not experi- The error function will typically consist of a weighted sum
ence any interruptions in a given year. Calibrating these similar to the following:
customers to historical data is misleading, making about 10
years of historical data for each customer desirable. Unfor- Error = α (µ − µ ') + β (σ − σ ')
2 2
(5)
tunately, most feeders change enough over 10 years to
make this method impractical.
An error function can also use a histogram of customer Unlike the χ2 error, this function allows relative weights
interruptions as its basis. The historical histogram could be to be assigned to mean and variance discrepancies (α and
compared to the predicted histogram and parameters ad- β). For example, a relatively large a value will ensure that
justed to minimize the chi-squared error (χ2): predicted average customer reliability reflects historical
average customer reliability while allowing relatively large
mismatches in standard deviation.
∑ (h h− p )
n Once an error measure is defined, failure rate model pa-
χ2 = i i
(4) rameters can be adjusted so that error is minimized. Since
i =1 i this process is over determined, the authors suggest using
Table 3 for initial parameter values and using gradient de-
scent or hill climbing techniques for parameter adjustment. 50
Calibration is computationally intensive since error sensi- Historical
tivity to parameters must be computed by actual parameter
40 Uncalibrated
perturbation, but calibration need only be performed once.
Calibrated
% of Customers
V. APPLICATION TO TEST SYSTEM 30
The methodologies described in the previous two sec-
tions have been applied to a test system derived from an
actual overhead distribution system in the Southern U.S. 20
This model consists of three substations, 13 feeders, 130
miles of exposure, and a peak load of 100 MVA serving
10
13,000 customers. The analytical model consists of 4100
components.
Customer historical failures are computed from four- 0
year historical averages. Equipment condition for this sys- 0.0 0.5 1.0 1.5 2.0 2.5 3.0
tem was not available, and was therefore assigned for ran- Inte rruptions (per year)
domly for individual components based on a normal distri-
bution with a 0.5 mean and a 0.2 standard deviation. Figure 2. Historical Versus Predicted Customer Interruptions
Calibration for this test system is performed based on
the chi-squared error of customer interruptions. Initial fail-
ure rates for all components are assigned based on λ(½)
values in Table 3, and initial failure rates are computed
based on condition and the parameters in Table 3. Calibra-
tion is performed by a variable-step local search that guar-
antees local optimality.
A summary of calibration results for overhead lines is
shown in Figure 2, and a visualization of calibrated results
in shown in Figure 3. The shape of the uncalibrated histo-
gram is similar to the historical histogram, but with a mean
and mode worse than historical values. After calibration,
the modes align, but the predicted histogram retains a
slightly smaller variance. In fact, the historical histogram is
subject to stochastic variance, and the inability of the ex-
pected value calibration to match this variance is immate-
rial and perhaps beneficial.
Uncalibrated and calibrated failure rate parameters are
shown in Table 4, and corresponding failure rate functions
are shown in Figure 4. In effect, the calibration for this
system did not change the failure rates for lines with good Figure 3. Visualization of Calibrated Results
condition (less than 0.2), but drastically reduced the failure
rates for lines with worse-than-average condition (greater 0.7
than 0.5). These results are not unexpected, since this par- Uncalibrated
ticular service territory is relatively homogeneous in both 0.6
Calibrated
Failure Rate (/yr)
0.2
VI. CONCLUSIONS
Equipment failure rate models are required for electric
utilities to plan, engineer, and operate their system at the
highest levels of reliability for the lowest possible cost.
Detailed models based on historical data and statistical
regression are not feasible at the present time, but this pa-
per presents an interpolation method based on normalized
condition scores and best/average/worst condition assump-
tions.
The equipment failure rate model developed in this pa-
per allows condition heterogeneity to be reflected in
equipment failure rates. Doing so more accurately reflects
component criticality in system models, and allows the
distribution of customer reliability to be more accurately
reflected. Further, a calibration method has been presented
that allows condition-mapping parameters to be tuned so
that the predicted distribution of reliability matches the
historical distribution of reliability. Finally, the use of this
condition-based approach allows the impact of mainte-
nance activities on condition to be anticipated and reflected
in system models, enabling the efficacy of maintenance
budgets to be compared with capital and operational budg-
ets.
This model is heuristic by nature, but adds a fundamen-
tal level of richness and usefulness to reliability modeling,
especially when parameters are calibrated to historical
data. In the short run, gathering more detailed information
on equipment failure rates and condition will strengthen
this approach. In the long run, this same information can
ultimately be used to develop explicit failure rate models
that eliminate the normalized condition assessment re-
quirement.
VII. REFERENCES
[1] EPRI Report EL-2018, Development of Distribution Reliability and
Risk Analysis Models, Aug. 1981.
[2] S. R. Gilligan, “A Method for Estimating the Reliability of Distribu-
tion Circuits,” IEEE Transactions on Power Delivery, Vol. 7, No. 2,
April 1992, pp. 694-698.
[4] R.E. Brown, S. Gupta, S.S. Venkata, R.D. Christie, and R. Fletcher,
“Distribution System Reliability Assessment Using Hierarchical
Markov Modeling,” IEEE Transactions on Power Delivery, Vol. 11,
No. 4, Oct., 1996, pp. 1929-1934.
[5] Y-Y Hsu, L-M Chen, J-L Chen, et al., “Application of a Microcom-
puter-Based Database Management System to Distribution System Re-
liability Evaluation,” IEEE Transactions on Power Delivery, Vol. 5,
No. 1, Jan. 1990, pp. 343-350.
[6] C.M. Warren, “The Effect of Reducing Momentary Outages on Distri-
bution Reliability Indices,” IEEE Transactions on Power Delivery,
Vol. 7, No. 3, July, 1992, pp. 1610-1615.
[7] R. Brown, S. Gupta, S.S. Venkata, R.D. Christie, and R. Fletcher,
‘Distribution System Reliability Assessment: Momentary Interruptions
and Storms,’ IEEE PES Summer Meeting, Denver, CO, June, 1996.
[8] R. E. Brown and J. J. Burke, “Managing the Risk of Performance
Based Rates,” IEEE Transactions on Power Systems, Vol. 15, No. 2,
May 2000, pp. 893-898.
[9] L. V. Trussell, “Engineering Analysis in GIS,” DistribuTECH Confer-
ence, Miami, FL, Feb. 2002.
[10] R. E. Brown and M. M. Marshall, “Budget-Constrained Planning to
Optimize Power System Reliability,” IEEE Transactions on Power
Systems, Vol. 15, No. 2, May 2000, pp. 887-892.
[11] R. E. Brown, J. R. Ochoa, “Distribution System Reliability: Default
Data and Model Validation,” IEEE Transactions on Power Systems,
Vol. 13, No. 2, May 1998, pp. 704-709.
[12] M. A. Rios, D. S. Kirschen, D. Jayaweera, D. P. Nedic, and R. N.
Allan, “Value of security: modeling time-dependent phenomena and
weather conditions,” IEEE Transactions on Power Systems, Vol. 17,
No. 3 , Aug 2002, pp. 543 –548.
[13] R. N. Allen, R. Billinton, I. Sjarief, L. Goel, and K. S. So, “A Reliabil-
ity Test System for Educational Purposes - Basic Distribution System
Data and Results,” IEEE Transactions on Power Systems, Vol. 6, No.
2, May 1991.
[14] R. M. Bucci, R. V. Rebbapragada, A. J. McElroy, E. A. Chebli and S.
Driller, “Failure Predic-tion of Underground Distribution Feeder Ca-
bles,” IEEE Transactions on Power Delivery, Vol. 9, No. 4, Oct.
1994, pp. 1943-1955.
[15] D. T. Radmer, P. A. Kuntz, R. D. Christie, S. S. Venkata, and R. H.
Fletcher, “Predicting vegetation-related failure rates for overhead dis-
tribution feeders,” IEEE Transactions on Power Delivery, Vol. 17,
No. 4, Oct. 2002, pp. 1170-1175.
[16] S. Gupta, A. Pahwa, R. E. Brown and S. Das, “A Fuzzy Model for
Overhead Distribution Feeders Failure Rates,” NAPS 2002: 34th An-
nual North American Power Symposium, Tempe, AZ, Oct. 2002.
[17] J. B. Bowles, “Commentary-caution: constant failure-rate models may
be hazardous to your design,” IEEE Transactions of Reliability, Vol.
51, No. 3, Sept. 2002, pp. 375-377.
[18] R. E. Brown, Electric Power Distribution Reliability, Marcel Dekker,
Inc., 2002.
VIII.BIOGRAPHIES
Richard E. Brown is a principal consultant with KEMA, and specializes in
distribution reliability and asset management. He is the author or co-author
of more than 50 technical papers and the book Electric Power Distribution
Reliability. Dr. Brown received his BSEE, MSEE, and PhD from the Uni-
versity of Washington and his MBA from the University of North Carolina
at Chapel Hill. He is a registered professional engineer, and can be reached
at rebrown@kema.com.