IEEE Standard 1366 –Classifying (SAIDI, SAIFI, CAIDI) Into .

IEEE Standard 1366 – ClassifyingReliability (SAIDI, SAIFI, CAIDI) intoNormal, Major Event andCatastrophic DaysRich ChristieUniversity of WashingtonEE 500E/ME 523October 11, 2012October 11, 2012Catastrophic Days1

Overview IEEE Standard 1366 Major Event Days Catastrophic Days– Heuristic– Box and Whiskers– Robust EstimationOctober 11, 2012Catastrophic Days2

IEEE Standard 1366 Need to compare utilities– If regulators compare utilities, the comparison shouldbe as equitable as possible First issued in 1998, then 2001, 2003 Product of the IEEE Distribution Design WorkingGroupOctober 11, 2012Catastrophic Days3

IEEE Standard 1366 Defines 12 indices– SAIFI, SAIDI, CAIDI, CTAIDI, CAIFI, ASAI, CEMIn, ASIFI,ASIDI, MAIFI, MAIFIE, CEMSMIn Defines how indices are calculated– Standardizes Computation– How many outages is a recloser event?– How long before an outage is sustained?– What is a customer?October 11, 2012Catastrophic Days4

IEEE Standard 1366 Defines how to separate reliability into normaland major event reliability– Major Event Days (MEDs)October 11, 2012Catastrophic Days5

Major Event Days Some days, reliability ri is a whole lot worsethan other days– ri is SAIDI/day, actually unreliabilty Usual cause is severe weather: hurricanes,windstorms, tornadoes, earthquakes, icestorms, rolling blackouts, terrorist attacks These are Major Event Days (MED) Problem: Exactly which days are MED?October 11, 2012Catastrophic Days6

Phenomenological MEDsDesignates a catastrophic event which exceeds reasonable designor operational limits of the electric power system and during whichat least 10% of the customers within an operating area experiencea sustained interruption during a 24 hour period. In 1366‐1998Reflected broad range of existing practiceSubjective: “catastrophic,” “reasonable”Inequitable (10% criterion)No one design limitNo standard event typesOctober 11, 2012Catastrophic Days7

10% CriterionABSame geographic phenomenon (e.g. storm track)affects more than 10% of B, less than 10% of A.Should be a major event for both, or neither inequitable to larger utility.October 11, 2012Catastrophic Days8

Frequency Criteria Agree on average frequency of MEDs, e.g. “onaverage, 3 MEDs/year”– Quantitative– Equitable to different sized utilities– Easy to understand– Translates to probability theory, e.g. “3σ”– Consistent with design criteria (withstand 1 in Nyear events)October 11, 2012Catastrophic Days9

Probability of Occurrence Frequency of occurrence f is probability ofoccurrence pfp 365October 11, 2012Catastrophic Days10

Reliability Threshold TMED Find threshold TMED from probability p andprobability distributionpdff(ri)p(ri TMED)TMEDDaily Reliability ri MEDs are days with reliability ri TMEDOctober 11, 2012Catastrophic Days11

Probability Distribution 3σ only works for Gaussian (Normal) distribution Examine distribution of daily SAIDI:40Bin CountBin Count1000203 yrs of utility data00010r, SAIDI/day(a)2001020r, SAIDI/day(b) Not Normal!October 11, 2012Catastrophic Days12

Log‐Normal Natural logs of the sample data are normallydistributed Sample data itself is skew5 years of data, anonymous utility U2October 11, 2012Catastrophic Days13

Log‐Normal Best fit of distributions tests Computationally tractable– Pragmatically important that method beaccessible to typical utility engineer Weak theoretical reasons to go with log‐normal– Loosely, normal process with lower limit has log‐normal distributionOctober 11, 2012Catastrophic Days14

Log‐Normal Not completely Log‐Normal – note ends5 years of data, anonymous utility U2October 11, 2012Catastrophic Days15

Finding TMED Five years of data Find average and standard deviation ofdistribution ofn ln of daily SAIDI1 ln ri ni 11 n2 lnr in 1 i 1 Compute TMEDTMED exp( 2.5 )October 11, 2012Catastrophic Days16

Finding TMED Why 2.5 (giving the “2.5β Method”)? Theoretical number of MEDs per year: 2.43 Real reason is that the Working Groupmembers liked the results using 2.5 betterthan 2 or 3. Liked means:– Does not identify too many or too few MEDs– Identifies days that ought to be MEDs as MEDs– Better MED consistency among subdivisionsOctober 11, 2012Catastrophic Days17

2.5β Method Method still subjective – but less so Adopted in P1366‐2001Anonymous utility U29October 11, 2012Catastrophic Days18

Catastrophic Days Some days are really, really worse than otherdays – catastrophic days 2.5β removes these days from normalreliability But catastrophic days affect the value of TMEDfor the next five years This affects the number of MEDs identified This affects normal reliability valuesOctober 11, 2012Catastrophic Days19

Catastrophic DaysU29 had a possiblecatastrophic day in 1998October 11, 2012Catastrophic Days20

Catastrophic DaysYR9798990001020304October 11, 0Catastrophic OCATMEDS614932882614103398221

Catastrophic Days What to do? Outlier removal problem– Identify outliers– Omit them from the TMED calculation How?– Heuristic (Xβ)– Box and Whiskers– Robust EstimationOctober 11, 2012Catastrophic Days22

Heuristic Work by Jim Bouford, TRC Engineers LLC A Catastrophic Day has SAIDI Xβ– X found heuristically 10 utility data sets with subjective “catastrophicdays” Vary X, examine identified catastrophic days X 4.14 gave good results X 4.15 or X 4.16 did not Clearly not a viable methodOctober 11, 2012Catastrophic Days23

Box and Whiskers Work by Heidemarie Caswell, Pacific Power Use Box and Whisker plot to identify outlyingCatastrophic DaysQ3 3 IQR3rd Quartile (Q3)Inter‐Quartile Range IQR Q3 – Q1Median1st Quartile (Q1)Q1 ‐ 3 IQROctober 11, 2012Catastrophic Days24

Box and Whiskers Tested on a dozen utility data sets Subjective assessment – unsatisfactory Why?– IQR is a robust estimator of standard deviation, β–.– Whiskers at– Given 4.14β, seems unlikely 4.725 would be betterOctober 11, 2012Catastrophic Days25

Robust Estimation Work by me Sample average and standard deviation areestimates of process average and standarddeviation There are other ways to estimate– Median estimates average ˆ ln rn / 2 – Difference of quartile values (Inter‐Quartile Range,IQR) estimates standard deviationIQR ln rn / 4 ln r3n / 4 October 11, 2012Catastrophic DaysIQRˆ 1.3526

Robust Estimation So, just use robust estimates andof α and βOctober 11, 2012Catastrophic Daysinstead27

Robust Estimation Example– Sample set 0.5, 2.0, 3.1, 3.9, 4.6, 5.4, 6.1, 6.9, 8.0,9.5 (artificial, normal)– Mean 5.0, robust estimate of mean 5.0– Standard deviation 2.76, robust estimate 2.81 With outlier – replace last sample by 100– Mean 14.1, robust estimate of mean 5.0– Standard deviation 30.3, robust estimate 2.81 Looks pretty good for the exampleOctober 11, 2012Catastrophic Days28

Robust Estimation More accurate when outliers are present Less accurate when outliers are not β2.151.98TMED10.97.59PARAMETERData from U2, which did not have a potential catastrophic day Working Group members did not like theroutine inaccuracyOctober 11, 2012Catastrophic Days29

Conclusions 2.5β does a pretty good job with catastrophicdays.– Utilities still want a method to identify them. No proposed method is subjectivelysatisfactory. The search continues.October 11, 2012Catastrophic Days30

Finding T MED Five years of data Find average and standard deviation of distribution of ln of daily SAIDI Compute T MED October 11, 2012 Catastrophic Days 16 n i r i n 1 ln 1 n i r i n 1 ln 2 1 1 T MED exp( 2.5 )