AProbabilistic Approach forAnimal-Caused Outages inOverhead Distribution Systems
Swati Sahai · 2006
Faults caused byanimals areasignificant factor in Utilities usehistorical outage datatocompute reliability thereliability ofdistribution systems. Examination ofhistorical indices andmakeoperation andmaintenance decisions to dataonoutages recorded byautility inKansas showed that the improve reliability ofthesystem. Increased computational occurrences ofanimal-caused faults, mostofwhich arecaused by powerhasallowed useofproactive methodsoffinding squirrels, aredependent onweather conditions andthetimeof distribution systemreliability. Thesemethodsallow theyear.Itwasobserved thatthemajority ofanimal-caused identification of lessreliable partsof thesystemfor outages takeplace onfair weather days, whicharethedayswith temperature between 40degrees F and85degrees F withno improvements inadvance rather thanwaiting forthefault to other weather activity. Also, thebreeding cycle andbehavioral takeplace.Thesefactors havebeenaddressed inseveral patterns ofsquirrels overtheyearhasastrong correlation withpapers. Specifically, vegetation andtreerelated failures are theanimal-caused outages indifferent monthsoftheyear.discussed in(4-6),lightning related failures arediscussed in Population density ofsquirrels will haveaninfluence onthe (7,8),andanimal related failures arediscussed in(9). numberofoutages caused byanimals inoverhead distribution Inthispaper, failures inoverhead distribution systems systems, butreliable dataonanimal population density could not caused byanimals areexamined. Occurrence offaults dueto befound, thusan adequate direct relationship couldnotbe animals indifferent weather conditions aswellasatdifferent established. Inthis paper, aBayesian modelisconstructed for times oftheyear,areinvestigated. Basedonthis investigation, prediction ofanimal-caused outages inoverhead distribution systems given thetype ofmonthandthenumberoffair weathera mathematical causeandeffect relationship isdeveloped daysperweek.Themodeluseshistorical datafortraining between weather andbehavioral patterns ofanimals, and purposes andpredicts thenumberofanimal-caused outages per animal-caused outages. A Bayesian modelisbuilt torepresent weekgiven acombination ofmonthtypeandnumberoffairthecause-effect relationship andforprediction offaults dueto weather daysperweek.A confidence interval isbuilt aroundthe Ip p predicted values ofanimal-caused outages, andthepercentage of an theobserve numbeultsin'a perdall observed values lying within theconfidence limits arecomputedwithin theconfidence bounds, theutilities don'tneedany inorder toobtain accuracy ofthepredictions. Also, thetime-action. However, iftheobserved numberoffaults ishigher sequences ofmonthly averaged predictions arecompared withthantheupperlimit, theutilities needto do further theobserved values. Themodelisillustrated withseven years of examination totakecorrective actions. dataobtained foraservice territory ofautility inKansas.