Integration of Asset and Outage Management Tasks for Distribution Systems

Transcription

1 1 Integraton of Asset and Outage Management Tasks for Dstrbuton Systems Y. Dong, Student Member, IEEE, V. Aravnthan, Student Member, IEEE, M. Kezunovc, Fellow, IEEE, and W. Jewell, Fellow, IEEE Abstract Integraton of asset management and outage management tasks for dstrbuton system s proposed and dscussed. The necessty of ntegraton s presented, followed by a descrpton of the concept of ntegraton. The optmzaton of asset and outage management tasks based on the ntegrated processng and evaluaton of the nfluence of optmzaton on the cost of system outage s elaborated. Potental beneft of ntegraton n dstrbuton system s analyzed n terms of system relablty and return of nvestment for utltes. Index Terms outage management, asset management, dstrbuton system, relablty, rsk-based assessment F I. ITRODUCTIO aults n dstrbuton system may cause nterrupton of power supply to customers. Snce dstrbuton systems n general encounter hgh frequency of faults caused by weather, component wear and other reasons, the need to reduce outage tme caused by faults s requred for several reasons.: a) better servce to customers. Customers requrement on the qualty of servce s constantly growng. As an example, senstve loads n modern ndustry such as chp manufacture and ore smelter are very senstve to nterruptons n power supply. The consequence of falure s more severe nowadays than a decade before; b) return on nvestment for utlty shareholders. The most drect mpact of faults on the proft s the loss n customer bllng, as well as mantenance expense. Relablty ndces defned n IEEE Standard 1366 are used to evaluate the mpact of faults on power dstrbuton performance [1]. The System Average Interrupton Duraton Index (SAIDI) and System Average Interrupton Frequency Index (SAIFI) are two most wdely used ndces. The lower the value of SAIDI and SAIFI, the better the performance n terms of relablty. Accordng to a survey done by the IEEE Workng group on dstrbuton relablty, n the year 2007, the average SAIDI derved from SAIDIs provded by 61 Ths work was supported by Power Systems Engneerng Research Center (PSerc). Project o: T-36. Y. Dong s wth the Department of Electrcal and Computer Engneerng, Texas A&M Unversty, College Staton, TX , USA (e-mal: dongyma@neo.tamu.edu). V. Aravnthan s wth the Department of Electrcal and Computer Engneerng, Wchta State Unversty, Wchta, KA USA (e-mal: vxaravnthan@wchta.edu). M. Kezunovc s wth the Department of Electrcal Engneerng, Texas A&M Unversty, College Staton, TX USA (e-mal: kezunov@ece.tamu.edu). W. Jewell s wth the Department of Electrcal Engneerng, Wchta State Unversty, Wchta, KA USA (e-mal: ward.jewell@wchta.edu). utltes s mn/(customer*year), and the average SAIFI s 1.71 /(customer*year) [2]. Currently the mprovement n dstrbuton performance s hampered by four major defects: a) lack of data. Besde voltage and current measurement at substatons, few montorng devces for measurements are nstalled n a dstrbuton system; b) agng of equpment. Most of the prmary equpment nstalled n the USA dstrbuton system s pretty old, n some nstances over years. c) neffectve processng of faults and mantenance schedulng caused by the lack of data. The fault locaton s currently based on trouble calls and manual swtchng [3] whle mantenance s performed ether wth a run-to-falure strategy or wth a fxed ahead-of-the-tme planned schedule [4], whch does not requre operatonal data; d) ndependent plannng and operaton of asset and outage management. Those two functons are planned ndependently even though the equpment that may record and collect relevant data from the fled maybe common to both applcatons.. Technologes have been proposed to reduce the frequency and duraton of faults. For outage management, effort has been made to better process the trouble calls [5], supplement nformaton from trouble calls wth AMR system and other sources [6], and to nvestgate varous methods to locate faults [7]-[9]. For asset management, condton-based mantenance has been proposed to prevent component falure and reduce cost by montorng real-tme electrcal quanttes and assessng condton of equpment [10, 11]. The new technologes n both asset management and outage management use non-operatonal data, whch s recorded n the feld ntellgent electronc devces (IEDs), and reveals the current condton of the system. Ths paper consders the overlappng of IED database use by outage and condton-based asset management and proposes the concept of ntegraton of asset management and outage management tasks. The expected benefts from ntegraton nclude: savngs n IED nstallaton expendtures, effcent collecton and use of non-operatonal data, reduced falure cost and better system relablty, and fnally more return on nvestment. The concept of ntegraton s presented n secton II, followed by benefts of ntegraton n secton III, whch deals wth optmzaton of asset and outage management tasks under the ntegrated database. Impact of ntegraton on the benefts n asset and outage management s dscussed n secton IV and V respectvely. Secton VI contans concluson, and s followed by acknowledgements and references.

2 2 II. COCEPT OF ITEGRATIO A tradtonal dstrbuton utlty busness process approach s llustrated n Fgure 1. In ths approach, outage analyss s prmarly based on nputs from outage detecton, tellng whch customers are connected, and ncdent verfcaton reportng (IVR), tellng whch customers have reported loss of power. Asset management s prmarly based on off-lne data wthout extensve use of operatonal and/or condton based nonoperatonal data. Wth the development of new technology n fault locaton and mantenance predcton, system falures may be reduced n terms of frequency and duraton. management of equpment assets leadng to optmzed equpment mantenance practces. Ths wll reduce the rsk of outages, as measured by relablty ndces, energy not served, cost of falure, or other measures. The optmzaton may be mplemented usng an asset management concept that selects and schedules mantenance tasks to mnmze outage rsk. Fg.2: Integrated asset management and outage management tasks Fg.1: Tradtonal dstrbuton utlty busness process One of the constrants to mplement those technologes s the avalablty of data. Condton-based mantenance, for nstance, requres real-tme feld-recorded data to do the condton assessment, e.g. voltage, load current, etc. On the other hand, to mplement a model-based fault locaton algorthm, an accurate system model s requred, ncludng system topology, on/off status of swtchng devces, parameters of components, etc. [9] It can be seen from the dscusson above that the flow of data requred to mprove the busness process s no longer as shown n Fg.1. Outage management and asset management now share the need for certan data and models. It s more effcent to generate an ntegrated database. Integratng the outage and asset management tasks through the use of data and models of common nterest should enhance the effcency and effectveness of the overall busness process because t prevents ether duplcaton or lack of nvestment n nstallng montorng devces, and collectng and storng data. Ths strategy of usng extensve feld data provdes two benefts: a) due to mproved mantenance, prmary equpment wll fal less frequently, reducng the number of forced outages; b) due to more precse locaton of a faults and better predcton of the equpment health, outage restoraton practces wll be far more effcent and effectve. The beneft can be evaluated from two aspects: a) System relablty. Ths s reflected by the mpact on relablty ndces. b) Return on nvestment of utltes. Ths s measured by optmzaton n captal and operatng expense. The mproved busness process should explore the correlaton of outage management wth rsk-based The ntegrated asset management and outage management tasks are shown n Fg.2. Fault locaton and condton assessment retreve feld-recorded operatonal and nonoperatonal data, as well as system models and confguraton data from a common database. Based on ths data, the reducton n falure cost s evaluated n an ntegrated rskbased assessment program. As very few data can be acqured from a dstrbuton system, and the data s wth poor qualty, algorthms that s flexble n the number of nputs and s robust to naccurate data s developed, whch s ntroduced n the followng sectons. III. OPTIMIZATIO OF ASSET MAAGEMET TASKS One of major challenge faced by the utltes s the allocaton of ther resources for the expandng system whle mantanng the system relablty. There are dfferent methods followed n the ndustry to schedule the mantenance and replacement of the components. Most of the methods do not consder the mantenance schedule by optmzng the cost of mantenance and relablty ndces. Some state regulatory commssons requre utltes to schedule mantenance cycles (tme based mantenance) to nsure system relablty. Some regulators have been nstead settng mnmum relablty standards, allowng the utltes to move from tme based mantenance to condton based mantenance, and relablty centered mantenance practces. These are more cost-effectve and also more focused at nsurng relablty. Ths work focuses on optmzng the cost of component mantenance schedule for utltes whle ensurng the mnmum relablty requrements. In the event a utlty s not restrcted wth any relablty regulatons, the utlty could use ths technque as a tool to optmze ther mantenance task to reduce the energy not served (ES) so that the revenue s mzed. The proposed technque could be llustrated usng the followng steps.

3 3 1. Identfyng the Crtera for Equpment Condton Assessment and Allocatng Relablty Dstrbutons In order to wsely allocate the predctve mantenance budget, t s mportant to dentfy the condton of the Most of the utltes perform routne component assessment and decde the condton of the Ths type of assessment cannot be generalzed and every tme t needs nspecton before predctng the health level of the Alternatve to ths would be to use the relablty models. When usng relablty models the most common practce s to use the average falure rates. Even though constant falure rate models are faster, easly tractable and needs very few data for calculatons, they wll not accurately predct the condton [12]. Especally wth age, components wll have ncreasng falure rate and there s a hgh probablty that the constant model wll overstate the condton of the Use of constant falure rate models s very much common n power ndustry [13], and mportance of developng technques to model the falure of the components as a functon of tme n an effcent way must consdered. Updatng the relablty models based on the perodc component nspecton nformaton wll generalze the component condton assessment and ths can be seen as a very powerful tool n predctng the health level of components [14]. Most of the components used n the power ndustry are made out of several parts. In terms of relablty each part n a component could be consdered as a separate subsystem. Further the utlty wll have nformaton about other factors that affect the lfe of a component, eg: age, loadng, frequency of mantenance, fault hstory, envronmental condton etc. We would lke to adopt the approach gven n [14], authors consder each condton that affect a component as a separate crteron and try to fnd the condton of a component based on relablty analyss and nspecton data of each crteron and ther mportance to the components healthy functonng. It s very vtal to dentfy approprate crterons for each These crterons can t be generalzed for smlar components. Crterons may vary wth topologcal locaton of the component, manufacturers, experence wth partcular type of component, etc. As a part of ths project, [15] gves a detaled methodology to dentfy the crterons of power transformers and crcut breakers. These crterons are practcal as they are based on ther manufacturer equpment database, hstorcal falure causes and mantenance actvtes. Smlar approach could be taken to fnd the crterons of other components. s based on the crterons effect on the falure of the component and the number of mantenance / replacement needed for a crteron durng the lfe of the Fg.3 descrbes the selecton of crterons. For each crteron of a component, we would lke to assgn a falure dstrbuton. Falure rate dstrbuton for a crteron wll be assgned based on the standards and regulatons, manufacturer data on each crteron, hstorcal data and expected lfetme of the component [15]. Usng a partcular falure dstrbuton for all the components and ther crterons may result n obtanng naccurate falure rates. The proposed technque allows us to use dfferent falure rates for components and ts crterons. Some of the common dstrbutons that can be used are gven n table 1. TABLE I: TYPICAL DISTRIBUTIOS Type of Falure rate Dstrbuton Constant wth tme Exponental, Webull Increasng wth tme ormal, Webull Decreasng wth tme Gamma, Webull Increase and then decrease Lognormal wth tme When the age of a component s consdered, older the component, the probablty to fll ncreases, thus t wll be a ncreasng falure rate. Probablty of falure due to the topologcal locaton or geographcal locaton wll not change, thus we can use a constant falure rate model for these crterons. But the more the experence we have wth a partcular type of component, we wll be able to predct the performance of the component much better, thus the falure rate of ths crteron would be decreasng wth tme. 2. Compute falure rate model for each component At the dstrbuton level, condton data for many components are not avalable. If a utlty s to move from tme based to relablty based mantenance a database of condton data for all equpment must be developed and mantaned. From ths data, falure rates can be calculated based on the followng dscusson. Fg.3: Crteron Selecton Flowchart [15] It should be noted that not all crterons have same mportance when t comes to healthy functonng of a We use a smlar approach taken by [14] to weght the mportance of each crteron. Weghtng of each crteron Fg. 4: Topology of SF6 Crcut Breaker

4 4 To calculate the falure rate of a component at any gven tme t many approaches are taken n the lterature. In ths analyss we would lke to consder each crteron of the component as a subsystem and use a relablty topology (eg. seres, parallel, seres-parallel, parallel seres, etc.) based on the relatonshp of each crteron wth other crterons of the component based on effect of the component falure. Fg.4 llustrates the topology for a SF6 crcut breaker as an example Each crteron wll be assgned a weghted falure rate based on the dscusson n step 1. Weghted falure rates for each crteron and topology of the crterons are used to determne the falure rate model of the Once falure rate model of for component s determned, the next step s to fnd the condton of each Real tme equpment montorng data, falure rate model, component ratngs suppled by the manufactures and the standards and regulatons are used to determne the condton of each The condton of each component s consdered n relablty pont of vew. Fg.5 llustrates the condton assessment and the output from the condton assessment can be a factor of, propostonal to falure rate, remanng or qualtatve assessment of health of the Fg.5: Condton Assessment Flowchart 3. Allocate the requred level of mantenance for each Dstrbuton system performance can be degraded both by controllable events (e.g., lack of mantenance of components and rregular tree trmmng) and uncontrollable events (e.g., lghtnng and accdents). When the performance of a utlty s consdered, t s not logcal to measure the performance affected by the uncontrollable events. Thus n ths analyss the relablty ndces that are used would nclude only the events that can be controlled by the utltes. The subscrpt C wll be used to ndcate that the relablty ndces are calculated based only on the controllable events. Once the condton of each component s computed, based on the performance / relablty requrements (requred SAIDI C, requred SAIFI C, requred CIME C, and mum allowed ES C etc.) the utlty should be able to schedule ts mantenance. As a part of ths work we have proposed an algorthm to acheve the requred mprovement of each component n such a way, that the total cost of mprovng the condton of components n the system s mnmzed [16]. For each component, an mportant ndex would be allocated. Important ndex would be a functon of load type, locaton of the component, avalablty of redundant components n the system, tme taken to mantan or replace the component upon falure, revenue loss upon falure of the Based on the requred mantenance and the mportant ndex of each component, components wll be gven a rank. Rank 1 would be gven to the component whch has hgh rsk. Rankng of the components wll gve qualtatve nformaton for the future plannng. Fg.6 llustrates the procedures nvolved n rankng the components. Fg 6: Optmal Component Rankng Flowchart 4. Ensure the requred mantenance s cost effectve than replacng the Components ranked hgh are defectve and needs mmedate attenton compared to the ones ranked low. Some of the hgher rank components can be crtcally faulty and t may be economcally compettve to replace the component than mantanng t. Therefore at ths stage budgetary calculaton must be done to see whether t s cost effectve to mantan the By mantanng a component we wll mprove but f the component s really bad, the replacement wll mprove the relablty by a huge margn. Ths wll result n the utlty achevng the requred performance level, by not mprovng the components whch have least ratng. Ths budgetary calculaton should nclude comparson between the remanng lfe of the component by mantenance and the mantenance cost versus the nvestment. Fg.7 shows ths comparson. Fg.7: Mantenance Vs. Replacement If replacng the component s cost effectve, then utlty should take necessary acton to replace the component and check the next component n the queue (ranked next) cost effectveness. If the analyss shows t s cost effectve to mantan the component, go to the next step. 5. Guarantee that the mantenance/replacement of all the components wll be under the allocated budget Ths step s smlar to the prevous step. Here we want to ensure, that the requred mantenance wll not exceed the budget lmtatons. Out of the components whch were not replaced, once agan the hgh ranked components wll get preference as they are the major contrbutors to the poor performance. Fg.8 explans the procedure.

5 5 Fg.8: Is mantenance cost effectve? In ths work we are consderng an optmal schedulng scheme. All the parts n the component would be consdered n smlar way to that of decdng the crterons. In ths analyss we mnmze the cost of mprovng the parts whle achevng the desred falure rate. If the mantenance cost s wthn the allocated budget, the component wll be scheduled for mantenance. The next component n the queue wll be consdered. The process wll start from step 3. If the mantenance exceeds the allocated budget go to step Incase the component can t be mantaned derate the Fg.9: Component Deratng and Reconfguraton In the event the component mantenance exceeds the allocated budget, the proposed method allocates a new ratng for the component (derate). The deratng wll be based on the system topology. We must ensure that the deratng wll not cause overloadng the component, f the deratng causes overloadng to the component, then the system must be reconfgured and part of the load should be allocated to supportng feeders/ laterals n a way none of the components are overloaded. If we can t acheve a reconfgured system, wthout overloadng any of the components, then the utlty could leave the part of load that can t be suppled wth reconfgurng the system, wth the component and prepare for the falure. Process s explaned n Fg.9. IV. OPTIMIZATIO OF OUTAGE MAAGEMET TASKS Accurate fault locaton n dstrbuton systems stll remans one of the man challenges n the utlty ndustry. Whle many dfferent fault locaton algorthms are proposed so far [5-10], fndng exact fault locaton s qute often a major part of the overall reparng and restorng tme. The queston that stll faces the developers of the algorthms s how to mprove the accuracy of the algorthms through mprovements n data recordng and collecton. The work reported n ths paper llustrates the ssues assocated wth mplementng accurate fault locaton n dstrbuton systems. It has been recognzed that fault locaton depends on data avalable for the mplementaton; hence the noton of performng a senstvty study of the fault locaton algorthm due to the change n avalable data s conveyed. Implementaton of an algorthm that takes advantage of spare measurements of voltage sags caused by the faults s ntroduced to llustrate how mprovements n fault locaton may be acheved f an extensve feld recorded data s used [17]. Ths fault locaton approach has been dscussed n more detals n other related references publshed earler [18]. 1. Fault locaton algorthm selecton ow that the asset and outage management share an ntegrated database, the accessble data for fault locaton s more elaborate than what was avalable when just a typcal data base for outage management was consdered. The expanson of data brngs not only a larger quantty but also a varety,.e. more types of data. Ths makes possble to mplement several fault locaton algorthm that applcable for a gven fault case. To mprove the accuracy of fault locaton, the senstvty of fault locaton algorthm to type, quantty and qualty of data s studed, and the fault locaton algorthm selecton s done based on the result. The proposed study ams at revealng the senstvty of fault locaton algorthms to several nfluencng factors: a) Pre-fault load condton; b) Dstance of fault pont from measurements; c) Fault mpedance; d) Branch gong out from the node between the fault pont and measurement; e) Imprecse feld-recorded data. The followng steps n the senstvty study are defned. Frst, fault scenaros assocated wth the nfluencng factors above are generated and smulated n ATP. Then, fault locaton algorthms are appled and the accuracy recorded. The mnmum set of data and data accuracy requrement s then determned for each algorthm. After ths study, one wll be able to select a fault locaton algorthm from the lst of algorthms and apply to a partcular fault based on the faulted area and avalablty of data. The result from the selected algorthm wll be the most accurate one, so feld crew wll be able to pnpont the fault wthn the shortest tme possble. Thus the overall tme to carry out fault processng wll be the shortest possble and the SAIDI wll be reduced sgnfcantly. 2. ew fault locaton algorthm development Fault locaton algorthms for dstrbuton systems should have the ablty to cope wth nsuffcent data, because of the general lack of wdespread use of data recordng devces n dstrbuton systems. An algorthm proposed n [16] uses voltage sags recorded from the sparsely nstalled power qualty meters (PQM) to detect the faulted node by comparng the measured and calculated values assumng fault occurred at dfferent nodes. Dfferences n pre-fault and fault voltage magntudes recorded by sparse voltage measurements n the system are utlzed. The merts of applyng ths algorthm as a dstrbuton system fault locaton method are as follows: a) It deals wth the realty of nsuffcent measurements

6 6 (just from PQM) n dstrbuton system, although the accuracy of the algorthm s affected by the number and placement of the measurements. b) It mnmzes the mpact of fault mpedance on the accuracy by consderng fault as a specal load connected to the faulted node. c) It takes nto account the characterstcs of dstrbuton system: non-transposed feeders, sngle-phased lne sectons and nodes, and radal topology. d) It provdes a lst of lkely fault locatons so that feld crew can start wth the most lkely fault locaton frst and move down the lst untl the fault s found. The flow chart of the algorthm s shown n Fg.10. From Measurement Devces Pre-Fault Voltage and Current Phasors Measured at the Root ode of the Feeder Durng-Fault Voltage and Current Phasors measured at the Root ode of the Feeder = 1, total feeder nodes Assgn for all nodes the Durng-Fault Voltage Measured at the Root ode Fault Current Computaton Durng-Fault Voltage Magntudes measured at the Remote odes Durng-Fault Load Flow From Feeder Database Load feeder data Settng the load model Dstrbuton Transformer Power Ratng Estmaton (fgure 2) Was the convergence acheved? Durng-Fault Voltage Msmatches Computaton for all Remote odes Computaton of the Fault Locaton Indces Fg.11 Flow chart of refned algorthm Objectve: mn{ PL PL' = ΔP } L mn{ } mn{ SW ( P ' P ') + ( P ' P ') + ( P ' P ') = ΔP } (1) A B B C C A φ Fg.10 Flow chart of voltage-sag based algorthm [16] Fault Locaton Rankng Lkeky Faulted node Process performed usng durng-fault quanttes Refnement of the voltage-sag based algorthm s studed. The erroneous system model and feld-recorded data are taken nto consderaton and an evaluaton of accuracy s done before runnng the man algorthm. Two ndces, J and RI are ntroduced to quantfy the nfluence of number of measurements and data qualty respectvely. If both J and RI are larger than a pre-set threshold, the algorthm s consdered as not applcable to the case. In the decson-makng stage, the dfferences n accuracy of measurements are ntroduced by means of weghted least square functon. Ths prevents error caused by sngle erroneous data. Fg.11 shows the flow chart of the refned algorthm. 3. Fault solaton strategy To further reduce SAIDI, fault solaton strategy ams at restorng power to the largest number of customers by system reconfguraton durng the perod of mantanng and replacng of faled components. An optmzaton problem mnmzng the loss of load, number of swtches nvolved and phase mbalance s formed as follows: s.t. V mn P ' P Tj λ λ SWm V ' V Tj SWm where: varables wth are the values after reconfguraton; P s the load at node ; L ΔP L s the total loss of load after reconfguraton; P, P, P are three phase power at root nodes of the feeders; A B C ΔP φ s the 3-phase load unbalance at root nodes of the feeders; s the total number of swtches nvolved n the SW reconfguraton; V ' s the voltage magntude at node after reconfguraton, mn wth V and V as lower and upper lmts of node voltage; P s the output power transformer j, whose prmary sde s Tj connected to the transmsson system and the upper lmt s P ; Tj λ s the falure rate of swtch m, wth SWm λ as the SWm mum tolerable value. (2)

7 7 V. OPTIMIZED OUTAGE COST THROUGH RISK-BASED ASSESSMET 1. Optmzed outage cost through rsk-based assessment Rsk-based analyss s used to estmate outage cost n ths paper. The rsk s formed usng selectve relablty ndces reflectng nterests of both customers and utltes. Reducton n rsk s comprsed of two parts: reducton from mantenance, ΔRsk AM and reducton from refnng fault locaton and hence other outage management tasks, ΔRsk OM. The consequence of equpment falure can be expressed as the weghted sum of SAIFI, SAIDI, ES (energy not served) and DevRsk (cost of mantenance) [19]. a) Effect on customer satsfacton: nk SAIFI( = λ( (3) n k d j j= 1 SAIDI( = λ ( (4) b) Revenue loss of utlty: ES ( = λ( P j d j (5) c) Cost of equpment falure: MTTF DecRsk( = Cost( { λ ( + (1 + r) } (6) where: λ( s the falure rate of component k; s the number of nterrupted customers for each sustaned nk nterrupton; s the total number of customers served for the area; Pj s the load connected at load pont j ; d j s the duraton of nterrupton experenced by the jth customer; Cost( s the cost of reparng component k; r s the rate of return acqured from deferrng replacement of a component; MTTF s the mean-tme-to-falure of component k. Snce mantenance changes the factors of falure rate (λ) and mean-tme-to-falure (MTTF), reducton n rsk obtaned from mantanng a component k can be expressed as follows: SAIFI ( SAIDI( ΔRsk AM ( = α1 Δλ( + α2 Δλ( λ( λ( ES ( + α3 Δλ( + α4 λ( DevRsk( + ΔMTTF ( MTTF ( where α 1 ~α 4 are the weghtng factors. DevRsk( Δλ( λ( Smlar to expresson of AM, the consequence of nterrupton from the outage management can be comprsed of weghted sum of SAIDI, ASIDI (average servce avalablty ndex), MAIFI (momentary average nterrupton event frequency ndex) and MED (major event day). (1) Effect on customer satsfacton: (7) r SAIDI( ) = (8) (2) Revenue loss of utlty: r L ASIDI( ) = (9) (3) Penalty for mportant customers senstve to momentary nterruptons: IM m MAIFI( ) = (10) (4) Cost for reconfguraton: MED( ) = { SAIDI( ) SAIDI( ) T MED } (11) where Š ƒ ˆ ƒ Š Š ˆ ˆ ƒ Š ƒ Š ƒž ˆ Š Ž ƒ ˆ ƒ Š Š ˆ ƒ Š ˆ ˆ ƒ Š ƒ Š ƒœ ƒ ˆ ƒ Š Š Ž ƒž Snce fault locaton practces change the duraton of fault (r), number of nterruptons (IM) and the range of affected area ( m ), rsk reducton n one nterrupton event s expressed as follows: SAIDI( ) ASIDI ( ΔRskOM ( ) = β1 Δr + β2 Δr r r MAIFI ( ) + β3 ΔIM IM MED( ) + β4 Δr r where β 1 ~β 4 are the weghtng factors. MAIFI ( ) + Δ m m (12) The overall reducton of rsk obtaned n a reportng perod s expressed as a lnear combnaton of ΔRsk AM and ΔRsk OM. Δ Rsk = ΔRsk ( + ΔRsk ( ) (13) AM 2. Optmzaton of captal nvestment The optmzaton of nvestment s based on the rsk-based assessment of outage cost. The nvestment s dstrbuted among payng for nstallaton of new montorng devces, mprovng communcaton and database nfrastructure, and budgetng the equpment repar/replacement cost so that the mum reducton n outage cost can be acheved. The topc of optmzaton of captal cost s not further explored n ths paper and dscusson of how optmzed mantenance and outage management tasks may mpact the strategy for captal nvestment wll be reported n the future. OM VI. ITEGRATIO BEEFITS As can be seen from the above dscussons, both asset and outage management tasks can be enhanced from the ntegraton. The mpact of ntegraton provdes benefts n

8 8 reducng number of scheduled (forced) outages and duraton of random (fault) outages. Hence the mproved performance of asset management and outage management has several postve mpacts: a) Customers mpacts: The mprovement n system relablty as measured by relablty ndces ndcates that the requrement of better servce s met. Ths s reflected by the mprovement n the values of the ndvdual relablty ndces as a result of better data recordng and collecton practces comng out of the ntegraton concept f ntegrated outage and asset management. b) Utlty mpacts: The nvestment n the equpment, nformaton nfrastructure and labor s more effcently utlzed f the optmzaton technques for asset and outage management proposed n the paper are used. The return on nvestment s ncreased and may be assessed by the means of rsk-based the reducton of outage and revenue ncrease due to relablty mprovements. V. COCLUSIOS The ntegraton of asset and outage management tasks s proposed n ths paper. The man contrbutons of ths paper nclude: Current development n asset and outage management are analyzed, and possblty for an ntegraton s ponted out; An ntegraton of database for asset and outage management tasks s proposed; Optmzaton of asset and outage management tasks usng ntegrated database s presented; A method to evaluate the mpact on the reducton of falure cost brought by ntegraton s outlned. Implementaton of the proposed ntegraton and a quantfed evaluaton of the benefts of ntegraton wll be presented n future. VII. ACKOWLEDGMET The authors gratefully acknowledge the contrbutons of Rodrgo A.F. Perera on the fault locaton technque mentoned n ths paper. VIII. REFERECES [1] IEEE gude for electrc power dstrbuton relablty ndces, IEEE Standard , May [2] Pooran Ramachandran, IEEE benchmarkng 2007 results, IEEE Workng Group of Dstrbuton Relablty, Oct [3] James orthcote-green, Robert Wlson, Control and automaton of electrcal power dstrbuton systems, ew York: Taylor & Francs, 2006, p [4] IEEE/PES Task Force on Impact of Mantenance Strategy on Relablty of the Relablty, Rsk and Probablty Applcatons Subcommttee, The present status of mantenance strateges and the mpact of mantenance on relablty, IEEE Trans. Power Systems, vol. 16, ssue 4, ov. 2001, p [5] C.. Lu, M.T. Tsay, Y.J. Hwang, Y.C Ln., An artfcal neural network based trouble call analyss, IEEE Trans. Power Delvery, vol.9, Issue 3, July 1994, p [6] K. Srdharan,. Schulz, Knowledge-based system for dstrbuton system outage locatng usng comprehensve nformaton, IEEE Trans. Power Systems, vol.17, Issue 2, May 2002, p [7] Jun Zhu, D.L. Lubkeman, A.A. Grgs, Automated fault locaton and dagnoss on electrc power dstrbuton feeders, IEEE Trans. Power Delvery, vol.12, Issue 2, Aprl 1997, p [8] Myeon-Song Cho, Seung-Jae Lee, Duck-Su Lee, Bo-Gun Jn, A new fault locaton algorthm usng drect crcut analyss for dstrbuton systems, IEEE Trans. Power Delvery, vol.19, Issue 1, Jan.2004, p [9] Z. Galjasevc, A Abur, Fault locaton usng voltage measurements, IEEE Trans. Power Delvery, vol.17, Issue 2, Aprl 2002, p [10] Ward Jewell, Project Leader, Joseph Warner, James McCalley, Yuan L, Sree Rama Kumar Yeddanapud, Rsk-Based Resource Allocaton for Dstrbuton System Mantenance, Fnal Project Report, PSERC Publcaton 06-26, Aug [Onlne]. Avalable: [11] Satsh att, Rsk Based Mantenance Optmzaton Usng Probablstc Mantenance Quantfcaton Models of Crcut Breakers, Ph.D. Dssertaton, Department of Electrcal and Computer Engneerng, Texas A&M Unversty, College Staton, Dec [12] J.B. Bowles, Commentary Cauton: Constant Falure Rate Models may be Hazardous to your desgn, IEEE Trans. on relablty, Vol. 51, o. 3, Sep [13] Brad Retterath, S.S. Venkata, Al A. Chowdhury, Impact of Tme- Varyng Falure Rates on Dstrbuton Relablty, n Proc. 8 th Internatonal Conference on Probablstc Methods Appled to Power System, Iowa State Unversty, Ames, Iowa, September 12-16, [14] Rchard E Brown, George Frmpong and H. Lee Wlls. Falure Rate Modelng Usng Equpment Inspecton Data, IEEE Trans. on power systems, Vol. 19, o. 2, May [15] Prasad Dongale, Equpment Condton Assessment and ts Importance n Estmaton and Predcton of Power System Relablty, M.S. Thess, Dept. of Elect. Eng., Wchta State Unversty Jan [16] Vsvakumar Aravnthan, Ward Jewell, Optmal Relablty Allocaton for Radal Dstrbuton System wth Performance Based Rate, n Proc. of the 2008 Fronters of Power Conference, Oklahoma State Unversty, Stllwater, October [17] Rodrgo A.F. Perera, Mladen Kezunovc and Jose R.S. Mantovan, Fault locaton algorthm for prmary dstrbuton feeders based on voltage sags, IEEE Trans. on Power Delvery, (accepted, n press) [18] Rodrgo A. F. Perera, Mladen Kezunovc, José R. S. Mantovan, Fault locaton algorthm for prmary dstrbuton feeders based on voltage sags. 16th Power System Computaton Conference, July [19] S.R.K Yeddanapud,. Yuan L, J.D. McCalley, A.A. Chowdhury, W.T.Jewell, Rsk-based allocaton of dstrbuton system mantenance, IEEE Trans. on power system, Vol. 23, o. 2, May IX. BIOGRAPHIES Yma Dong (S 07) receved her B.S. and M.S. degrees from orth Chna Electrc Power Unversty, Bejng, Chna, n 2005, 2007 respectvely, all n electrc engneerng. She has been wth Texas A&M Unversty pursung her Ph.D. degree snce August Her research nterests nclude applcatons n power system protecton, dgtal smulaton, power system analyss and control. Vsvakumar Aravnthan (S 04) receved hs Bachelor s and Masters degree n Electrcal Engneerng from Unversty of Moratuwa, Sr Lanka n 2002 and 2004 respectvely and receved hs MS n Electrcal Engneerng from Wchta State Unversty n Currently, he s pursung hs Ph.D. at Wchta State Unversty. Hs research nterests nclude power dstrbuton, power qualty, power system analyss and controls. Mladen Kezunovc (S 77-M 80 SM 85 F 99) receved the Dpl. Ing., M.S. and Ph.D. degrees n electrcal engneerng n 1974, 1977 and 1980, respectvely. Currently, he s the Eugene E. Webb Professor and Ste Drector of Power Engneerng Research Center (PSerc).. Hs man research nterests are dgtal smulators and smulaton methods for relay testng as well as applcaton of ntellgent methods to power system montorng, control, and protecton. Dr. Kezunovc s a Fellow of the IEEE, member of CIGRE and Regstered Professonal Engneer n Texas. Ward Jewell (F 03) teaches electrc power systems and electrc machnery as a professor of Electrcal Engneerng at Wchta State Unversty. He s Ste Drector for the Power System Engneerng Research Center (PSerc). Dr. Jewell performs research n electrc power qualty and advanced energy technologes and has been wth Wchta State Unversty snce 1987.