Smart Grid Infrastructure for Distribution Systems and Applications

Transcription

1 Smart Grd Infrastructure for Dstrbuton Systems and Applcatons Sas Melopoulos, George Condes, Rene Huang, Evangelos Farantatos, Sungyun Cho, Yonghee Lee, and Xuebe Yu School of Electrcal and Computer Engneerng Georga Insttute of Technology Atlanta, Georga e-mal: Abstract: Ths paper presents a new smart grd nfrastructure for actve dstrbuton systems that wll allow contnuous and accurate montorng of dstrbuton system operatons and customer utlzaton of electrc power. The nfrastructure allows a complete array of applcatons. The paper dscusses four specfc applcatons: (a) protecton aganst downed conductors, (b) load levelzaton, (c) loss mnmzaton and (d) relablty enhancement. Index Terms PMU, GPS-synchronzaton, Data Accuracy, Energy Functon, Transent Stablty. Glossary GPS: Global Postonng System, PMU: Phasor Measurement Unt, SE: State Estmaton, Infrastructure: Techncal structures (meters, communcatons, software, etc.) that support an operaton. Introducton The dstrbuton system s a vtal part of a smart grd nfrastructure especally n the lght of recent trends towards dstrbuted green electrc power and customer nvolvement. As these trends contnue, the dstrbuton system wll be transformed from a system wth passve loads nto an actve dstrbuton system wth customer and utlty owned resources. Customers may own photovoltac roofs, small wnd turbnes, bacup power, electrc vehcles or pluggable hybrd cars, and other resources that may become avalable n the future. Utltes may elect to nstall storage systems at strategc locatons, dynamc VAR support systems, addton of reclosers, swtches, etc. Proper management of ths system can result n substantal savngs that wll beneft both end users of electrcty as well as utltes. Savngs wll come from the ablty to levelze electrc loads, mnmze losses and mprove relablty. In addton, proper management of the actve dstrbuton systems of the future wll create addtonal benefts that wll ncrease the safety of these systems by enablng the dentfcaton of unsafe condtons such as down conductors and other faulty condtons for whch present approaches are not fully relable. It s also mportant to mentoned that these systems can also provde mmedate trouble reports such as locaton of faults and other detals that wll enable utltes to remedy these problems wth mnmum cost. We are provdng a proposal for an nfrastructure and applcatons software that wll enable the stated objectves. The word nfrastructure s used to mean the hardware (meters, communcaton systems, etc) as well the necessary software that wll support the functons of a modern dstrbuton management system. Proposed nfrastructure Our vson s to create an nfrastructure for dstrbuton systems that wll enable the goals and objectves of the smart grd. For our purposes, the smart grd objectves that we are focusng on s the ablty to operate and coordnate a dstrbuton system wth ts customer owned apparatus and possble resources such as customer owned PHEVs, photovoltac roofs, wnd turbnes, etc. as well as utlty owned dstrbuted recourses such as sectonalzng devces, storage devces, capactors, SVCs, etc. The tools that we propose to use for achevng ths goal must be a combnaton of advanced hardware and software tools that wll enable the accurate and frequent montorng of the overall dstrbuton system and the nterconnected customer load and resources and wll optmze the control of the system n real tme. Therefore, as n any control system, we separate the components of such a system nto two major components: (a) A hgh fdelty real tme montorng system that wll enable full awareness of the system status and the operatng condtons and (b) Advanced optmzaton algorthms that wll generate real tme control sgnals based on the objectves of specfc applcatons. A bloc dagram contanng the major tools we suggest for ths realzaton s llustrated n Fgure 1. In the subsequent paragraphs we present more detals on each of these components. The most crtcal parts of a real tme montorng system are an advanced meterng nfrastructure (AMI) /11 $ IEEE 1

2 and a three phase state estmaton algorthm based on a combnaton of synchronzed measurements and nonsynchronzed measurements. Fgure 1: Major Components of Proposed Infrastructure As far as the AMI s concerned, t s expected that house meters wll be massvely deployed wthn the next few year n the houses (for example as part of smart applances), enablng actve control and montorng of resdental loads. It s unequvocal that data obtaned from house meters are not suffcent to acheve full observablty of the entre dstrbuton feeder. However, n order to be able to optmally control and coordnate all avalable devces (such as capactor bans, voltage regulators and utlty owned dstrbuted generaton resources or storage devces) we need to acheve full observabtlty of the feeder. In order to acheve ths objectve, we suggest the deployment of a novel smart meterng devce whch we wll refer to as "Unversal GPS Synchronzed Meter". We propose ths to be deployed on the poles of dstrbuton lnes and along the dstrbuton feeders. The major characterstcs of the UGPSSM are the followng Low Cost Power Autonomous Synchronzed Voltage and Current Phasor measurng capablty Two way communcatons capablty The major components of such a devce are llustrated n Fgure. Whle such devces wth varous capabltes are commercally avalable today t s mportant to note that ther cost s relatvely hgh and may dscourage massve deployment. However the technology exsts to mnaturze these devces wth dramatc cost reductons. Research towards ths development s under way. Fgure : Conceptual Vew of Self-Powered GPS- Synchronzed, Communcatons-Enabled Smart Meter Note that t s crtcal to have one devce per phase of the dstrbuton lne snce mbalances and asymmetres are a common feature of dstrbuton systems and have to be taen nto account for the modelng and estmaton procedure. The energy harvester s one of the ey components of the proposed devce. It wll be able to harvest suffcent energy from the magnetc feld of the lne n order to provde the necessary nput power to the varous electronc devces such as the mcroprocessor, the communcatons module etc. The power harvester can be desgned as a smple col wrapped around a hgh permeablty magnetc core. Another ey and novel concept n the proposed devce s that the measurements are GPS-synchronzed. GPS-synchronzed measurements transform the state estmaton process nto a robust method for provdng the system model n real tme. Ths feature can be mplemented smply by the ntegraton of a GPS cloc and phase-locng the GPS cloc wth the analog to dgtal converter that wll provde the devce wth PMU capabltes. The ntroducton of ths technology n the proposed meters (UGPSSMs) that wll be deployed n the dstrbuton system wll provde hghly accurate measurements both n magntude and phase (typcal PMU accuracy s less than 1 μs and magntude accuracy better than 0.%). As a result each meterng devce wll be able to measure the voltage and current phasor of the node t s connected to wth a globally vald tme reference. These synchronzed and globally vald measurements wll be tmed tagged and wll be sent and collected n a central locaton (substaton or Modern Dstrbuton Management System) where they wll be used n the state estmaton algorthm, along wth nonsynchronzed measurements obtaned from house meters or SCADA measurements. The state estmaton algorthm based on GPS-synchronzed measurements s a drect computaton mmune to convergence problems of the tradtonal state estmator. Specfcally, f the

3 state estmaton s based only on GPS-synchronzed measurements, the formulated problem s lnear and can be solved drectly. However, snce we propose that we use all avalable measurements (synchronzed and nonsynchronzed) the formulated problem s non-lnear, thus an teratve algorthm s needed for the soluton, as explaned n detal s the followng correspondng secton. The state estmator also performs bad data dentfcaton and removal and fnally provdes the operator the true state of the dstrbuton system n real tme. The results of the state estmator (SE) can be used as an nput to advanced optmzaton algorthms wth several objectves such as mnmzaton of losses, mnmzaton of pea load, relablty mprovement etc. The formulaton of the optmzaton problems has to tae nto account all the controllable devces n the networ such as nverters, storage devces etc, but also ther operatng constrants. Fnally, addtonal constrants such as handlng of crtcal loads, customer nconvenences, prce sgnals etc. may also be taen nto account as t wll become clear n the dscusson of the applcatons. Detals on the State Estmaton algorthm, the communcatons nfrastructure that s needed to support the proposed nfrastructure and the formulaton of optmzaton algorthms sutable for varous applcatons are presented n the next sectons. Note that the dgtzed data are sent to the process bus va hgh speed communcaton meda ncludng wreless communcaton of ggabt speed. Hgh frequency (GHz) pont to pont communcatons are cost effectve. Ths proposed approach s advantageous compared to typcal Smart Meter/AMI applcatons where the data transmsson rate s n the order of mnutes. Of course t s recognzed that the quantty of the data that wll have to be processed s bg. That s the reason why we propose a decentralzed approach where the data are processed n the substaton level, where the state estmator wll be performed. In ths case, the dstances are relatvely short and therefore tme latences ntroduced by the communcaton system are very short and wll not mpact the performance of the system. Snce the data are dgtal wth accurate tme tags to mcrosecond accuracy, the accuracy and synchronzaton of the data wll reman ntact. Also dgtal transmsson prevents data dstorton or loss that occurs wth tradtonal analog transmsson va electrcally wred cables. Communcatons Snce the prevously descrbed nfrastructure s desgned for dstrbuted data acquston and usng the data for extractng the real tme operatng condtons of a dstrbuton system, a dstrbuted relable and hgh speed communcaton system s essental.. In order to meet the requrement of hgh speed communcatons, we proposed the deployment of the meterng system llustrated n Fgure. We refer to ths meter as the Unversal GPS Synchronzed Meter (UGPSSM). The UGPSSMs measure analog data such as voltage and current wth hgh samplng rates. The dgtal data are transmtted va hgh speed communcaton meda to a process bus as t s llustrated n Fgure 3. The transmsson dstances are relatvely short snce process buses can be establshed at strategcally placed locatons. IEDs connected to the process bus can process the data nto phasors (fundamental, harmoncs, other dsturbances, etc.) and transmt the results to the staton bus va another communcaton nfrastructure. At the substaton the data wll be used n the state estmator procedure. The output of the state estmator wll be the real tme operatng condton of the entre feeder. Fgure 3: Communcaton Infrastructure for Dstrbuton Systems n Smart Grd Fgure 3 llustrates the dagram of the proposed communcaton nfrastructure capable of hgh speed and relable data acquston system. The proposed nfrastructure conssts of three parts: process level, bay level, and staton level as descrbed n IEC standard [6]. (These parts are ntally defned n the feld of substaton automaton, but can also be appled to dstrbuton systems). At the process bus level, the UGPSSMs act as Mergng Unts (MUs) that measure analog quanttes, convert the data nto dgtal and transmt the dgtal data to the process bus. At the bay level, there are Intellgent Electronc Devces (IEDs) that process the data and execute protecton functons, control, or both functons. In addton the IEDs wll transmt the processed data to the state estmator va the staton bus. Specfcally the ndcated computer wth ts 3

4 own Human Machne Interface (HMI) wll collect the data from the staton bus (.e. from all IEDs) and feed t to the state estmator operatng n the ndcated computer. Note that n general, the data traffc s much hgher at the process bus (sampled data) than the staton bus. Therefore, the process bus should be equpped wth communcaton meda of hgher data rate such as Gga bts per second (Gbps), and wth smpler communcaton protocols. Below we provde some addtonal descrptons of the components shown n Fgure 3. UGPSSM: Ths component s an Advanced Meter that performs the role of MU (mergng unt). It senses voltage and current analog data wth hgh samplng rate and converts the measured data from analog to dgtal. The dgtal data s sent to the process bus drectly or va a concentrator wth hgh data rates (Gbps). Wreless communcaton, especally, s used n ths data communcaton between UGPSSM and process bus. Ths tas requres a protocol that does not burden the data transmsson wth extra nformaton that wll reduce the throughput. Furthermore the UGPSSM has the capablty to perform GPS-synchronzed measurements and to provde accurate tme stamps and verfcaton of the GPS-synchronzaton. Concentrator: The concentrator connects one or more hgh speed channels to a sngle hgher speed lne. The data concentrator can also serve as a protocol converter.. I/O: Ths devce connects the concentrator to the process bus. IED: IEDs n bay level collects the measurement data as well as status data from UGPSSM, and then process the data to extract phasors, etc. Ths data (phasor data) are lower rate data and are streamed nto the staton computer va the staton bus for state estmaton and other functons. The real tme model derved by the state estmator can be used to determne control and protecton needs. For example, as t wll be descrbed later, the real tme model wll be used to optmze the operaton of the dstrbuton feeder for system control needs (such as loss mnmzaton, levelzng load, etc.) and to dentfy and protect aganst downed conductors. The computed controls are transmtted bac to the IEDs whch may ntate the control acton.. Protocol: When consderng the nteroperablty of devces and the avalablty of data, the global communcaton standards are defntely necessary. Recommended are two types of nternatonal communcaton standards n the feld of power system, IEC and/or IEEE C As mentoned earler, the newly developed protocol wll be used for communcaton between UGPSSM and concentrator. IEC 61850: IEC s an nternatonal standard for communcaton networs and systems n substatons, and provdes the nteroperable data communcaton between mult-vendor IEDs. Moreover, ths standard plays a pvotal role n hgh speed communcaton that enables accurate and fast montorng, protecton, and control over the smart grd. IEC specfes sampled values that UGPSSMs are sensng from nstrumentaton transformers, and thus communcaton between concentrator and IEDs follows ths protocol. In the proposed nfrastructure, the real-tme modelng depends on hgh samplng rate, and thus sampled values are tme crtcal. Accordng to IEC that defnes the types of messages, the transmsson of sampled values omts the transport layer and the networ layer n OSI seven layers; the data pacet of sampled values n applcaton layer s drectly mapped to data ln layer. Therefore data process tme s reduced by mang concse the Ethernet frame, whch eventually leads to faster data communcaton. IEC deals wth the communcaton between IEDs and local control center by mappng Abstract Communcaton Servce Interface (ACSI) to Manufacture Message Specfcaton (MMS). IEEE C (Synchrophasor): The IEEE standard for Synchrophasor data, C37.118, defnes streamng synchronzed phasor measurements. It covers Synchrophasor message formats, tme synchronzaton, and tme taggng of data for real tme communcaton wth other Phase Measurement Unts (PMUs) or Phasor Data Concentrators (PDCs). A phasor networ s used so that the measured data and nformaton can be easly accessed and nterfaced wth other networ composton elements. Phase measurement estmaton wll requre samplng the raw data over certan cycles and tme taggng at a certan tme. C defnes tme tag as the tme of the theoretcal phasor that the estmated phasor represents [7]. In order to synchronze phasor estmaton wthn 1 mcrosecond accuracy, a synchronzng source shall be provded. The presently acceptable opton s the Global Postonng System (GPS). Tme s manly provded by GPS, and IRIG-B format s commonly used. The C standard defnes the format of messages of PMUs for real tme data transmsson (streamng data). Snce we propose to use PMUs and PDCs for dstrbuton automaton, we also propose to conform to the IEEE standard C In ths paper, every UGPSSM s to transmt data to PDC, and then, data accumulated by PDC are sent to process bus. 4

5 Dstrbuton system state estmaton State Estmaton s expected to be the most crtcal functon n the proposed nfrastructure. We wll refer to ths nfrastructure as Advanced Dstrbuton Management System (ADMS). Its functon s to flter all the avalable measurement data and provde the operator wth an accurate and relable real tme model of the system. In the followng paragraphs we elaborate on the major components of a state estmaton,.e. models nvolved, state, measurements, state estmaton algorthm, detecton and dentfcaton of bad data and error analyss. Feeder Modelng: System mbalances and asymmetres are a common occurrence n dstrbuton systems due to sngle phase laterals and loads. In order for the estmaton algorthm to be able to capture these phenomena, the power system has to be modeled on a three phase base (ncludng neutral and grounds) usng the physcal model of each component. An example of ths approach s shown n Fgure 4. Phase Conductors Shelds/Neutrals Type Sze AAC COSMOS AAC PANSY Tower/Pole Type AGC-DP-1 Crcut Number 1 Structure Name D-POLE-A Tower/Pole Ground Impedance (Ohms) R = 50.0 X = 0.0 Get From GIS Lne Length (mles) Lne Span Length (mles) Sol Resstvty (Ohm-Meters) Bus Name, Sde 1 FDR Falure & Repar Rates Falure Rate (per year) 1.0 Repar Rate (per year) 1.0 Type Sze Crcut Number lne Insulated Shelds Transposed Phases Transposed Shelds Read GPS Coordnates Fgure 4: Dstrbuton Lne Physcal Model State Set: The state s defned as the mnmum nformaton that completely descrbes the operatng condtons of the dstrbuton system, so t s defned as the set of the voltage phasors of every node of the system. So for every three phase bus of the system the state s defned as: ~ ~ ~ ~ ~ = V V V V [ ] T V, A, B, C, N Bus Name, Sde POLE1 Operatng Voltage (V) 13.8 Insulaton Levels (V) FOW (Front of Wave) 390 BIL (Basc Insulaton Level) 80 AC (AC Wthstand) 150 Note that n a rectangular coordnates formulaton there are two states for each phasor voltage, the real and the magnary part. So the actual states are: [ V V V V V V V V ] T V =, A, R, A, I, B, R, B, I, C, R, C, I, N, R, N, I where subscrpts R and I denote Real and Imagnary part. Measurement Set and Model: The measurements can be categorzed nto GPS-synchronzed and nonsynchronzed measurements. The non-synchronzed measurements are avalable from dstrbuton SCADA, reclosers, capactor controllers, etc. The synchronzed measurements are avalable from the advanced meter descrbed n the secton Proposed Infrastructure. In our formulaton the measurements are expressed as nonlnear functons n terms of the states of order at most quadratc. The generc form of the measurement s as follows: z = c + a x + b j x x j +,,, η, j where z : s the measured value c : s the constant term a, : are the lnear coeffcents b, : are the nonlnear coeffcents η : s the error term The GPS-synchronzed phasor measurement set conssts of voltage and current measurements, both magntude and phase n all three phases. Usng a rectangular coordnate formulaton, these measurements can be expressed as lnear functons n terms of the states. On the contrary non-synchronzed measurements consst of non synchronzed voltage or current phasors, voltage or current magntudes, actve and reactve power flows. Such measurements are typcally obtaned va analog measurement devces and are n general related to the system state as a nonlnear functon (at most quadratc n our case). In addton to the actual measurements the approach s facltated by a number of pseudo-measurements that are properly ntroduced [4,5] State Estmaton Algorthm: The problem s defned as a statc estmaton problem where the objectve s to estmate the states x usng the least squares approach. Thus the problem s formulated as follows: T Mnmze J ( x) = η Wη, where: η = z h(x), and W s a dagonal matrx whose non-zero entres are equal to the nverse of the varance of the measurement errors: 5

6 W 1 dag. σ v = The best estmate of the system state s obtaned from the followng teratve algorthm: j+ ˆ 1 j T 1 T x = xˆ + ( H WH) H W( z h(ˆ x j )), where xˆ refers to the best estmate of the state vector and H s the Jacoban matrx of the measurement equatons. State Estmaton Accuracy Quantfcaton: The accuracy of the state estmate and the estmaton confdence levels s obtaned va the standard chsquare test and the statstcal propertes of the state estmator. The ch-square test defnes the probablty that the dstrbuton of the measurement errors are wthn the expected bounds. Gven the number of measurements m and the number of states n, the degrees of freedom can smply be calculated as ν=m-n. Calculatng the value ζ of the objectve functon based on the state estmates, the estmaton confdence level s gven by the probablty: Pr [ χ ζ ] = 1.0 Pr[ χ ζ ] = 1.0 Pr( ζ, v). For an acceptable confdence level, the accuracy of the soluton s computed va the covarance (or nformaton) matrx. The covarance matrx of the state s defned as x [( x x)( x x) ] T C = E ˆ ˆ, where x denotes the true state value and xˆ the estmated value, and computed as C T ( H ) 1 = WH x. Once the nformaton matrx of the soluton has been computed, the standard devaton of a component of the soluton vector s gven by σ (, ), x = C x a Scalablty: Our goal s to be able to apply ths nfrastructure on a dstrbuton system ndependently of the sze of the system, and wth no mpact on the performance and computatonal burden. As a result we propose a scheme where the state estmaton algorthm s performed n the substaton level. More precsely, typcal substatons have -1 feeders. By collectng all the measurements of the feeders n the substaton and gven the model of the feeders, the state estmaton algorthm can be executed n the substaton level usng just a hgh end personal computer. Then, only the results of the analyss, that s the estmated states, are sent to the DMS where they can be processed along wth the results from other substatons and reconstruct the real tme model of the whole system that the operator s nterested n. The advantages of ths approach are tremendous snce ths decentralzed approach leads to mnmzaton of communcatons burden and latences and also the computaton tme s mnmzed. Fnally, the state estmator problem s much smaller n sze and therefore powerful hypothess testng methods can be appled for both bad data and topology errors dentfcaton wthout substantal deteroraton of the computatonal effcency, whch s not the case n a centralzed approach due to enormous number of hypotheses. Applcaton: downed conductor protecton The proposed nfrastructure s the enabler to solve a dstrbuton protecton problem for whch we presently do not have acceptable solutons and result n numerous fataltes each year. The problem s that of an open conductor. Occasonally, a phase conductor breas and falls on the ground wthout mang contact wth the neutral conductor, as t s llustrated n Fgure 5. In general the contact mpedance of a downed phase conductor wth the ground (sol) s relatvely hgh resultng n fault currents that may be below load currents. In ths case a downed conductor remans on the ground energzed untl someone nterrupts the crcut or an accdent may happen. where C x (, ) s the th dagonal entry of the C. x The estmates of the measurements can also be computed as: b ˆ = h( xˆ, yˆ). and ther covarance matrx s proved to be: T 1 T ( H WH ) H Cov ( bˆ) = H. 6

7 Fgure 5: A Downed Energzed Phase Conductor Efforts to dentfy these condtons and nterrupt the crcut resulted n systems that are ether prohbtvely expensve or not 100% relable. In the early 90s the authors developed a specal relay that wll detect a downed conductor []. The relay was relyng on communcatons va the neutral and therefore ts proper operaton requred that the neutral s ntact, a reasonable assumpton. In evaluatng the technology [] the authors recognzed that the deployment of ths technology as a dedcated system wth the only functon of detectng downed conductors s relatvely expensve and suggested that dstrbuton automaton may provde a better approach. Smart grd actvtes have superseded dstrbuton automaton. In partcular the proposed smart grd nfrastructure can provde an excellent 100% soluton to ths unsolved protecton problem. Ths s accomplshed as follows. The proposed nfrastructure provdes frequent updates of the state of the system. Therefore the electrc current flow and voltage s computed at each node of the system. When a downed conductor occurs the voltage on the source sde of the conductor wll be stll hgh whle the electrc current wll be relatvely small. On the other sde of the downed conductor the electrc current wll be practcally zero and the voltage abnormal (n general small). Note that the state estmaton wll dentfy ths condton and t wll explan the measurements as a dscontnuty on the phase conductor. It wll also dentfy the phase (or phases n case there s the unusual case of mult-down-conductors). The proposed state estmaton wll also provde the confdence level by whch ths condton s dentfed. Once the condton s dentfed the proper control wll be generated to trp the nearest upstream protectve devce. Applcaton: load levelzaton The proposed nfrastructure s the enabler to effectvely acheve levelzaton of the dstrbuton system electrc load wthout ncurrng nconvenences to the customers. Nowadays the electrc utlty nfrastructure costs are drven prmarly by the need to serve the load durng the pea demand perod. Therefore, t s desrable to shave pea demand, or levelze the dstrbuton system electrc load n order to defer generaton, transmsson and dstrbuton equpment upgrades, and reduce or avod the necessty to dspatch hgher cost generaton assets. A common and practcal way to acheve a reducton n pea load s the applcaton of battery energy storage systems [11,1]. Levelng load nvolves storng electrc energy n a battery durng the off-pea perod, and extractng t durng the pea perod. Wth the rapd development of the renewable energes, such as photovoltac arrays, there s also a trend to store the energy generated by the renewable technologes n the batteres and use them durng the pea load perod. Another alternatve for levelng the dstrbuton system load s through demand-sde management technques such as drect load control of resdental systems by nstallng the so called smart applances whch can turn the load off durng the pea load perod when they receve a dsconnecton command from the substaton level control center. In our approach we merge the above approaches together to get a better load levelzaton result. However, n order to acheve ths goal t s mperatve to have a hgh fdelty montorng system whch can provde contnuous and accurate nformaton of the dstrbuton system status and operatng condtons. Such nd of nformaton s crtcally essental for mang decsons on when to levelze the load and how we can coordnate the approaches we mentoned prevously to mnmze the pea load. Apart from real tme nformaton of the current status and operatng condtons of the dstrbuton system, t s also crtcal to have an, as accurate as possble, evaluaton of the dstrbuton system loadng throughout the plannng perod. Ths mples that short term load forecastng s another essental component n the proposed nfrastructure for levelng the dstrbuton system load. The nfrastructure we proposed at the prevous secton wll allow contnuous and accurate montorng of dstrbuton system operatons and customer utlzaton of electrc power, and then the levelzaton of the dstrbuton system load can be treated as an optmzaton problem. The goal of the optmzaton problem s to mnmze the pea load of a dstrbuton system for the whole day perod usng the approaches we mentoned prevously. Snce the load of the dstrbuton system has some statstcal characterstcs and wll change over the tme, we dvde the tme perod (one day) nto n small tme ntervals as t 1, t,, t n and we assume that the load wll not 7

8 change durng each tme nterval. Based on these, the optmzaton problem can be formulated as follows: * mn X where: X X * * g( x( t ~ ~ * > Re( V1 ( t1) I1 ( t1)) ~ ~ * > Re( V ( t ) I ( )) ), u( t 1 ), L( t n 1 t n s. t g( x( t1), u( t1), L( t1)) = 0 n n n )) = 0 Constrants Set of Controllable Devces (DG, PHEVs, storage etc) Our goal s to mnmze the pea load, so the objectve s to mnmze the mum value of the loads of the dstrbuton system over all the tme ntervals. The constrant equatons: g( x( t ), u( t ), L( t )) = 0, = 1,, n represent the power flow equatons of the dstrbuton system for each tme nterval and the vectors x t ), u t ) and L t ) ( ( ( represent the state varables, control varables and the statstcal load varables of the dstrbuton system respectvely. Fnally, the constrants set at each tme nterval represents the operaton constrans of the devces we used for the levelzaton of the load, such as the energy n the batteres of a PHEV, the photovoltac system status, the status of smart applances, etc. Each specfc devce wll have a specfc and detaled constrant model for the optmzaton problem. Below we present the constrant model of a battery storage as an example and we assume that the batteres are connected to the dstrbuton feeder through a converter. * ( ) ( ) = ( ) + ( ) ( ) s( 1 ) = s( ) Δ + s( ) s, s( ) s, Qs,mn Qs( t) Qs, 0 E ( t ) E s s, ( ) ( ) P t Q t I s t ubs t jubs t V t V t E t P t t E t st : P P t P t : the tme nterval ndex s s s s Δ t : the tme between 1 t and t u bs : the status of the swtch of the storage battery. V s : voltage of the node the storage devce s connected wth. P, Q : the real and reactve power of the storage s s devce E : the energy stored n the storage devce. s P, P : the mnmum and mum real power s,mn s, the storage devce can provde. Q, Q : the mnmum and mum reactve s,mn s, power the storage devce can provde. E s, : the capacty of the storage devce. The defnton of the state varables x and control varables u for the storage devce s as follows: x = V s, u bs u = [ P, Q, E ] s s s Note that the soluton of above problem provdes the requred controls to levelze the load for each resource n the system. Ths approach s qute dfferent from the more smple approach of sendng prce sgnals to customers and expect that the customers wll respond to prce sgnals. The prce sgnals are nown to generate unwanted responses such as shftng of peas nstead of levelzng the electrc load. Applcaton: loss mnmzaton Transmsson losses across the dstrbuton feeders, s one of the major ssues for dstrbuton systems engneers snce the exstence of power systems. Typcally they vary from 3%-8% and can be categorzed nto ohmc losses, losses from reactve power flow and losses due to the harmonc currents that result from nonlnear loads of the system. Mnmzaton of the dstrbuton losses s one of the major goals of the Smart Grd concept that can result n better performance of the networ wth the upper goal beng, a low cost and envronmentally frendly energy utlzaton. The most common practce that s used nowadays n order to reduce dstrbuton system losses s Volt/Var control. Var compensaton s mplemented nowadays by the use of capactor bans that are placed n crtcal buses of the system and supply reactve power to support and optmze the voltage profle of the system. Real tme control actons can be mplemented up to some extend through swtched capactor bans; however these are placed only at dscrete ponts of the 8

9 system and nject dscrete levels of reactve power. Moreover the control actons are based on local nformaton (local voltage level). The proposed nfrastructure s the enabler to effectvely mnmze dstrbuton system losses n a manner transparent to the end user. A three phase state estmaton algorthm as descrbed n secton Proposed Infrastructure that s performed multple tmes per second usng GPS synchronzed measurement data obtaned from AMI, enables real tme montorng of the feeders wth extremely hgh accuracy and confdence level. The mportance of the three phase formulaton s more unequvocal n ths applcaton, snce a sgnfcant amount of the losses result from system mbalances and asymmetres that cannot be captured usng conventonal state estmaton approaches based on sngle phase, postve sequence equvalents. Upon beng able to montor the system n real tme, thus beng fully aware of the state of the system wth a confdence level close to 100%, the proposed nfrastructure opens up the capablty of the applcaton of a coordnated Volt/Var control scheme throughout the length of the feeder, deally at every bus, usng not only utlty resources (capactor bans) but also customer resources. And ths s applcable because one of the ey concepts of the Smart Grd are the dstrbuted renewable sources and storage devces even down to the house level, whch wll connect to the grd va nverters that have real tme voltage and power factor control capablty. Volt/Var control can be exercsed based on an optmzaton algorthm that wll use the results of the state estmaton, wth the objectve beng the mnmzaton of losses. The problem can be formulated as an optmal power flow problem, wth ts general form beng as follows: mn f (x) : Total Losses s.t g ( x, u) = 0: Power Flow Equatons h ( x, u) 0 : Functonal Constrants u u u : Control constrants, mn, where x are the state varables and u the control varables. Total losses can be expressed n terms of the system states.e. the node voltages as follows: ~ ~ * ~ ~ * f ( x) = Re{( V I + V I )} phase, phase, phase where : crcut connected to buses, j j, phase j, phase Fgure 6: Dstrbuton Lne Losses Computaton (n Green) The power flow equatons can be expressed based on the quadratc power flow formulaton as descrbed n [10]. Functonal constrants nclude lmtatons such as voltage magntude of phase angle of buses. Fnally control constrants relate to dstrbuted resources ( PVs, PHEVs, storage devces) such as power factor varaton lmts, actve power capablty or energy storage lmts. Table 1 summarzes typcal constrants that are ncluded n the formulaton of the optmzaton algorthm. TABLE 1 CONSTRAINTS OF THE LOSS MINIMIZATION OPTIMIZATION PROBLEM Functonal Constrants Control Constrants Bus Voltage Magntude/ Angle Crcut loadng Avalable Actve and Reactve Power from PVs, PHEVs, Wnd Turbnes, etc Energy Storage lmts of storage devces Inverter power factor control lmts Consder for example a PV system that s connected to bus of the system va an nverter. Then the followng constrants should be satsfed: ~ ~ V * I = u1 P + Pmn u1 P Q u Q ε mn mn V j u Q P Q u Q ε u1 P ~ V V,mn 0 u u 1, ISV 9

10 where u 1, u : control varables desgn ε,ε mn : constants dependng on the nverter Tang nto account all the operatng constrants of all dstrbuted generaton devces, renewable energy sources, storage devces that are connected to the grd, t s clear that we deal wth a large scale and challengng optmzaton problem. However the control defned wth ths problem mnmze the losses n the dstrbuton system usng the avalable resources. Applcaton: relablty enhancement Another applcaton of the proposed nfrastructure s relablty enhancement. Snce the proposed scheme provdes the real tme of the dstrbuton feeder wth hgh fdelty and n case of a fault dentfes the locaton of the fault, t s relatvely easy to optmze the relablty of the system n real tme,.e. to respond extremely fast. Specfcally, for a gven condton (fault at a specfc locaton, breaer and swtch status, etc.) one can determne the optmal reconfguraton of the system n order to mnmze the affected customers and to perform these actons n an automatc way usng the proposed nfrastructure. We wll refer to ths problem as relablty enhancement and the formulaton of the problem s descrbed below. Ω S L stf.. ( V, P, Q ) = 0 V V V P P P Q Q Q I I j G G mn mn G G G mn G G G j j, Ω Where F s the power flow functon n the affected area, S L s the load at bus, V s the voltage magntude at bus, P G s the actve power generaton at bus, Q G s the reactve power generaton at bus, I j s the current flowng between bus and bus j, Ω s the set of buses n the affected area. The objectve of ths optmzaton problem s to mze the total loads that can be connected to the grd n the affected area, mprovng relablty of the grd as much as possble. The problem s subject to the power flow restrcton, bus voltage regulaton, lmtatons of the actve and reactve power output of avalable DGs (dstrbuted generatons) n the affected area as well as the constrants of transmsson lnes. The soluton can provde us the optmal set ponts of DGs and the detaled nformaton of loads that the grd s able to supply. To mze the ablty of the proposed system to enhance the relablty of the dstrbuton system t wll be expedent to place swtches, reclosers n strategc locatons of the system (ths can be done n the desgn phase). Reference [8] provdes a method for optmal swtch placement. Conclusons Ths paper descrbed a comprehensve approach for a smart grd mplementaton on a dstrbuton system. The basc objectve s that the nfrastructure should enable all the desrable functons of optmzng the operaton of the dstrbuton feeder for mum beneft to utltes and customers ale. These goals are achevable only va a system that wll enable accurate and frequent montorng of the dstrbuton system. The montorng functon extends to the customer and ncludes montorng and control of customer resources. A requrement placed on the proposed nfrastructure s that the customer should not be nconvenenced. The proposed scheme acheves the goals of the smart grd wth mnmal or no nconvenence to the customer. Implementaton of the proposed scheme wll requre the development of low cost meters that wll have the capabltes descrbed n the UGPSSM. Presently technology exsts to massvely manufacture these meters at very low cost. References 1. A. P. Sas Melopoulos and Fan Zhang, Multphase Power Flow and State Estmaton for Power Dstrbuton Systems,' IEEE Transactons on Power Systems, Vol. 11, No., pp , May Art Westrom, A. P. Sas Melopoulos, G. J. Condes, and A. H. Ayoub, "Open Conductor Detector System," IEEE Transactons on Power Delvery, Vol 7, No. 3, pp , July B. Fardanesh, S. Zelngher, A. P. Sas Melopoulos, G. Condes and Jm Ingleson, Multfunctonal Synchronzed Measurement Networ, IEEE Computer Applcatons n Power, Volume 11, Number 1, pp 6-30, January A. P. Sas Melopoulos, George J. Condes, Floyd Galvan, Bruce Fardanesh and Paul Myrda, Delverng Accurate and Tmely Data to All: Model Based Substaton Automaton Applcatons for Advanced Data Avalablty, IEEE Power & Energy Magazne, Volume 5, No: 3, pp 74-86, May/June

11 5. Sas Melopoulos, George Condes, George Stefopoulos, Terry Conrad and Clnton Hedrngton, Dstrbuted State Estmator va the SuperCalbrator Approach, Protecton, Automaton and Control World, pp 38-44, Autumn Communcaton Networs and Systems n Substatons, IEC 61850, 1st ed., IEEE Standard for Synchrophasors for Power Systems, Std C , IEEE Power Engneerng Socety. 8. Ymng Mao and Karen N. Mu, Swtch Placement to Improve System Relablty for Radal Dstrbuton Systems wth Dstrbuted Generaton, IEEE Transactons on Power Syetem, Vol 18, No. 4, pp , November A. P. S. Melopoulos and G. K. Stefopoulos, Characterzaton of state estmaton bases, Proceedngs of the 004 Internatonal Conference on Probablstc Methods Appled to Power Systems, pp , Ames, IA, Sept. 1-16, A. P. Sas Melopoulos, State estmaton for mega RTOs, presented at the 00 IEEE Power Engneerng Socety Wnter Meetng, New Yor, NY, Jan. 7-31, I. Papc, Smulaton model for dschargng a lead-acd battery energy storage system for load levelng ; IEEE Transacton on Energy Converson, Volume 1, Issue, June 006, Pages: C. Venu; Y. Rffonneau; S. Bacha; Y. Baghzouz, Battery Storage System szng n dstrbuton feeders wth dstrbuted photovoltac systems, IEEE PowerTech Conference, Bucharest, Romana, A. P. Sas Melopoulos, S. W. Kang, G. J. Condes, R. Dougal, Anmaton and Vsualzaton of Spot Prces va Quadratzed Power Flow Analyss, Proceedngs of the 36 th Hawa Internatonal Conference on System Scences (HICSS 03) 14. K.M. Rogers; R. Klump; H. Khurana; T.J. Overbye; "An Authentcated Control Framewor for Dstrbuted Voltage Support on the Smart Grd", IEEE Transactons on Smart Grd, Vol. 1, No. 1, pp , Jun F. L; W. Qao; H. Sun; H. Wan; P. Zhang; "Smart Transmsson Grd: Vson and Framewor", IEEE Transactons on Smart Grd, Vol. 1, No., pp , Sep R. E. Brown; "Impact of Smart Grd on dstrbuton system desgn", IEEE Power and Energy Socety General Meetng, Pttsburgh, PA, July M. Ppattanasomporn; H. Feroze; S. Rahman; "Multagent systems n a dstrbuted smart grd: Desgn and mplementaton", IEEE Power Systems Conference and Exposton, Seattle, WA, A. Ipach; "Smart grd of the future wth large scale DR/DER penetraton", IEEE Power Systems Conference and Exposton, Seattle, WA, A. Chuang; M. McGranaghan; "Functons of a local controller to coordnate dstrbuted resources n a smart grd", IEEE Power and Energy Socety General Meetng, Pttsburgh, PA, July V.K. Sood; D. Fscher; J.M. Elund; T. Brown; "Developng a communcaton nfrastructure for the Smart Grd", IEEE Electrcal Power & Energy Conference (EPEC), Montreal, QC, D.G. Hart; "Usng AMI to realze the Smart Grd", IEEE Power and Energy Socety General Meetng, Pttsburgh, PA, July