Optimal Management of Microgrids

Transcription

1 Universia Poliècnica de Caalunya Facula de Maemàiques i Esadísica Maser hesis Opimal Managemen of Microgrids Lucía Igualada González Advisor: F. Javier Heredia, Crisina Corchero (IREC) Deparmen of Saisics and Operaions Research, Research Group GNOM (Group on Numerical Opimizaion and Modeling)

2

3 To he only hing ha I can believe: he Mahemaics, he Science...

4 Preface The proec of his hesis was carried ou in he IREC (Insiu de Recerca en Energia de Caalunya) in he Area of Elecricia i Elecrònica de Poència wihin he group of Energy Economics, under suppor of he KIC Innoenergy wihin he framework of he KIC-EVCITY proec.

5 Agradecimienos Mis más sinceros agradecimienos: v Al n dimensional mundo de las Maemáicas por enseñarme el camino hacia la Invesigación Operaiva. A odos los profesores, ano de la Licenciaura como del Máser, que han marcado algún puno y a pare en mi camino. A Vicor Jiménez porque núnca le agradecí aquella charla y Alfredo Marín por adornar odos aquellos grafos con su humor paricular. Del Máser quiero hacer una mención especial a F.Javier Heredia y Crisina Corchero no solo por ser mis uores y iempo dedicado, si no además por animarme a omar la decisión que ha dado como fruo ese rabao. Al IREC, y su grupo de Energy Economics por darme la oporunidad de hacer mi TFM con ellos y enrar en el mundo de la energía, mercados y microgrids. Agradecer a Miguel Cruz su paciencia, aporaciones y ánimos. Y por supueso a odos esos ingenieros del área de Elecricia i Elecrònica de Poència que me han enseñado desde diferenciar poencia de energía a aprender nuevos lenguaes de programación. Por responder a odas mis pregunas sobre esas cosas an básicas, por odos los desayunos, risas, fricadas varias, guerras por el hueco bao la mesa y acividades fuera de la oficina que han hecho esa experiencia más ineresane y claro que sí, diverida. A esa muer admirable llena de hisorias que conar y que me abrió las pueras de su casa nada mas llegar. Gracias Paquia. A esa nueva familia laina que he enconrado enre los compañeros de clase. Colombianos, mexicanos y venezolanos, gracias por darle rimo a mi esancia en Barcelona. Y en especial a Anonio por aguanarme día a día, animarme y consenirme. A los amigos de siempre por seguir presenes a pesar de la disancia, sobre odo a mis peardis por hacer mis pequeñas vacaciones en el sur siempre más diveridas. Al LOL y sus rolls por ser mi más inservible disracción y alimenar a la pequeña friki que llevo denro (o no an denro). Y por úlimo, mis padres. Por acepar la locura de venirme sin nada y con an poco iempo de anelación, por hacer posible mi esancia aquí y por la mochila presada. En definiiva, por ayudarme a seguir mi camino y animarme siempre ane las peores noicias del mundo. Gracias.

6

7 Absrac Keywords: Smar grids, microgrid, energy sysem opimizaion, elecric vehicle Curren mankind is facing a global dilemma wih energy demand increasing, while diminishing radiional energy resources. Increase energy efficiency and susainabiliy are becoming more necessary. In his framework, smar grids and microgrids are he key in he near fuure where a decenralizaion of energy generaion is expeced. An advanage of hese ype of grids is ha balancing beween energy generaion, sorage, and consumpion can be realized mos efficienly he closer he physical locaion of generaion, sorage and demand is he conroller. This reduces he need for cenralized communicaion, enables auonomous operaions of increasingly smaller secions of he disribuion grid and decreases he losses by disan disribuion. Wihin his framework and from he poin of view of microgrid energy managemen, economic scheduling for generaion devices, sorage sysems and loads is a crucial problem. Performance an opimizaion process is necessary o minimize he operaing coss while several operaional consrains are aken ino accoun. Energy managemen is carried ou by MCC (Microgrid Cenral Conroller) in hree seps: eriary, secondary and primary conrols. The firs managemen sep is execued one day-ahead and has wo obecives. The firs is economic opimizaion using a program based on an Economic Dispach and an Uni Commimen problem. The second obecive is o improve he profiabiliy of he supply and demand balance by ineracing wih he grid and aking advanage of he V2G (vehicle o grid) capabiliy of he charging spo, and o generae a schedule over all componens of he microgrid. The res of he conrols are execued on day of operaion in order o adus he oupu power levels. The secondary conrol receives he scheduling plan creaed by eriary conrol and aking ino accoun curren daa, correcs he power oupus of generaion unis.exchanged power wih he grid and sorage saes of charge programmed by he eriary conrol are ensured.finally, he primary conrol regulaes he energy flow in real ime and ensures a proper operaion o address any unexpeced issues alhough, his conrol is no considered in he proec. The Energy Managemen Sysem has been esed over differen scenarios. One of hem is based on a smar house wih a phoovolaic module, a micro wind urbine and one elecric vehicle charging spo. The household load can be divided ino hree differen profiles: criical, adusable and shifable loads. The seleced profiles, boh mobiliy of EV and household load, have been measured during a working day. The oher scenario is based on a large building where one micro gas urbine and one sorage device have been added o he res of unis. Afer analysing he resuls, several conclusions have been deduced such as a change in curve of load and a lower cos for he user. Generally, he consumpion over peak periods is decreased or is almos zero in some es cases, while he demand overnigh is increased.

8 Conens Chaper 1. Inroducion 1 1. Moivaions and obecives 3 2. Conens 3 Chaper 2. Microgrid s regulaory environmen 5 1. Presen siuaion 5 2. Trend oward Disribued Energy Sources (DER) 7 3. Prospecive 8 Chaper 3. Microgrid componens modelling Technical specificaion of microgrid componens Mahemaical modelling of microgrid s componens 15 Chaper 4. The microgrid cenral conroller opimizaion problem Overview of he microgrid cenral conroller problem Teriary conrol opimizaion problem Secondary conrol opimizaion problem 29 Chaper 5. Implemenaion and resuls Implemenaion The IREC s microgrid emulaor Tes cases Resuls for he eriary conrol problem Resuls for he secondary conrol problem 47 Chaper 6. Conclusions and furher research 51 References 53 i

9 Chaper 1 Inroducion The obecive of his proec is o develop an energy managemen algorihm for microgrids hrough wo opimizaion problems. To guaranee a beer undersanding of he obecive, some conceps and definiions associaed wih his framework should be explained. The firs wo conceps are DG (Disribued Generaion) and DER (Disribued Energy Resources). Mos counries generae elecriciy in large cenralized faciliies, such as coal, nuclear or hydropower plans. However, in he fuure a significan porion of energy consumpion may be supplied by domesic or small scale generaion. Disribued generaion, also called on-sie generaion or decenralized energy, generaes elecriciy from many small energy sources and in mos cases, from small renewable sources causing lower environmenal impac. Disribued generaion improves supply securiy due o he reducion in he amoun of energy los in ransmiing elecriciy because he elecriciy is generaed near where i is required, perhaps even in he same building. This also reduces he size and number of power lines ha may be insalled and mainained. DER sysems are he conrollable load, he power generaion echnologies and sorage devices in small-scale (2 kw o 10 MW range). Curren connecion pracice is based on fi and forge approach. This policy leads o inefficiency and cosly invesmen in disribuion infrasrucure. A he momen, DER can only displace energy produced by cenral generaion i.e., decrease he energy amoun generaed by radiional generaion unis, bu canno displace/decrease insalled generaion capaciy. The lack of conrollabiliy of DER implies ha sysem conrol and securiy mus coninue o be provided by cenral generaion. However, previous proecs have demonsraed ha he implemenaion of a microgrid sysem will provide a solid framework for a full inegraion of DG and demand. Anoher key facor is he smar grid. The erm smar grid refers o a modernizaion of he elecriciy delivery sysem so ha i moniors, proecs, and auomaically opimizes he operaion of is inerconneced elemens. These elemens include he cenral and disribued generaion hrough he high volage ransmission nework, indusrial users and building auomaion sysem, energy sorage insallaions, and end use consumers and heir hermosas, elecric vehicles, appliances, and oher household device [2]. 1

10 2 1. INTRODUCTION When hese conceps are assimilaed, undersanding he Microgrid definiion is clearer. A microgrid is a localized group of disribued energy recourses ha can be operaed coordinaely as an energy generaor, as an energy sorage and as a load. I normally operaes conneced o a radiional cenralized grid (macrogrid or general grid) o provide maximum elecrical efficiency wih a minimum incidence o loads profile in he local power grid [8]. Such sysems can someimes be operaed in emergency operaion, also called islanded mode, if he single poin of common coupling wih he macrogrid has been disconneced. Alhough a microgrid is capable of handling boh saes, is main benefis will arise from grid-conneced [9]. In addiion, his microgrid concep serves muliple economic, echnical, and environmenal aims, and reduces he number of inermediary paries. An example of microgrid is shown in Figure 1: L.V. Grid Macrogrid Microgrid Elecric Vehicle LC Conrollable load LC Criical load LC Solar Power MC Wind Power MC MT gas MC MC Sorage Temperaure and solar irradiaion Wind Speed Microgrid Cenral Conroller - + Fig. 1. Microgrid diagram. The diagram shows generaion unis, load sources, a sorage device, he conroller for each componen, he poin of common connecion and he connecion wih he grid. All microgrid componens will be explained in he hird chaper.

11 1. Moivaions and obecives 2. CONTENTS 3 The origin of his proec lies in he imporan role ha microgrids could play in he near fuure due o iniiaives where sraegies focus on more efficien elecriciy use. One of hese iniiaives is he 2020 Sraegy in he European Elecrical Secor where commied o CO 2 reducion, increasing renewables in he generaion mix and efficien energy managemen. Included in sraegies for CO 2 reducion is he promoion incremen of EV (elecric vehicle) use and how, a he same ime, i could also help he inegraion of renewable power sources such as wind power by soring he energy surplus produced during windy periods and providing i back o he microgrid during high load periods. To ake advanage of hese acions, a V2G sysem is necessary. V2G is a sysem in which plug-in elecric vehicles communicae wih he power grid delivering elecriciy ino he grid or decreasing or shifing heir charging rae. Thus, aking ino accoun he possible benefis provided by elecric vehicle use and inegraion in microgrids, he following obecives have been se o conribue o european energy iniiaives. General obecive: Develop an opimal energy managemen algorihm for he IREC microgrid faciliies in order o improve he research abou hese small grids. Specific obecives: To formulae problems associaed wih he differen conrol levels of he microgrids. To obain a GAMS based implemenaion for he off-line soluion of he eriary opimal conrol problem. To develop a C based implemenaion for he on-line soluion of he secondary microgrid opimal problem. To implemen he algorihm for he secondary opimal conrol problem in he microgrid s emulaor of he IREC. To use he opimal energy managemen algorihm o es he inroducion of he microgrids in he energy marke and, a he same ime, o sudy he elecric vehicle as a new componen in he microgrid. 2. Conens The documen srucure is as follow: Chaper 1: Inroducion. This chaper inroduces all he basic conceps needed o undersand wha is a microgrid and is framework. The second secion oulines he moivaions of he proec and is obecives. The las par of he chaper summarizes of he conens.

12 4 1. INTRODUCTION Chaper 2: Microgrid s regulaory environmen. This chaper presens a brief inroducion of he Spanish energy marke and is curren regulaion associaed wih microgrid framework. The chaper ends wih a prospecive of a new agen for he paricipaion of microgrids in he energy marke. Chaper 3: Microgrid s componens modelling. Chaper 3 defines he considered componens of he microgrid included in he es cases. For each componen is presened is definiion, main characerisics and in he las secion, is modelling. Chaper 4: The microgrid cenral conroller opimizaion problem. This chaper includes a brief inroducion of microgrid conroller and he opimizaion problems associaed wih hem. For he wo problems, eriary and secondary conrol, he mahemaical formulaion is presened considering he microgrid s componens modelling. Chaper 5: Implemenaion and resuls. Chaper 5 includes he opimal managemen algorihm, a descripion of microgrid s emulaor of he IREC and he es cases. There are wo main es cases splied ino hree cases wih differen hypohesis. Finally, he resuls obained wih he algorihm for each es case and a comparison beween hem, are commened. Chaper 6: Conclusions and furher research. General conclusions of he resuls generaed by his proec and furher research are discussed. Appendix. This appendix provides a glossary where he symbols and abbreviaions used in his proec are described.

13 Chaper 2 Microgrid s regulaory environmen 1. Presen siuaion The curren siuaion in he Spanish Elecriciy Marke has been based on a liberalized sysem since December 1997, when he Elecriciy Secor Ac provided for he creaion of an independen marke operaor ha would uphold a liberalized sysem for he generaion and supply of elecriciy. The organizaion of he elecriciy marke was sipulaed hrough Law 54/1997 and Royal Decree 2019/1997 whose basic principle was o allow o rade for producers, resellers and consumers. The Minisry of Economy mus approve he rules and erms governing he operaion and selemen of he producion marke according o European regulaions. In addiion since 1 January 2003, consumers have acquired free marke saus and can herefore choose heir supplier. To paricipae in he MIBEL (Mercado Ibérico de la Elecricidad) here are wo main ways. The firs opion is he rading by bilaeral conracs beween he generaion company and a qualified consumer. The conrac sipulaes he elecriciy amoun ha he company mus supply, he price and he duraion of he delivery period. The second and mos imporan way are he regulaed markes, hey consis of seven sessions where he agens involved in he elecriciy marke can paricipae o buy or sale energy. The firs and main session is he daily marke, followed by six subsequen sessions o he inraday markes disribued hroughou 24 hours of he day. In addiion, a process of echnical managemen enables rading o guaranee he securiy and reliabiliy of he elecriciy sysem. The operaion of he naional elecriciy sysem depends on wo independen eniies he MO (Marke Operaor) and he SO (Sysem Operaor). The main role of each operaor is economic managemen for MO (in Spain he Operador del Mercado Ibérico de la Energía Polo Español, S.A. OMEL) and echnical managemen for SO (Red Elécrica de España, S.A). The coordinaion beween hese wo eniies is essenial in order o guaranee ha marke ransacions are physically feasible and fulfill securiy crieria. In he Figure 1 he general diagram of he presen Spanish Energy Marke is shown wih he differen agens, operaors and he possible flows beween hem: 5

14 6 2. MICROGRID S REGULATORY ENVIRONMENT Fig. 1. General diagram of Spanish Energy Marke. Included in generaion unis framework, i is imporan o ake ino accoun he special regime o inroduce microgrids in he presen framework. The main difference beween ordinary and special regimes is he opporuniy o choose beween wo differen economic mechanisms. One opion is o ransfer all power o he sysem hrough disribued and ranspor neworks, receiving a regulaed ariff depending on he generaion echnology. The generaion unis in special regime can sell he power in he MIBEL as in he case of ordinary regimes. In his case, he price will be he price resuled in he organized marke, complemened wih a premium depending on he echnology. The ype and power of elecrical insallaions which can operae in a special regime are regulaed in aricle 79.7 of he law wih number 54/1997. However, he possibiliy of a regulaed ariff is no available for new insallaions, pending a new regulaion.

15 2. TREND TOWARD DISTRIBUTED ENERGY SOURCES (DER) 7 2. Trend oward Disribued Energy Sources (DER) 2.1. European framework. Since 1991 when he firs energy efficiency program was approved by he European council, he laws concerning savings and energeic efficiency have been linked o he developmen of he EU, is energy policy and he requiremens derived from climae change. The legislaions wih more feaures in his sphere have been he acion plan of 200/2006 and he Green Book of 2006 abou European sraegy for energy where he Commission noes ha he EU can reduce energy consumpion expeced for 2020 by 20%. This level of energy savings will srenghen he compeiiveness of European indusry and will allow he EU o mee is Kyoo commimens and reduce CO 2 emissions. In January 2008, he European Commission presened a draf Direcive on he promoion of he use of energy from Renewable Energy Sources (RES) which conains a series of elemens o creae he necessary legislaive framework o enable a 20% renewable energy share o become a realiy. Direcive 2009/28/EC of he European Parliamen and he Council of April 23, concerns he promoion of he use of RES and saes ha each Member Sae shall esablish a Naional Acion Plan on Renewable Energy (NAPRE) o achieve he naional arges se in he Direcive. For his, he Plan de Accion Nacional de Energias Renovables has been creaed in Spain, where he framework and aims for each ype of renewable echnology are se. The evoluion of energy consumpion in Spain, rising oil prices and he inensificaion of he savings plans and energy efficiency have been aken ino accoun in preparing an energy map for In more deail, i is anicipaed ha in 2012 he paricipaion of renewable energies will be 15% (compared o he expeced value 11%) and in 2016, paricipaion will be 18%. Wihin he Acion Plan, he increase of NZEB (ne zero energy buildings) will be adoped aking ino accoun he approval of he European direcive (Direcive 2010/31/UE) ha requires all new buildings buil, from 31/12/2020, mus be of his ype. A definiion of NZEB can be: a building or insallaion which creae as much energy as i consumes, by on-sie renewable energy, hereby providing a significan reducion in boh demand and emission Spanish framework. To guaranee he European arges, wo Royal Decrees have recenly been approved. RD (Royal Decree) 1699/2011 concerns he managemen and echnical condiions for connecion o grid. In his RD is regulaed he gradual enry of faciliies of elecrical energy producion of low power (Microgrids), o modify he cenralized model by promoing a new sysem of increasingly disribued generaion. This change will involve lower losses of energy in he grid, he reducion of economic invesmen in ranspor and disribuion and primary energy savings; and for consumers, a beer elecrical auonomy and securiy of supply.

16 8 2. MICROGRID S REGULATORY ENVIRONMENT Possible he mos imporan concep regulaed in his RD is he possibiliy of working island, a main characerisic of microgrids. Abou his, aricle 12 says: 1. Connecion schemes should respond o he principle of minimizing losses in he sysem, favoring he mainenance of securiy and qualiy of supply and enabling he island o work on is own consumpion, never feed oher nework users. Moreover, he creaion of a new royal decree abou own consumpions is menioned in he Aricle 18 of his RD: The Miniser of Indusry, Tourism and Trade, wihin four monhs afer he enry ino force of his Royal Decree, will raise a royal decree proposal o he Governmen regulaing he adminisraive, echnical and economic elecrical energy consumpion produced wihin he grid of a consumer for heir own consumpion. And he possibiliy of exporing and imporing power a he same ime bu independenly: In cases where he producion faciliy will sell only he surplus energy will be allowed he opion of insalling a single measuremen device wih records independen of generaion and consumpion. In his case, require he signing of wo conracs for access, one for generaion and oher one for consumpion. A previous Royal Decree, 647/2011, has creaed a new paricipan specifically o manage he recharges of plug in elecric vehicle (an imporan componen in our sudies of microgrids), namely gesor de cargas (charging manager) and shown in he Figure 1. The charging managers are hose mercanile companies ha provide services of energy recharge defined in he ar. 9 h) of he law wih number 54/1997 which, being consumers, are eniled o he resale of elecric energy for energeic recharge services for elecric vehicle. In addiion, he model o be implemened will no be of concession bu one in which any eniy can obain adminisraive permission o insall recharging poss, buy and resell energy wihin he service recharge of elecric vehicle baeries [12]. In he fuure, oher expeced operaions for hese managers is he possibiliy o receive se poins from Disribuion Sysem Operaor (DSO) hrough a managemen cener o opimize he nework operaion. 3. Prospecive Currenly, here are hree main agens in he Spanish marke; generaion company, marke operaor and reailer. Since he emergence of microgrids, a new agen called he aggregaor is proposed. Is role would be o represen a group of one or more microgrids for heir managemen in differen sessions of he energy marke. More specifically, he aggregaor would receive supply needs and generaion offers from each microgrid hen, wih his informaion, could offer buying and selling bids in he differen markes of he sysem by opimizaion of he overall benefi of he

17 3. PROSPECTIVE 9 represened microgrids. Afer he paricipaion in he marke, he aggregaor migh communicae he se poins for generaion or buying of energy o each microgrid. In addiion, he aggregaor should provide he informaion abou each microgrid o is corresponding disribuor. The aggregaor adds anoher possibiliy ino he sysem. The microgrid could paricipae ino he ransporaion nework acions and o provide services o is disribuion nework. Fig. 2. General diagram of Spanish Energy Marke wih microgrids.

18

19 Chaper 3 Microgrid componens modelling 1. Technical specificaion of microgrid componens The considered componens of he microgrids include microgrid s considered componens have been he mos suiable elemens for he es cases, which are briefly described in Chaper 5. These elemens are presened below, and likewise, he main feaures for he modelling of each componen will be deailed laer. Fig. 1. Represenaion of a microgrid Demand. The elecric demand of a Microgrid can be grouped ino hree differen profiles depending on he way in which his demand is managed or no. - Criical demand The criical demand (or non conrollable demand) comes from devices or sysems which demand mus be insanly supplied and does no shif hrough ime, or when is posponemen could creae a user dissaisfacion i.e., he demand generaed by he lighing canno be posponed 11

20 12 3. MICROGRID COMPONENTS MODELLING because i would creae a blackou lighing [3]. Some examples of non-conrollable devices are: Illuminaion House equipmen Elevaors - Adusable demand These devices can no be oally managed, bu can be adused. If a demand managemen is necessary, he user (or remoely) can change a characerisic of he device decreasing he level of consumpion during a pre esablished period. For example: Air condiioning and cenral heaing - Shifable demand The demand profile for hese devices will be shifable in he planning horizon as explained laer. For example: Washing machines Elecric Vehicle: he demand from EV is also shifable hrough ime, bu i mus be aken ino accoun he mobiliy profile ha bounds he available ime o mee he load 1.2. Phoovolaic module. A phoovolaic module (PV) or solar panel is an inerconneced collecion of solar cells combined ino one iem only which is used o conver energy conained in he sun rays ino elecriciy. When an insallaion conains several panels, is commonly referred as phoovolaic array. Arrays are a grea way o increase he poenial of a solar elecriciy sysem, o provide a greaer oupu of elecriciy. However in his proec, a single module is only considered. To model he PV module, he parameers below are needed o describe is funcioning: Environmenal facors such as solar radiaion and ambien emperaure Technical values, for example: efficiency facor, nominal capaciy insalled, surface area and nominal operaing cell emperaure Oher adusable facors: iling of plaque Boh echnical and environmenal facor, which are needed o he consrains se below, mus be aken ino accoun for such modelling. Capaciy of supply depending on solar irradiaion and ambien emperaure Micro-wind urbine. Micro wind urbine is a general erm used for small scale wind urbines ha are ypically deployed near buildings.i is used o generae elecriciy from wind and ransmiing i direcly o he owner. In his case, he parameers involved in he modelling of micro-wind urbines are he following:

21 1. TECHNICAL SPECIFICATION OF MICROGRID COMPONENTS 13 In a simplified model, he only environmenal inpu is wind speed. Depending on he ype of modelling, oher facors such as ype of ambien (humid, dry...) can be aken ino accoun. Differen speeds of urbine: cu-in, cu-ou and raed speed. Technical inpus such as efficiency facor, nominal capaciy insalled and swep area Boh echnical and environmenal facor, which are needed o he consrains se below, mus be aken ino accoun for such modelling The uni iniial operaion saus Capaciy of supply depending on wind speed 1.4. Micro gas urbine. The hird kind of generaion uni is a micro gas urbine whose basic principles are based on micro combusion. The process consiss in burning a mixure of compressed air and fuel under specific condiions of consan pressure. Consequenly, he mechanic energy is generaed when he ho gas has expanded. The micro gas urbine could generae he energy ha renewable energy unis can no supply. The following echnical inpus should be considered: Insalled capaciy Maximum and minimum insananeous power Ramp down and ramp up limis Minimum sar-up and shu-down imes The consrains o be aken ino accoun mus comply wih all he above facors. Iniial sae of connecion Maximum and minimum insananeous power Ramp down and ramp up limis Minimum sar-up and shu-down imes 1.5. Elecric Vehicle. A charging spo conneced o he microgrid for elecric vehicles has been considered due o he benefis provided by he EV menioned in Chaper 1. On oher hand, here are wo imporan conceps relaed o baeries. SOC: Sae of Charge is defined as how much a baery is charged (or how full i is). If a baery is half charged, hen he sae of charge is 50%. This is commonly wrien as 50% SOC. DOD: Deph of Discharge is he opposie sae of charge. For insance, i is how much a baery is discharged. If a baery is fully discharged (empy) hen is deph of discharge is 100%, commonly wrien as 100% DOD, hen he sae of charge is 0%.

22 14 3. MICROGRID COMPONENTS MODELLING The DOD acquires imporance in relaion o lihium ion baeries, because if a lihium ion baery is fully discharged frequenly, is performance is drasically affeced. However, hese ype of baeries are one of he hree mos common ypes of baeries used for elecric and hybrid elecric vehicles due o is loading capaciy, lifecycle performance and environmenal impac. The parameers of he EV model are: Mobiliy profile: demand and available ime o V2G Minimum SOC allowed before leaving Efficiency of baery Energy sorage capaciy Maximum insananeous power Maximum deph of discharge (DOD) The consrains ha mus be considered are: Baery change balance: how he sae of charge varies from one period o anoher considering he amoun charged and discharged during his period Minimum level of ending charge: minimum amoun of baery charge a which he vehicle can leave he saion once decided i will be conneced Sorage capaciy Limiaions by mobiliy 1.6. Sorage device. An energy sorage sysem is defined as any device soring energy in any form such as chemical (baeries), hermal, mechanical (flywheel), elecrical (capacior), or anoher ype of energy for is usage a anoher ime. Since he use of he elecric vehicle baery is limied by he mobiliy profile, a sorage device is considered for a beer performance of he microgrid allowing a higher self-supply. In his way, he needed parameers are: Efficiency of baery Energy sorage capaciy Maximum number of cycles Maximum insananeous power of charging and discharging Maximum deph of discharge (DOD) Iniial baery s sae of charge Final baery s sae of charge The consrains ha mus be aken ino accoun are similar o he consrains of EV: Sorage balance from one period o anoher period Limis of insananeous charging/discharging power

23 2. MATHEMATICAL MODELLING OF MICROGRID S COMPONENTS 15 Limis of recommended DOD 1.7. Poin of inerconnecion o he grid. The microgrid has a poin of inerconnecion o he grid. The access will be regulaed hrough a ariff wih wo erms. The firs erm is a consan depending on he capaciy and use of he access. The second erm is variable in ime and depends on he amoun of exchanged energy wih he grid. For insance, if he microgrid needs exernal power, his erm will be a cos; bu if he microgrid has power surplus, his erm will represen a economic profi o he microgrid owners. 2. Mahemaical modelling of microgrid s componens Each componen has been modelling according wih he needs of he opimizaion problems, which will be presened in Chaper 4, and he characerisics presened above Micro wind urbine. On he microgrid basis, he only imporan hing of he micro wind urbine is he power ha he urbine can produce in each period. Due o his fac, an unique se of consrain is necessary o guaranee hese bounds. Le p W he wind power signal a ime hen, he variables will be bounded beween zero(if wind has no enough speed) and available wind power : 0 p W P W, T where T will be a se of ime periods. To calculae he available wind power P W been considered [4]:, he following piecewise funcion has 0 if v < v in or v v co P W = A W η(av 3 bn W ) if v ci v v r A W ηn W if v r v < v co N W v3 ci where boh a = vr 3 vci 3 and b = vr 3 vci 3 are consans, η is he efficiency facor, N W (kw ) is nominal power and A W is he swep area by wind urbine blades. In addiion, his funcion depends on wind speed v (m/s) and hree speed characerisics of he urbine: v ci cu-in speed, v r raed speed and v co cu-ou speed (m/s) as seen he Figure 2.

24 16 3. MICROGRID COMPONENTS MODELLING Fig. 2. Wind urbine power oupu. The wind speed v, is no obained direcly because he wind daa are aken a a reference heigh (v ref a h r ). Thus, he wind speed a hub heigh h can be calculaed by using a power-law equaion: v = v ref ( h h r ) α where α is he power law exponen which value is usually 1/7 if here is no a specific sie daa [13] Solar power available. This case is similar o he previous subsecion in he sense ha he se of consrains is he same kind. The power provides by he phoovolaic module is echnically bounded by is characerisics and weaher facors. Le p P V be a signal solar power a period hen he consrains are: 0 p P V P P V Where P P V is he maximum available solar power a inerval. The used model o esimae hese bounds is par of anoher IREC inernal proec where he following linear equaion was developed o calculae he generaed power by PV modules: P P V = (a 1 Ir + a 2 )T C + a 3 Ir + a 4 Where he coefficiens a were calculaed using leas squares mehodology. The erm T C is he cell emperaure which was calculaed under sandard condiions by

25 2. MATHEMATICAL MODELLING OF MICROGRID S COMPONENTS 17 a simple linear model using he direc normal irradiaion Ir and ambien emperaure T a. Once again, leas squares mehodology has been used o carry ou he calculaion: T C = b 1 + b 2 T a + b 3 Ir 2.3. Micro gas urbine. To schedule when sar-up or shu-down he micro gas urbine, he formulaion of hesis [6] have been followed. Le u MT {0, 1} binary variable expressing he off-on operaing saus of he micro gas urbine over he h inerval (u MT = 1 if commied,u MT = 0 if uncommied). The values of u MT and u MT +1 mus obey cerain operaing rules in order o mee he limis of he minimum ime in service and idle. Thus, i is necessary o inroduce wo exra binary variables e MT and a MT for each u MT. Le e MT {0, 1} be a sar-up indicaor. I has a value of one in all inervals where he MT uni has changed from u MT 1 = 0 o u MT = 1, and zero elsewhere. Similarly, a MT {0, 1} is a shu-down indicaor. I should have a value of one in all inervals where u MT 1 = 1 changes he value o u MT = 0, and zero oherwise. The relaion beween hese binary variables is modelled hrough he following ses of consrains: u MT u MT 1 e MT + a MT = 0 T σ e MT + a MT + k= τ k=+1 a MT k 1 T e MT k 1 T where σ and τ are he minimum beween T and plus he minimum sar-up and shu-down imes respecively. Le p MT be a posiive variable for he signal power o micro gas urbine a period. These semiconinuous variables ake values in he se {0} {P MT, P MT } depending on he value of u MT. If he micro gas urbine is in operaion sae a ime hen u MT = 1 and he power available has a range of possible values beween he wo bounds of insananeous power P MT and P MT, in oherwise, he micro gas urbine is idle and p MT is zero. u MT P MT p MT u MT P MT, T On oher hand, if he micro gas urbine is commied, he variable p MT mus ake values wihin a range depending on he value of he previous variable p MT 1. The difference beween he generaion level a period 1 and generaion level a period, if i varies, mus vary wihin a range (R l,r u ). The range depends on he machine characerisics and hese consrains are called ramp limis.

26 18 3. MICROGRID COMPONENTS MODELLING R l p MT p MT 1 R u, T \{1} 2.4. Sorage device. As already menioned above, he sorage device can be charged, if he oal power oupu of he generaion unis is higher han he energy demand, or can provide power if i is discharged. Due o his dual process, wo ses of posiive variables are needed o model he performance of he sorage device. Le p Sc be a posiive variable for he charged power o sorage device a period. Le also p Sd be a posiive variable for he discharged power from he sorage device a period. Boh variables ake posiive values beween echnical bounds bu are never higher han zero in he same period: p Sd P Sd x S T p S P Sc (1 x S ) T where x S {0, 1} are binary variables expressing he direcion operaing saus of he sorage over he h inerval (x S = 1 if discharged,x S = 0 if charged). Binary variables are needed because in he obecive funcion of he eriary and second opimizaion problems, only he discharged process has a cos, as will be seen laer. This way, he charged power is separaed of he discharged power. Le SOC S be a SOC indicaor a period. The sorage balance is ensured by he following equaion where, from an inerval o anoher, he amoun of charged/discharged power is aking ino accoun wih is corresponding efficiency facor e c and e d : N S SOC S = N S SOC S 1 + (p Sc /e c e d p Sd ) = T \{1} N S SOC S 1 = N S SOC S 0 + (p Sc 1 /e c e d p Sd 1 ) SOC S T = SOC S F where N S is he sorage s nominal capaciy. The efficiency facor is a performance index, which expresses he relaionship beween he provided power for he microgrid and he real sored power in he form of percenage. The efficiency facor have been considered from he microgrid poin of view. For insance, if he baery requires power hen, he microgrid mus provide more amoun of power han he required o ake ino accoun he losses. In many ypes of baeries, he full sored energy canno be compleely discharged wihou causing serious damage o he sorage device. Thus, here are recommended maximum and minimum rae for he DOD. Taking ino accoun hese recommended DOD and he relaion beween SOC and DOD (DOD + SOC = 1), he bounds for he sae of charge in each are he following: 1 DOD SOC s 1 DOD, T 2.5. Elecric vehicle. The model for a single elecrical vehicle is similar o he sorage device modelling bu considering he mobiliy profile.

27 2. MATHEMATICAL MODELLING OF MICROGRID S COMPONENTS 19 Le U EV T a se of periods where he EV is conneced o he microgrid. The variables associaed wih he provided power p EV d and required power p EV c beween he microgrid and he EV, are defined in his se. Since he previous case, if a process (charge/discharge)is no acive a period hen, he bound for he associaed variable, is zero in his period. However, if he process EV is acive hen he bound will be he allowed maximum P which is he same in wo process: p EV d p EV c P EV x EV (1 x EV ) EV P where x EV are binary variables expressing he direcion operaing saus of he EV baery over he sh period in U EV (x EV = 1 if he EV is charged, x EV = 0 in oherwise). Le D EV he required energy by he vehicle while is no conneced i.e., in he periods T \ U EV. This demand does no affec direcly o he Microgrid. However, i affecs o he sae of charge SOC EV which mus be calculaed in all periods considering he following piecewise funcion which depends on 1: { N EV SOC EV N EV SOC EV 1 + (p EV c /e = c e d p EV d ) if U EV N EV SOC EV 1 D EV if T \ U EV If = 1, he funcion is: N EV SOC EV 1 = N EV SOC EV 0 + P EV 0 where where N EV P EV 0 = { (p EV c 1 /e c e d p EV d 1 ) if 1 U EV D EV 1 if 1 T \ U EV is he EV baery s nominal capaciy and e c, e d are he efficiency facor. To guaranee a bes performance of he baery life, a maximum % for he DOD mus be imposed if he EV is conneced o he Microgrid. The value of his maximum will be higher in he periods before he nex displacemen o ensure he EV demand. Abou he upper bounds of SOC EV will be 100% due o he lihium ion baeries can be charged fully. SOC EV 1 DOD if, + 1 U EV SOC if U EV & + 1 T \ U EV 0 if T \ U EV SOC EV 1 T

28 20 3. MICROGRID COMPONENTS MODELLING 2.6. Demand Shifable demand. The shifable demand has a profile defined by a number of subinervals L wih L T and a power profile D SH in L. The decision will be o choose he period which o sar o supply his demand. Le x SH {0, 1} variables expressing he iniial inerval where he shifable demand sars o be supplied for he nex L hours. Thus, if x SH = 1 hen, only he variables associaed o x SH : d SH, d SH +1,, d SH +(L 1), ake posiive values, and more specifically, each d SH + akes he value DSH l+1 exacly. Only he firs T L + 1 binary variables can be defined because each x SH variable is associaed wih he nex L 1 d SH variables. Anoher way o see he relaionship beween he differen variables, i is from he power variables poin of view: each variable d SH is associaed o he following L 1 binary variables. This way, he consrains are: d SH = L l=1 D SH l x SH l+1 = 1..T Considering ha he shifable demand profile sars us once, one binary variable only can ake he value one: T (L 1) =1 x SH = Criical and adusable demand. Criical demand, D C, is forecased for each inerval and i mus be supplied compleely bu i can no be possible. To avoid an imbalance in he microgrid, an adusable rae demand has been inroduced. For insance, if a inerval he peak demand is oo high, he supply can be decreased unil an assumed maximum f C, aking ino accoun ha he criical demands no supplied has a penaly in he obecive funcion. Le d C be a posiive variable indicaing he supplied criical demand a period, and le also d A be a posiive variable indicaing he demand no supplied in he same period. Considering ha f C is a percenage of adusable demand, which is also a fracion of he criical demand, he consrains are he following: d C + d A = D C T d A f C D C T 2.7. Poin of inerconnecion o he grid. The inerconnecion poin also has wo acions: exporing and imporing power. In addiion, he consan cos of he access ariff mus be always paid regardless he microgrid buys or sells power o he grid. For hese facs, binary variables are needed again and also, wo ses of posiive variables o avoid he absolue value in he obecive funcion. Le x i be a indicaor of he flow direcion in he inerconnecion a period. x i akes he value 1 if he microgrid is selling he power surplus, and akes he value 0 when he grid provides power o he microgrid. Le p Ib be a posiive variable o

29 2. MATHEMATICAL MODELLING OF MICROGRID S COMPONENTS 21 calculae he amoun of acquired power from he grid a period sh, and le also p Is be a posiive variable o accoun he amoun of power sold o he grid a period sh. Considering he bounds of inerchangeable power beween he grid and microgrid and he above variables, which canno be higher han zero a same period, he consrains are as follows: { p Is I x P, T p Ib (1 x ) P I, T

30

31 Chaper 4 The microgrid cenral conroller opimizaion problem From he poin of view of he microgrids energy managemen, he economic scheduling of generaion unis, sorage sysems and loads is a crucial problem where he opimizaion mehods can be he mos imporan ool o face i. This managemen is carried ou by he MCC (Microgrid Cenral Conroller) receiving/sending signals o local conrollers of each componen and receiving exernal inpus (price, weaher,...). 1. Overview of he microgrid cenral conroller problem A microgrid mus ensure he supply of elecrical energy a any ime. For his fac, a supervisory conrol,eiher cenralized or decenralized, is necessary. This supervisory conrol is splied ino hree hierarchical levels: 1 Cenral Auonomous Managemen Conroller (CAMC) belonging o he disribuion nework operaor (DNO) 2 Microgrid cenral conroller 3 Local conrollers splied ino Microsource Conrollers (MCs) and Load Conrollers (LCs) CAMC DNO Microgrid MCC LC LC LC MC MC The firs conroller does no belong o he microgrid, i is delegaed of he macrogrid conrollers. The main inerface beween his conrol level and microgrid, is he MCC. The MCC is responsible of he opimizaion of is operaion according o he cos supply price, he microgrid conrollable componens and he expeced loads. To manage he microgrid componens, he MCC also mus coordinae local conrollers 23

32 24 4. THE MICROGRID CENTRAL CONTROLLER OPTIMIZATION PROBLEM associaed wih each componen. To carry ou hese obecives, he MCC has wo main funcional modules which are EMM (Energy Managemen Module) and PCM (Proecion Co-ordinaion Module). On his proec basis, he EMM operaions are imporan. The EMM communicaes he se poins, such as acive and reacive power oupu, volage and frequency, o each local conroller [1]. Thus, he EMM is in charged o ensure he profiabiliy of microgrid, which definiely is a key facor for is deploymen. Microgrid conrol sraegies can be divided in he following ypes: real-ime opimizaion, exper sysem conrol and decenralized conrol [14] Long-erm planning. The Long-erm managemen has a scope of 2-10 years. Is purpose is o design he opology of a new microgrid, wha echnologies will be included and heir insalled capaciy. Oher goals can be when o sar operaion in order o mach fuure loads and when o reire old unis Shor-erm managemen. The scope for shor-erm managemen is 1 day in his proec. Is main purpose is o decide when he unis mus be urn-on or urn-off, he amoun of generaed power a each period by each uni, o manage he sorage device use and he exchangeable power wih he general grid hrough he inerconnecion poin while meeing several operaional consrains. Included in his level of managemen are hree sublevels which can be disinguished depending on he ime frame (always in he conex of a day) and he performed funcions. These hree sublevels are described below: a) Teriary conrol: economic The eriary conrol has a scope of 24 hours wih periods of 15 minues and has wo obecives. The firs is he economic opimizaion using a program based on an Uni Commimen problem. The second is o improve he profiabiliy of he supply and demand balance by ineracing wih he grid. These aims are realized aking ino accoun daily forecass regarding weaher, energy price and demand daa. On economic opimizaion, he signals for he conrollable unis are calculaed allowing he sysem o find ou an opimal uni commimen considering fuure values by exploring he price differences beween on-peak and off-peak periods during a day. The final resul is a schedule of he power oupus for each period wihin he opimizaion range. As already menioned above, he opimizaion procedure depends on energy price. However, in a higher level, he procedure depends on he marke policy adoped in he microgrid operaion. The main role in he wo possible policies is played by he MCC. In he firs case, he MCC supplies he oal demand of he microgrid only using is local producion from generaion unis and sorage devices and, rying no expor power o he disribuion grid. In he second policy, he MCC ries o maximize he value of he microgrid paricipaing in he open marke i.e., buying and selling power o he grid [11]. If he microgrid wan o paricipae in he marke, a daily schedule of producion is necessary because he energy markes se heir operaions wih a day-ahead. This schedule could be he eriary conrol program.

33 1. OVERVIEW OF THE MICROGRID CENTRAL CONTROLLER PROBLEM 25 We will focus on he second policy bu evenually, an opimal soluion o our model could be a mixing of boh policies: wih and wihou ne energy exchange depending on he he ime period. b) Secondary conrol: power qualiy The secondary conrol runs each 15 minues wih periods of 30 seconds (more or less depending on he ime range o obain he real daa available), and i is in charge of power qualiy opimizaion and o minimize he average of all deviaions compared o he eriary conrol program. For doing so, he conroller mus ake ino accoun curren weaher daa, operaion daa of generaion unis while ensures he exchanged power wih he grid and sorages saes of charge programmed above. This opimizaion problem is resolved by a cascade programming (Fig. 1). Included in he procedure is an opimizaion problem in each ieraion. The soluion of one subproblem is used as a sar poin for he nex subproblem which will have one period ime less and daa more curren. Finally, he procedure combines he soluions of he subproblems o reach an overall soluion. Quarer hour Ieraion 1 X₁* 30 periods 30 seconds Ieraion 2 X₂* 29 periods Ieraion 3 X₃* 28 periods Las ier. 2 periods Secondary soluion X* = (X₁*, X₂*, X₃*,, X₂₈*) X₂₈* Fig. 1. Cascade procedure for he secondary conrol opimizaion problem. c) Primary conrol: power reliabiliy The primary conrols, also called momenary conrols, are execued in real-ime each second. Is obecive is ensured he balance beween generaion and demand agains any unexpeced issue. However, he primary conrol is no considered in his proec because i is no a opimizaion problem.

34 26 4. THE MICROGRID CENTRAL CONTROLLER OPTIMIZATION PROBLEM The eriary and secondary opimizaion problems are formulaed mahemaically below. 2. Teriary conrol opimizaion problem To formulae he model of he eriary conrol problem, we are going o ake ino accoun he mahemaical modelling of each microgrid componen, which have been explained in Chaper 3, bu adding he balance consrains beween demand and generaion and an economic obecive funcion Obecive funcion. The goal is o minimize he generaion coss of he micro gas urbine and he economic coss associaed wih he exchanged energy beween he grid and he microgrid: Min T C MT u e MT + C MT d a MT + C MT p MT + + T + T (C I2 (p Ib (C Sd p Sd + p Is ) + C I1 (p Ib p Is )) + C EV d p EV d + K A d A ) where: C MT u Sar-up cos of MT e C MT d Shu-down cos of MT e C MT Generaion cos of MT e/kwh C I2 Consan cos of access ariff e/kwh C I1 Variable cos of energy e/kwh In addiion, an economic cos for each baery discharging process, based on he reducion in he number of complee charge/discharge cycles a baery can perform before is nominal capaciy falls below 80% of is iniial capaciy [10], and an economic penaly for he no supplied demand, are evaluaed. These penalies have been esed by a sensiiviy analysis. One of he main problems wih elecric cars and is owners is he range anxiey concep. Range anxiey is he fear of running ou of elecriciy before desinaion has been reached. To minimize his poenial problem, he user has he opion o choose an obecive funcion less economical bu more secure. The following erm will be added o he obecive funcion in some es cases. Is new goal will be o minimize coss and o encourage a greaer charging baery. U EV K RA(1 SOC EV )N EV

35 2. TERTIARY CONTROL OPTIMIZATION PROBLEM 27 where K RA is he economic penaly whose value has been esed wih differen values by a sensiiviy analysis as will be explained in Chaper 5. This erm minimizes he energy amoun missing o have a fully charge a each period U EV Power balance consrain. To ensure a balancing beween generaed energy, sorage and load, he following consrains are essenials. p W + p P V + p MT + p EV d + p Sd + p Ib = d C + d SH + p EV c + p Sc + p Is, T On he lef hand side of he balance equaion are he variables associaed wih he generaed power by micro wind urbine, PV module and micro gas urbine, he discharged power from he sorage device or EV baery and he purchased power o he grid. On he righ hand side appears he variables associaed wih required power by he load, boh sorage and EV baeries and he sold surplus power o he grid Mahemaical Formulaion. Due o he size of he eriary conrol opimizaion problem, a summary of he formulaion is shown below: M in (1) Obecive Funcion: minimize overall cos (2) microgrid balance (3 8) Sorage device consrains (9 14) EV consrains (15 16) Shifable demand consrains s.. (17 18) Criial and adusable demand consrains (19 23) MT conrains (24 25) Inerconnecion poin consrains (26) Wind power consrains (27) Solar power consrains Finally, he mahemaical formulaion is presened.

36 28 4. THE MICROGRID CENTRAL CONTROLLER OPTIMIZATION PROBLEM Min T =1 T + C MT u e MT =1 T + =1 (C I2 (p Ib (C Sd p Sd + C MT d a MT + C MT p MT + + p Is ) + C I1 (p Ib p Is )) + C EV d p EV d + K A d A ) (1) s.. p W + p P V + p MT + p Sd + p i,b = d C + d SH + p EV p Sc + p Is T (2) N S SOC S = N S SOC S 1 + ( psc e c e d p Sd ) T \{1} (3) SOC1 s = SOC0( s psc 1 e c e d p Sd 1 ) (4) SOC T s = SOCs F (5) l d SOC s l u T (6) p Sd p Sc P Sd x S T (7) P Sc (1 x S ) T (8) N EV SOC EV N EV SOC EV = N EV SOC EV c 1 + ( pev ec e d p EV d = N EV SOC EV 1 D EV N EV SOC EV 1 = N EV SOC EV 0 + P EV ) U EV (9) T \U EV (10) 0 (11) SOC EV SOC EV 1 T (12) p EV c p EV d d SH = P EV x EV U EV (13) (1 x EV ) P EV U EV (14) L l=1 T L +1 =1 D SH l x SH l+1 T (15) x SH = 1 (16) d c + d a = D C T (17) d a f A D C T (18)

37 3. SECONDARY CONTROL OPTIMIZATION PROBLEM 29 R l p MT p MT 1 R u T (19) u MT P MT p MT u MT P MT T (20) u MT u MT 1 e MT + a MT = 0 T (21) σ e MT + a MT + k= τ k=+1 a MT k 1 T (22) e MT k 1 T (23) p Ib x I P I T (24) p Is (1 x I ) P I T (25) 0 p W P W T (26) 0 p P V P P V T (27) The equaions (1)-(27) define a mixed ineger linear program (MILP) and has been implemened in boh General Algebraic Modelling Sysem (GAMS) and C language wih he help of he Visual Sudio plaform. In boh cases he model has been solved using CPLEX wih sandard opions. 3. Secondary conrol opimizaion problem To formulae he secondary conrol opimizaion problem, i is no possible o use direcly he mahemaical modelling of microgrid componens presened in Chaper 3. In he secondary conrol, some variables of he mahemaical modellings become parameers fixed by he eriary conrol. For insance, he exchanged power wih he grid, sorage saes of charge and he commied periods for he MT are ensured in each quarer hour. These changes are explained below for each microgrid componen excep he micro wind urbine and he PV module because is models don change Mahemaical modelling of microgrid componens for he secondary conrol Micro gas urbine. In he micro gas urbine modelling for secondary conrol, he hree ses of binary variables u MT,e MT and a MT became fixed because he eriary conrol decided previously when he uni is commied. Le Ū MT be a operaion sae indicaor which is equal o one if he MT is commied a quarer hour and zero in oherwise. Thus, he news consrains are he following: Ū MT P MT p MT Ū MT P MT, J

38 30 4. THE MICROGRID CENTRAL CONTROLLER OPTIMIZATION PROBLEM where J will be he se of periods for he secondary conrol. operaion sae is consan in he se J. To noe ha he The ramp limis consrains remain equal: R l p MT p MT 1 R u, J\{1} The res of micro gas urbine consrains are no needed in secondary conrol Sorage device. The sorage device consrains remain equal o he consrains defined in Chaper 3. However, a new condiion abou he final sae of charge programmed by he eriary conrol SOC S F, is added. Le s S + be a slack measuring he posiive deviaion over he final SOC and le s S be a slack measuring he negaive deviaion from he expeced final SOC. Boh variables are bounded by a fracion f Ss of he final SOC, SOC S F. SOC S J + ss + s S = SOC S F 0 s S +, s S f Ss SOC S F Elecric Vehicle. Le he quarer hour o opimize by he secondary conrol hen, i is necessary o inroduce a parameer Ū EV which akes he value one if U EV and zero oherwise. Thus, he EV modelling is he following: 0 p EV c Ū EV EV P x EV J 0 p EV d Ū EV P EV (1 x EV ) J N EV SOC EV c = N EV SOC EV 1 + ( pev e c c N EV SOC EV 1 = N EV SOC EV 0 + ( pev 1 e d p EV d ) J\{1} e c e d p EV d 1 ) As in he sorage device modelling, he final SOC appears fixed by he eriary conrol. Noneheless, any unexpeced even could happen and he final sae of charge would vary. For his reason, wo exra posiive variables are necessary. Le s EV + be a slack measuring he posiive deviaion over he final SOC and le s EV be a slack measuring he negaive deviaion from he expeced final SOC. Boh variables are bounded by a fracion f EV s of he final SOC, SOC EV F. SOC EV J + s EV + s EV 0 s EV +, s EV = SOC EV F f EV s SOC EV F

39 3. SECONDARY CONTROL OPTIMIZATION PROBLEM Demand. In he secondary conrol opimizaion problem only have wo differen profiles of demand: criical and adusable. The shifable demand has been programmed by he eriary conrol and for he secondary conrol, his demand is only par of he criical demand in he commied quarer hours. Thus, he consrains are he ones defined in Chaper 3 for criical and adusable demand. d C + da = DC J d A f C D C J Poin of inerconnecion o he grid. The eriary conrol opimizaion problem programms he exchangeable power wih he general grid for each quarer hour. The secondary conrol mus ensure his amoun bu in average in he J inervals. Ib Is Le P and P he soluion of eriary conrol for he inerconnecion hen, he consrain for he average is: J p Ib p Is Ib Is = J ( P P ) Power balance consrain. There is an unique difference in he power balance consrain beween he eriary conrol and he secondary conrol: he shifable SH demand variable became an parameer D. The parameer value is zero if he eriary conrol didn decide o supply he shifable demand a quarer hour, and i is posiive in oherwise. p W + p P V + p MT + p EV d + p Sd + p Ib = d C + D SH + p EV c + p Sc + p Is, J As in he previous problem, on he lef hand side of he balance equaion are he variables associaed wih he generaed power by micro wind urbine, PV module and micro gas urbine, he discharged power from he sorage device or EV baery and he purchased power o he grid. On he righ hand side appears he variables associaed wih required power by he load, boh sorage and EV baeries and he sold surplus power o he grid Obecive funcion. As in he previous problem, he crieria is o minimize overall cos. Min J + J δc MT p MT δ(c Sd p Sd + C EV d p EV d + K A d A )+ where: C S (s S + + s S ) EV C (s EV + + s EV ) C MT Generaion cos of MT e/kwh

40 32 4. THE MICROGRID CENTRAL CONTROLLER OPTIMIZATION PROBLEM C Sd Discharged sorage cos e/kwh C EV d Discharged EV cos e/kwh K A Economic penaly of no supplied demand e/kwh and where C S EV and C are he shadow prices of he SOC consrains (in boh cases, sorage device and EV baery) a period in eriary conrol opimizaion problem. These consrains are equaliy equaions and an increase in he righ hand side means a free increase in he SOC value. As an free increase in he SOC value resuls ino a lower cos hen, he shadow prices are negaive. Thus, we should change he sign, becoming his value in a posiive erm, in order o penalize any deviaion Mahemaical Formulaion. The following summary is creaed o help he mahemaical formulaion visualizaion. The consrains are grouped by componens. M in (28) Obecive Funcion: minimize overall cos (37 43) EV consrains (44 45) Criial and adusable demand consrains s.. (46 47) MT conrains (29) microgrid balance (30 36) Sorage device consrains (48 50) Inerconnecion poin consrains (51) Wind power consrains (52) Solar power consrains Finally, he mahemaical formulaion is presened. Min J + J δc MT p MT δ(c Sd p Sd + C EV d p EV d + K A d A )+ C S (s S + + s S ) EV C (s EV + + s EV ) (28) s.. p W + p P V + p MT + p EV d + p Sd + p Ib = d C + pev c + p Sc + p Is + D SH J (29) N S SOC S δ(psc = N S SOC S 1 + e c ed p Sd ) J\{1} (30)

41 3. SECONDARY CONTROL OPTIMIZATION PROBLEM 33 N S SOC S 1 = N S SOC S 0 + δ( psc 1 e c ed p Sd 1 )) (31) SOC S J + ss + s S = SOC S F (32) SOC SOC S SOC J (33) p Sd P Sd x S J (34) p Sc P Sc (1 x S ) J (35) 0 s S +, s S f Ss SOC S F J (36) N EV SOC EV δ(pev c = N EV SOC EV 1 + e c e d p EV d )) J\{1} (37) N EV SOC EV 1 = (1 U EV )N EV SOC EV +1 + U EV SOC EV J + s EV + s EV SOC EV c + δ( pev 1 e c e d p EV d 1 )) (38) = SOC EV F (39) SOC SOC EV 1 J (40) 0 p EV c 0 p EV d Ū EV Ū EV P EV x EV J (41) P EV (1 x EV ) J (42) 0 s EV +, s EV f EV s SOC EV F J (43) d C + da = D J (44) 0 d A f A D J (45) R l p MT p MT 1 R u J\{1} (46) U MT P MT p MT U MT P MT J (47) p Ib I x P J (48) p Is (1 x ) P I J (49) p Ib p Is Ib Is = J ( P P ) J (50) 0 p W P W J (51) 0 p P V P P V J (52)

42 34 4. THE MICROGRID CENTRAL CONTROLLER OPTIMIZATION PROBLEM The equaions (28)-(52) define a mixed ineger linear program (MILP) and has been implemened in boh General Algebraic Modelling Sysem (GAMS) and C language wih he help of he Visual Sudio plaform. In boh cases he model has been solved using CPLEX wih sandard opions. However, he cascade procedure has only been implemened in C language.

43 Chaper 5 Implemenaion and resuls 1. Implemenaion The implemenaion has been done in GAMS and C language wih he help of he Visual Sudio plaform. GAMS has been used for esing he opimizaion problems by sensibiliy analysis of penalies. However, he algorihms for he IREC s microgrid faciliies can only be programmed in C language, which is also more efficien for programming of he cascade procedure. Differen configuraions can be chosen in he algorihm implemened in C language depending on load seleced ypes, generaion unis aken ino accoun and oher facors such as he range anxiey. Depending on he seleced configuraion, he program generaes he correc MILP problem and solves i using CPLEX (wih defaul opions). I creaes he scheduling graph associaed wih he soluion and creaes he needed files o he secondary conrol. In he secondary conrol algorihm, he daa of he elemens previously seleced are colleced for he corresponding quarer hour and hen, he cascade procedure is execued. Every hree hours, when here are new weaher forecas daa, he eriary problem is refreshed from he curren quarer hour. Aferwards, he secondary conrol problem coninues from he las correced period. 35

44 36 5. IMPLEMENTATION AND RESULTS In order o have a beer undersanding of he general procedure, he Figure 1 shows a flow diagram of he energy managemen algorihm. Selec Microgrid configuraion START Ge forecas daa for nex 48h Creae Teriary Problem Opimize Save scheduling Creae files for Secondary conrol YES Ge firs quarer hour Creae Secondary Problem Opimize Save firs inerval soluion Penulimae NO Eliminae firs inerval Updae new bounds Wai 30s Ge forecas daa Teriary conrol Secondary conrol Save global soluion of his quarer hour YES 3h NO Nex quarer hour Fig. 1. Flow diagram of Energy Managemen Sysem

45 1. IMPLEMENTATION 37 Figure 2 represens a flow diagram of he creae problem module for eriary and secondary problems. Fig. 2. Generic flow diagram of he creae problem module

46 38 5. IMPLEMENTATION AND RESULTS And Figure 3 shows how he secondary conrol updaes he iniial daa from one ieraion o he nex one by a flow diagram. Period Elemens E Ier. J Updae new bounds Updae mean expeced power for inerconnecion poin k k EV in E YES Updae iniial SOC SD in E YES Updae iniial SOC END Fig. 3. Flow diagram of he iniial daa in cascade procedure where he parameers SOC EV 1 and SOC EV 1 represen he opimal values of he variables SOC EV 1 and SOC EV 1 in he ieraion The IREC s microgrid emulaor IREC s microgrid is a 40kW low volage es plaform. This microgrid is composed by differen devices capable of emulaing he behaviour of real disribued energy resources wih profiles of generaion, sorage and load. Addiionally, a sor of four non-emulaed devices ha are going o be insalled on he emplacemen of he microgrid will provide real solar and wind generaion, 20kWh lihium sorage baery and a peaking power source and consumpion based on ulracapaciors. These devices are currenly programmed o work as emulaors bu, when available, hey are going o be used o ransform he energy coming from he renewable energy sources on he roofop of IREC s building ino he microgrid. The possibiliy o work wih emulaors gives grea flexibiliy o he whole microgrid in erms of giving he abiliy o reproduce any kind of siuaion and he possibiliy of performing any kind of es environmen independenly of he meeorological (wind and sun) condiions. The emulaors of IREC s microgrid are devices able o ac as any kind of elecrical nodes and able o perform any kind of V2G es. Due o ime limiaions, he resuls presened in his work have been obained compuaionally using a Dell Laiude E5400 Inel Duo wih 3.49 Gb of RAM memory and wo processor a 2.53GHz.

47 3. TEST CASES Tes cases 3.1. IREC s proec es case. A ask in he IREC s proecs is o es advanage of he V2G capabiliy of charging spo for minimizing he amoun of energy consumed from he uiliy grid during peak period pricing and, consequenly, he household energy bill. In order o validae he echnical viabiliy of such applicaion, his es case has been proposed o be carried ou in IREC s microgrid faciliies wih he opimizaion problems presened in his proec. The Figure 4 shows he seleced elemens ha paricipae in he IREC s proec. Load / Household PV Generaor Wind Generaor EV V2M [W] [W] [W] SOC Measuremens Microgrid managemen Weaher forecas Grid Operaor Side Daily marke prices Fig. 4. Seleced elemens for he IREC s proec: house load, renewable energy resources ha be insalled a he building and he elecrical vehicle charger spo. In addiion, he proec has been adaped o he acual capabiliies of EVs considering he resricions caused by he mobiliy needs of EV users. For doing so, his proec has been splied ino hree es cases: I.1 : The obecive of his es case is o guaranee a beer comparison of he effecs of he EV usage by disconnecing he V2G sysem. This means ha he discharging process is no available and he baery canno be used for supplying he household load. Thus, he discharging cos in he obecive funcion and he variables of discharging process are no considered and, consequenly, he binary variables are no needed. I.2 : The obecive of he second es case is o use he sorage sysem of EV for saving money by means of buying energy during low price periods (off-peak hours) and discharging of he baery when prices are highes (peak hours) doing use of V2G sysem. I.3 : In he las es case, he obecive funcion will be modified by adding he erm associaed wih he range anxiey. The obecive is o prioriize minimizaion of he range anxiey, alhough he discharging process is available. For all cases, he seleced profiles boh EV mobiliy and household load have been measured during a sandard working day. The weaher daa used in he proec were measured a summer. The source and deails of daa and parameers have no been specified because hey are par of IREC s inernal proecs.

48 40 5. IMPLEMENTATION AND RESULTS 3.2. Exended es case. The exended es is similar o he IREC s proec case bu including a sorage device and one micro gas urbine. Wih hese wo unis conneced o he microgrid, he resuls can be more ineresing due o a higher opporuniy o disribue he use of he exernal power during he day wihou creaing overload a peak hours. Furhermore, he shifable demand is esed in he exended case because i could be included in IREC s proec case laer. The Table I presens a summary of he main characerisics of all es cases. TABLE I Tes cases Tes cases IREC s proec Exended case I.1 V2G: off Range anxiey: off IREC s proec: I.2 V2G: on + Sorage Range anxiey: off + MT I.3 V2G: on Range anxiey: on 4. Resuls for he eriary conrol problem Table II provides he main informaion of he eriary conrol problem associaed wih every es case. TABLE II Teriary conrol opimizaion problem Tes Toal Binary Toal Execuion O.F Generaion case variables variables consrains ime value Cos [ms] [e] [e] I IREC I I I Exended I I Noe ha he problem has been finally solved for 192 periods due o he low compuaional cos and given ha he weaher forecas daa are always available for he nex 48h. The issues associaed wih final sae of charge a he las periods of he 24h are avoided considering his greaer horizon. However, he soluion graph shows he firss 96 periods (one day), which are he mos imporan IREC s proec es case. In he generaion scheduling figures for he eriary conrol soluion, he generaed power is in he op and he power supplied in he boom of he plo. In each period, he amoun of energy provided or demanded by each source or device is represened wih a differen color. The black

49 4. RESULTS FOR THE TERTIARY CONTROL PROBLEM 41 line represens he variable energy price in e/kwh and is values are shown in he secondary axis. The grey zones show he periods of connecion o he grid of he EV I.1 V2G off. The firs es case is he simples. The only goal is o choose he cheapes periods o charge he EV baery. However, as i will be seen laer, his opion is beer han no use he energy managemen program. The Figure 5 shows he resul of he es case where he V2G sysem is no available and he range anxiey is no considered. Fig. 5. Teriary energy scheduling of IREC s I.1 case As seen in he figure, he EV baery is charged during he las se of hours i is conneced o he microgrid. In his es, hese hours correspond o 1:45 5:00 a nigh. The elecric vechicle SOC reaches he 100% and i is no necessary oher charge during he day I.2 V2G on. Figure 6 represens he resul of he second es case. As can be observed in he figure, when he EV is conneced o he microgrid (grey zones), he EV baery is charged in off-peak periods such as during nigh hours, and discharged in on-peak periods namely in he afernoon. In addiion, in hese las periods, he microgrid does no need exernal power because i is cheaper o discharge he EV baery. I is a example of he profis derived from he V2G sysem.

50 42 5. IMPLEMENTATION AND RESULTS Fig. 6. Energy scheduling for he IREC s proec case wihou range anxiey I.3 Range anxiey on. Before presening he resul wih he range anxiey, i is necessary o analyse how his erm affecs o sae of charge. For doing so, he following sensiiviy analysis has been done: Fig. 7. Sensiiviy analysis for range anxiey penaly In Figure 7, i is observed how he elecric vehicle SOC changes a conneced periods (periods wih horizonal lines) depending on he value of range anxiey penaly. As can be seen, he values can be grouped in hree levels: low, medium and maximum; wih he corresponding values K RA = 0 (case I.2), K RA = where he EV baery is no discharged bu i is charged in off peak periods, and K RA = where he anxiey is highes and he EV baery is charged as soon as possible.

51 4. RESULTS FOR THE TERTIARY CONTROL PROBLEM 43 For his es case, he value K RA = is chosen. However, in he fuure, he microgrid owner will choose he modaliy more adequae for is necessiies. Wih he K RA value fixed, he es resuls are shown in Figure 8. Fig. 8. Teriary energy scheduling of IREC s I.3 case As can be seen, he range anxiey affecs o he EV baery performance. In his case, he EV baery is never discharged in order o ensure a beer SOC as soon as possible, bu wihou waiing for off-peak hours compleely. For his reason, he inerconnecion poin is always providing exernal power Comparison: IREC s proec case. In order o provide a beer general vision of he various resuls, Figure 9 shows a comparison beween he hree cases and he cos associaed wih he household load and EV wihou an energy managemen program. Fig. 9. Supply coss depending on he case In he plo, each bar is associaed wih he cos of a es case. The order of appearance of he es cases, lef o righ, is he following: I.0, I.3, I.1 and I.2.

52 44 5. IMPLEMENTATION AND RESULTS I.0 corresponds wih a case where V2G sysem is no available and he range anxiey is conneced. This case represens he behaviour of an owner wihou energy managemen sysem: always sar charging he baery as soon as he EV is conneced. The resuls of he cases I.3 and I.1 are very similar because he freedom o manage he devices is low. However, as i has be seen in he I.2 es case, where he V2G is conneced and he range anxiey is no considered, he saving reaches he 6, 7% of he cos Exended es case. In he exended es case, he adusable demand is only allowed in exreme siuaions i.e., if here is a unexpeced issue and he demand canno be compleely supplied. To regulae his, a sudy abou he differen penaly values (K) has been performed. In he Figure 10, he supplied demand behaviour is shown in several cases: Fig. 10. Sensiiviy analysis for adusable demand penaly Each line corresponds o a differen penaly value K A. The real demand is he orange poin line and he oher lines change beween his one and he K A = 0 blue solid line, where i is free o decrease he adusable load and he allowed minimum for adusable demand is reached. In his case, he value K A = 0.15 is seleced because is he firs value ha supplies he oal demand (knowing ha wih his daa i is possible o do) I.1 V2G off. In he firs es case, he possibiliy o use he discharging process in he EV baery and he range anxiey concep are no considered. On he oher hand, he sorage device can discharge power a any ime. Figure 11 shows he obained resuls. As can be seen in he figure, he sorage device is able o ake advanage of he difference beween on peak and off peak periods. The power required by he EV is supplied during he las nigh hours, which are cheapes, and does no need o be charged in he afernoon. I is also seen how he shifable demand is supplied

53 4. RESULTS FOR THE TERTIARY CONTROL PROBLEM 45 Fig. 11. Teriary energy scheduling: exended case, I.1 es. before he EV baery during off peak hours. In he on peak hours, he MT is urned on. In hese periods, he MT coss are high bu he ariff of inerconnecion is higher. To compensae his cos, he excess of power generaed by he MT is sold o he grid during he maor on peak hours I.2 V2G on/range anxiey off. As in he previous I.2 es case of IREC s proec, he energy managemen program has freedom o choose he periods of charging/discharging he EV baery. Furhermore, i is adds he possibiliy o use he sorage device and micro gas urbine. Fig. 12. Teriary resuls. Exended case wihou range anxiey In Figure 12 can be seen how he power discharged from EV baery and sorage device is enough o supply he demand during he on peak hours. The number

54 46 5. IMPLEMENTATION AND RESULTS of hours where he inerconnecion poin doesn provide power o he microgrid is higher han in he IREC s case. Noice also ha he micro gas urbine has never been commied due o is higher cos. Also i is shown how he shifable demand is supplied when he EV baery is full and he following periods are sill off peak hours I.3 Range anxiey on. Considering he same hypohesis as in he previous secion bu adding he range anxiey, he obained resuls have been summarized in Figure 13. Fig. 13. Teriary energy scheduling: exended case, I.3 es. The main difference wih he previous case is he micro gas urbine. Due o he fac ha he EV baery canno be discharged in his case, he sorage device is no enough o supply he oal load during he on peak hours and hen, he MT is urned on. As in he firs es case of his secion, o compensae he MT cos he surplus power is sold o he grid during he on peak hours. The shifable demand is supplied a lile earlier wih respec o previous case, and in addiion, he sorage device is used o provide power o he EV baery during he firs periods of connecion. For his reason, he range anxiey effecs decrease compared wih he IREC s case Comparison: exended case. The sar poin o compare he resuls obained in exended es cases is he above case I.0 where he range anxiey is considered and he V2G sysem is disconneced. The cos for 48h associaed wih his case is 16, 73e. If i is compared wih he more opimal case (I.2), he saving reaches he 5.5% of he cos.

55 5. RESULTS FOR THE SECONDARY CONTROL PROBLEM Resuls for he secondary conrol problem For his problem, single period has been seleced from he previous hree cases. The soluion obained wih his procedure and he comparison wih he eriary s soluion are shown below IREC s proec es case. The seleced period corresponds o he inerval 16: of he I.3 case where all considered unis are working a he same ime. The main characerisic of his problem are shown in Table III. The columns of oal variables and oal consrains are referred o he oal variables and consrains used hroughou he procedure. TABLE III SECONDARY CONTROL OPTIMIZATION PROBLEM. CASCADE OPTIMIZATION PROCEDURE (COP) Tes Toal Binary Toal Execuion O.F value O.F value case variables variables consrains ime firs ier. afer COP [ms] [e] [e] IREC I The idea is o ensure he inerconnecion se poins of eriary conrol soluion. This is due o he fac ha he energy marke regulaion esablishes economic penalies associaed wih he deviaions of he agreed schedules. Considering his hypohesis, he EV is he unique conrollable device in his case. Fig. 14. Secondary resul: IREC s proec case a period [16:15, 16:30] of I.3 es. Figure 14 shows as for every inerval of 30 seconds, he secondary conrol balances he oal power provided by he renewable energy sources and he commied power wih he grid, by loading o he EV baery.

56 48 5. IMPLEMENTATION AND RESULTS To compare hese resuls wih he energy scheduled by eriary conrol, he Figure 15 shows he expeced energy in blue and he energy scheduled by he cascade procedure in green. As in he previous figures, he generaed power is in he op and he supplied power is in he boom of he plo. Fig. 15. Expeced resuls by eriary conrol versus correced resuls by secondary conrol in IREC s proec case Due o he increase on he expeced wind energy and he decrease in he load, he elecric vehicle baery has higher charging han in he schedule of eriary conrol Exended es case. To es he exended case by he cascade opimizaion procedure for he secondary conrol, he ime range seleced is [16:00, 16:15] of I.2 subcase because his period is he unique inerval where he sorage device, EVB and he micro gas urbine are working a he same ime. The characerisics of his opimizaion procedure have been summarized in Table IV. TABLE IV SECONDARY CONTROL OPTIMIZATION PROBLEM. CASCADE OPTIMIZATION PROCEDURE (COP) Tes Toal Binary Toal Execuion O.F value O.F value case variables variables consrains ime firs ier. afer COP [ms] [e] [e] Exended I In his case, i can be seen how he secondary conrol disribues he load of he EV hroughou he quarer hour and decides he power generaed by he micro gas urbine depending on he power imbalance beween he res of componens.

57 5. RESULTS FOR THE SECONDARY CONTROL PROBLEM 49 Fig. 16. Secondary resul: Exended case a inerval [16:00, 16:15] of I.3 es. As in he previous cases, he Figure 16 shows he generaed power in he op and he demanded power in he boom of he plo. In he resuls, i is no observed any unexpeced even during his inerval and consequenly, he schedule of eriary conrol is correcly me. However, here are some small deviaions due o he increase on he household load as i can seen in he Figure 17. Fig. 17. Scheduled resuls of he eriary conrol versus correced resuls by secondary conrol in exended case