IEEE TRANSACTIONS ON SMART GRID 1 Smar Household Operaion Considering Bi-Direcional EV and ESS Uilizaion by Real-Time Pricing-Based DR Ozan Erdinc, Member, IEEE, Nikolaos G. Paerakis, Suden Member, IEEE, Tiago D. P. Mendes, Anasasios G. Bakirzis, Senior Member, IEEE, and João P. S. Caalão, Senior Member, IEEE Absrac As he smar grid soluions enable acive consumer paricipaion, demand response (DR) sraegies have drawn much ineres in he lieraure recenly, especially for residenial areas. As a new ype of consumer load in he elecric power sysem, elecric vehicles (EVs) also provide differen opporuniies, including he capabiliy of uilizing EVs as a sorage uni via vehicle-o-home (V2H) and vehicle-o-grid (V2G) opions insead of peak power procuremen from he grid. In his paper, as he main conribuion o he lieraure, a collaboraive evaluaion of dynamic-pricing and peak power limiing-based DR sraegies wih a bi-direcional uilizaion possibiliy for EV and energy sorage sysem (ESS) is realized. A mixed-ineger linear programming (MILP) framework-based modeling of a home energy managemen (HEM) srucure is provided for his purpose. A disribued small-scale renewable energy generaion sysem, he V2H and V2G capabiliies of an EV ogeher wih wo-way energy rading of ESS, and differen DR sraegies are all combined in a single HEM sysem for he firs ime in he lieraure. The impacs of differen EV owner consumer preferences ogeher wih he availabiliy of ESS and wo-way energy rading capabiliies on he reducion of oal elecriciy prices are examined wih case sudies. Index Terms Demand response (DR), elecric vehicle (EV), home energy managemen (HEM), smar household, vehicle-o-grid (V2G), vehicle-o-home (V2H). NOMENCLATURE The main nomenclaure used hroughou his paper is saed below. I is o be noed ha he parameers and variables can only be relaed o ime, so he respecive index is used and no furher reference is given o his fac. Manuscrip received April 24, 2014; revised July 18, 2014; acceped Augus 22, 2014. This work was suppored in par by FEDER funds (European Union) hrough COMPETE, in par by Poruguese funds hrough FCT under Projec FCOMP-01-0124-FEDER-020282 (Ref. PTDC/EEA-EEL/118519/2010) and Projec PEs-OE/EEI/LA0021/2013, and in par by he EU Sevenh Framework Programme FP7/2007-2013 under Gran Agreemen 309048. Paper no. TSG-00347-2014. O. Erdinc is wih he Deparmen of Elecrical-Elecronics Engineering, Isanbul Arel Universiy, Isanbul 34740, Turkey (e-mail: ozanerdinc@arel.edu.r). N. G. Paerakis, T. D. P. Mendes, and J. P. S. Caalão are wih he Universiy of Beira Inerior, Covilhã 6201-001, Porugal, also wih INESC-ID, Lisbon 1000-029, Porugal, and also wih IST, Universiy of Lisbon, Lisbon 1049-001, Porugal (e-mail: nikpaerak@gmail.com; iagomendesdi@gmail.com; caalao@ubi.p). A. G. Bakirzis is wih he Arisole Universiy of Thessaloniki, Thessaloniki 54006, Greece (e-mail: bakiana@eng.auh.gr). Color versions of one or more of he figures in his paper are available online a hp://ieeexplore.ieee.org. Digial Objec Idenifier 10.1109/TSG.2014.2352650 Abbreviaions DR Demand response. ESS Energy sorage sysem. EV Elecric vehicle. HEM Home energy managemen. MILP Mixed-ineger linear programming. PV Phoovolaic. V2G Vehicle-o-grid. V2H Vehicle-o-home. Indices Period of he day index in ime unis [h or min]. Parameers CE ESS Charging efficiency of he ESS. CE EV Charging efficiency of he EV. CR ESS Charging rae of he ESS [kw per ime inerval]. CR EV Charging rae of he EV [kw per ime inerval]. DE ESS Discharging efficiency of he ESS. DE EV Discharging efficiency of he EV. DR ESS Discharging rae of he ESS [kw per ime inerval]. DR EV Discharging rae of he EV [kw per ime inerval]. N 1 Maximum power ha can be drawn from he grid [kw]. N 2 Maximum power ha can be sold back o he grid [kw]. Household power demand [kw]. P oher P PV,pro Power produced by he PV [kw]. SOE ESS,ini Iniial sae-of-energy of he ESS [kwh]. SOE ESS,max Maximum allowed sae-of-energy of he ESS [kwh]. SOE ESS,min Minimum allowed sae-of-energy of he ESS [kwh].,ini,max,min T a T d Iniial sae-of-energy of he EV [kwh]. Maximum allowed sae-of-energy of he EV [kwh]. Minimum allowed sae-of-energy of he EV [kwh]. Arrival ime of EV o household. Deparure ime of EV from household. 1949-3053 c 2014 IEEE. Personal use is permied, bu republicaion/redisribuion requires IEEE permission. See hp://www.ieee.org/publicaions_sandards/publicaions/righs/index.hml for more informaion.
2 IEEE TRANSACTIONS ON SMART GRID T f,c T f,d Period a which EV should be fully charged. Period a which EV should be fully discharged, if applicable. Number of ime inervals in 1 h. ε 1 Prioriy parameer of PV. ε 2 Prioriy parameer of ESS. ε 3 Prioriy parameer of EV. λ buy Price of energy bough from he grid [cens/kwh]. λ sell Price of energy sold back o he grid [cens/kwh]. Variables P ESS,ch P ESS,dis P ESS,sold P ESS,used P EV,ch P EV,dis P EV,sold P EV,used P grid P PV,sold P PV,used P sold SOE ESS u ESS u EV u grid ESS charging power [kw]. ESS discharging power [kw]. Power injeced o grid from he ESS [kw]. Power used o saisfy household load from he ESS [kw]. EV charging power [kw]. EV discharging power [kw]. Power injeced o grid from he EV [kw]. Power used o saisfy household load from he EV [kw]. Power supplied by he grid [kw]. Power injeced o grid from he PV [kw]. Power used o saisfy household load from he PV [kw]. Toal power injeced o he grid [kw]. Sae-of-energy of he ESS [kwh]. Sae-of-energy of he EV [kwh]. Binary variable: 1 if ESS is charging during period, 0else. Binary variable: 1 if EV is charging during period, 0else. Binary variable: 1 if grid is supplying power during period, 0else. I. INTRODUCTION A. Moivaion and Background THE DEREGULATION of he elecric power indusry is a concern of invesors, regulaors, and oher paricipans of he elecriciy marke for more han a decade wih he aim of obaining a more efficien use of elecric energy and improved profis. As a recenly growing concep for an effecive deregulaion of he elecric power indusry, he smar grid issue has drawn significan aenion from developed counry governmens, declaring considerable invesmens. Smar grid is he vision for enhancing he efficiency of elecriciy uilizaion from he producion o end-user poins, ogeher wih effecively accommodaing all generaion and sorage opions and enabling consumer paricipaion in he demand-side. Coupled wih he growing imporance of smar grid vision, smar households ha can monior heir use of elecriciy in real-ime and ac in order o lower heir elecriciy bills have also been given specific imporance by he research, regarding possible demand-side acions [1], [2]. Demand-side acions for smar households in a smar grid generally focus on DR sraegies allowing ineracion beween uiliy and consumers. DR is a erm defined as changes in elecric usage by end-use cusomers from heir normal consumpion paerns in response o changes in he price of elecriciy over ime, or o incenive paymens designed o induce lower elecriciy use a imes of high wholesale marke prices or when sysem reliabiliy is jeopardized by he U.S. Deparmen of Energy (DOE) and comprises incenivebased programs and price-based programs (ime-of-use, criical peak pricing, dynamic pricing) [3], [4]. DR sraegies generally focus on shifing he elecric usage of consumers from peak o off-peak periods o reduce he sress on uiliy-handled asses such as disribuion ransformers, lines, ec., and may provide a valuable resource for he efficien operaion of he smar grid srucure [5], [6]. The uilizaion of DR sraegies can be considered maure for indusrial cusomers, bu his is a relaively new concep for residenial households responsible for nearly 40% of energy consumpion in he world [7]. There are many enabling echnologies for DR aciviies in residenial areas. Especially, HEM sysems and smar meers have he leading role in effecively applying DR sraegies. Wih he inroducion of differen kinds of elecric loads in he marke, he load shapes of households have sared o change significanly. As a new ype of end-user appliance/load, EVs have recenly gained more imporance, as he elecrificaion of he ranspor secor which radiionally is a major fossil fuel consumer is a opic of curren ineres [8]. EVs have a differen srucure wih challenges and opporuniies ha should be examined in deail. Energy needs of EVs as a load can reach o he levels of new power plan insallaion requiremens. The recommended charging level of a Chevy Vol, a small sized EV, is 3.3 kw [9], which can even exceed he oal insalled power of many individual homes in an insular area. Besides, EVs can also be employed as a resource, especially during peak periods wih he possibiliy of V2H and V2G opions. B. Lieraure Overview There are several recen sudies dealing wih DR sraegies for he opimum appliance operaion of smar households. Chen e al. [10] and Tsui and Chan [11] developed an opimizaion sraegy for he effecive operaion of a household wih a price signal-based DR. Li and Hong [12] proposed a user-expeced price -based DR sraegy for a smar household, including also a baery-based ESS aiming a lowering he oal elecriciy cos by charging and discharging he ESS a off-peak and peak price periods, respecively. However, he impac of including an addiional EV load ha can also be helpful for peak clipping in cerain periods when EV is a home and he possibiliy of an own producion faciliy have no been evaluaed in [12]. Zhao e al. [13] considered he HEM sraegy-based conrol of a smar household including PV-based own producion faciliy and availabiliy of EV and ESS. However, V2H and
ERDINC e al.: SMART HOUSEHOLD OPERATION CONSIDERING BI-DIRECTIONAL EV AND ESS UTILIZATION 3 furher possible V2G operaing modes of EV have no been aken ino accoun in [13]. Resegar e al. [14] developed a smar home load commimen sraegy considering all he possible operaing modes of EV and ESS. However, ha paper negleced he impac of an exra peak power limiing sraegy ha is probable o be imposed by a load serving eniy (LSE), no considered also in [10] [13]. Pipaanasomporn e al. [15] and Kuzlu e al. [16] presened a HEM sraegy considering peak power limiing DR sraegy for a smar household, including boh smar appliances and EV charging. Shao e al. [17] also invesigaed EV for DR-based load shaping of a disribuion ransformer serving a neighborhood. References [15] [17] did no provide an opimum operaing sraegy considering price variabiliy wih he aim of obaining he lowes daily cos apar from jus limiing he peak power drawn from he grid by he household in cerain periods. Maallanas e al. [18] applied an HEM sysem based on neural neworks wih experimenal resuls for a household including PV and ESS. However, he impacs of varying price as well as oher ypes of DR sraegies have no been evaluaed in [18]. De Angelis e al. [19] performed he evaluaion of a HEM sraegy considering he elecrical and hermal consrains imposed by he overall power balance and consumer preferences. Chen e al. [20] provided an appliance scheduling in a smar home considering dynamic prices and appliance usage paerns of consumer. Missaoui e al. [21] also provided a smar building energy managemen sraegy based on price variaions and exernal condiions as well as comfor requiremens. The pricing daa-based energy managemen has also been suggesed by Hu and Li [22] ogeher wih a hardware demonsraion. Erdinc [23] considered boh pricing and peak power limiing DR, bu negleced he possibiliy of wo-way energy rading possibiliy for EV and ESS wih he grid, which can furher improve he economic advanage of he HEM srucure by increased flexibiliy. These papers ogeher wih many oher sudies no referred here have provided valuable conribuions o he applicaion of smar grid conceps in household areas. However, many of he menioned papers failed o address disribued renewable energy conribuion o reduce load demand on uiliy side, V2H opion of EV o lower he demand peak periods, and wo-way energy rading capabiliy of EV (wih V2G) and a possible ESS ogeher wih differen DR sraegies. C. Conribuions In his paper, a MILP model of he HEM srucure is provided o invesigae a collaboraive evaluaion of a dynamicpricing based DR sraegy, a disribued small-scale renewable energy generaion sysem, he V2H capabiliy of an EV ogeher wih wo-way energy rading of EV (using V2G opion) and ESS. To he bes knowledge of he auhors, his is he firs sudy in he lieraure combining all of he aforemenioned operaional possibiliies in a single HEM sysem formulaed in a MILP framework, which is he main novely of his paper. Fig. 1. Block diagram of a fundamenal DR sraegy for smar households. Differen case sudies are conduced considering he impacs of having a HEM sysem, an EV capable of providing V2H and V2G opions, and an addiional ESS under differen DR sraegies. The impacs of all case sudies in erms of consumer elecriciy bill reducion performance are evaluaed wih relevan comparisons. Besides, real-ime measured load demand and normalized PV-based disribued energy resource producion daa are uilized. D. Organizaion This paper is organized as follows. Secion II provides he mehodology employed in his paper. Aferwards, Secion III includes he case sudies for evaluaing daily DR-based operaion sraegies for he smar household. Finally, he conclusion is presened in Secion IV. II. METHODOLOGY The block diagram of a fundamenal DR sraegy is presened in Fig. 1. The HEM sysem regulaes he operaion of he smar household considering price-based and oher signals from he LSE, producion of small-scale own faciliies, load consumpion of smar appliances, ec., ogeher wih differen consumer preferences as seen from Fig. 1. The res of his secion presens he proposed model. The objecive is o minimize he oal daily cos of elecriciy consumpion. The cos is he difference beween he energy bough from he grid and he energy sold back o he grid by he household-owned asses ha are able o provide energy (PV, ESS, and EV). The price variables are ime dependen, a fac ha implies ime varying prices for boh bough and sold energy. The second par of he objecive funcion in (1) imposes an arificial penaly o he energy provided by he differen resources. The ε parameers have sufficienly small posiive values (such as e-7, 2e-7, and 3e-7) ha are deermined by assumpions, so ha he oal cos is no affeced. This echnique serves he need of having a prioriy in selling energy from he resources. Smaller relaive value of a specific ε forces he HEM sysem o sell firs all he energy available from ha resource
4 IEEE TRANSACTIONS ON SMART GRID before selling energy from anoher Minimize TC = ( P grid λbuy ( ε 1 PPV,sold + ) Psold λsell + ε 2 PESS,sold + ε 3 PEV,sold. (1) In (1), he opimizaion variables are he oal power bough from he grid a ime (P grid ), and he oal power sold back o he grid (P sold ) which comprises power values sold from PV, ESS, and EV (P PV,sold, P ESS,sold, and P EV,sold ). In his paper, we consider ha he HEM sysem firs sells energy from he PV, nex from he ESS, and finally from he EV baery, which means ε 1 <ε 2 <ε 3. The consrains presened hereafer comprise he basic body of he HEM sysem operaion. The model can be easily exended and adaped o oher more specific implemenaions (e.g., by furher modeling specific smar-appliances such as HVAC, waer heaers, appliances wih cycling operaion and/or cusomer s conrac deails). Any ime granulariy can be used simply by selecing he appropriae. For insance, for a 15-min inerval he coefficien mus be 4, as 1 h comprises four 15-min inervals. A. Power Balance P grid + P PV,used + P EV,used = P oher + P ESS,used + P EV,ch ) + P ESS,ch,. (2) Equaion (2) saes ha he load consising of he residenial load (P oher ), he charging needs of he EV (P EV,ch ), and he ESS (P ESS,ch ) is eiher saisfied by he grid (P grid )or by he combined procuremen of energy by he PV, he ESS, and he EV (P PV,used, P EV,used, and P ESS,used ). B. ESS Modeling P ESS,used + P ESS,sold P ESS,ch = P ESS,dis DE ESS, (3) CR ESS u ESS, (4) ( ) 1 u ESS, (5) P ESS,dis DR ESS SOE ESS = SOE ESS 1 + CE ESS PESS,ch PESS,dis, 1 (6) SOE ESS = SOE ESS,ini, if = 1 (7) SOE ESS SOE ESS,max, (8) SOE ESS SOE ESS,min,. (9) Equaion (3) enforces he fac ha he acual power provided by he ESS discharge (P ESS,dis DE ESS ) can be used o cover a porion of he household needs (P ESS,used ) or injeced back o he grid (P ESS,sold ). Consrains (4) and (5) impose a limi on he charging (P ESS,ch ) and discharging (P ESS,dis ) power of he ESS. The idle ESS sae can be described by any of hese consrains by he ime he respecive power variable is allowed o have zero value. Equaions (6) (9) describe he sae-of-energy of he ESS. Consrain (6) forces he sae-of-energy a every inerval (SOE ESS ) o have he value ha i had a he previous inerval (SOE ESS 1 ) plus he acual amoun of energy ha is ransferred o he baery if i is charging a ha inerval minus he energy ha is subraced if he baery is discharging during ha inerval. A he beginning of he ime horizon he saeof-energy of he ESS coincides wih he iniial sae-of-energy of he ESS (SOE ESS,ini ), as described by (7). Consrain (8) limis he sae-of-energy of he baery o be less han he ESS capaciy (SOE ESS,max ). Similarly, (9) prevens he deep discharge of he baery by imposing a leas sae-of-energy limi (SOE ESS,min ). C. EV Modeling P EV,used + P EV,sold P EV,ch P EV,dis = P EV,dis DE EV, [T a, T d ] (10) CR EV u EV, [T a, T d ] (11) DR EV (1 u EV ), [T a, T d ] (12) = 1 + CE EV PEV,ch PEV,dis, [Ta, T d ] (13) =,ini, if = T a (14),max, [T a, T d ] (15),min, [T a, T d ] (16) =,max, T f,c [T a, T d ] (17) =,min, if = T f,d [T a, T d ] (18) = P EV,used = P EV,sold = P EV,dis = P EV,ch = 0, / [T a, T d ]. (19) Equaion (10) enforces he fac ha he acual power provided by he EV discharge (P EV,dis DE EV ) can be used o cover a porion of he household needs (P EV,used ) or injeced back o he grid (P EV,sold ). Consrains (11) and (12) impose a limi on he charging (P EV,ch ) and discharging (P EV,dis )power of he EV. The idle EV sae can be described by any of hese consrains by he ime he respecive power variable is allowed o have zero value. Equaions (13) (17) describe he sae-of-energy of he EV. Consrain (13) forces he saeof-energy a every inerval ( ) o have he value ha i had a he previous inerval ( 1 ) plus he acual amoun of energy ha is ransferred o he EV baery if i is charging a ha inerval minus he energy ha is subraced if he EV baery is discharging during ha inerval. A he arrival ime of EV o household, he sae-of-energy of he EV coincides wih he iniial sae-of-energy of he EV (,ini ), as described by (14). Consrain (15) limis he sae-of-energy of he EV baery o be less han is capaciy (,max ).
ERDINC e al.: SMART HOUSEHOLD OPERATION CONSIDERING BI-DIRECTIONAL EV AND ESS UTILIZATION 5 Similarly, (16) prevens he deep discharge of he EV baery by imposing a leas sae-of-energy limi (,min ). Equaions (17) and (18) represen he opion of having he EV baery fully charged or discharged a he leas sae-ofenergy a preseleced ime inervals. Finally, (19) ensures ha all he variables relaed o EV modeling are zero apar from he ime inerval beween arrival ime of EV o household (T a ) and deparure ime of EV from household (T d ). D. PV Modeling P PV,used + P PV,sold = P PV,pro,. (20) Similarlyo(3) and (10), (20) enforces he fac ha he acual power provided by he PV (P PV,pro ) can be used o cover a porion of he household needs (P PV,used ) or injeced back o he grid (P PV,sold ). E. Toal Power Injeced o he Grid Fig. 2. Real-ime measured average household power demand. P sold = P PV,sold + P ESS,sold + P EV,sold,. (21) The oal amoun of power injeced o he grid (P sold ) consiss of he amoun of power provided by he PV (P PV,sold ), he ESS (P ESS,sold ), and he EV (P EV,sold ) as menioned before. This is enforced by (21). F. Power Transacion Resricions P grid N 1 u grid, ( ) (22) 1 u grid,. (23) P sold N 2 Equaions (22) and (23) implemen he logic of power exchange. If power from he grid is needed o be drawn, hen i is no possible o injec power back o he grid. The reverse case is also described by hese consrains. N 1 is a posiive ineger value ha imposes a limiaion on he power ha can be drawn from he grid. This limiaion may represen a resricion posed by he aggregaor or he responsible eniy for he enduser elecrificaion in order o face he siuaion where in is conrol area exis muliple households ha own HEM sysem. The implemenaion of a ime-varying peak power drawn from he grid limi as a differen DR sraegy can be easily adaped on his formulaion, by replacing he N 1 by a ime-dependen parameer. Similarly, N 2 imposes a limi on he power ha can be injeced back o he grid and also can be replaced by a ime-dependen parameer. Differen consumer opions and behavioral deails can be expressed by fixing he charging and discharging variables of he ESS and EV o be zero in he appropriae ime inervals. Differen policies (e.g., energy selling back opions) can be modeled by fixing he selling energy/power variables o zero or oher desired values. III. TEST AND RESULTS To evaluae he oal impac of differen case sudies in household operaion on consumer elecriciy bills, he MILP model is esed in GAMS v.24.1.3 using he solver Fig. 3. PV sysem power producion curve. CPLEX v.12 [24] and he relevan obained resuls are discussed in his secion. The real-ime measured load demand of an average house in Porugal is used in his paper. The nearly 140 meer-square household includes four habians wih differen elecric appliances, including fridge, TVs, microwave, washing machine and dishwasher, compuer, oven, ec. I should be noed ha he household includes a waer heaer using gas insead of elecriciy. The consumpion of each day in a period of one monh was recorded and he obained average power consumpion profile of his period is shown in Fig. 2. I is considered in his paper ha he household includes a small-scale PV sysem of 1 kw. The producion daa of he menioned PV sysem is he normalized version of a measured daily solar farm producion profile. The considered PV sysem power producion curve is given in Fig. 3. A bi-direcional EV operaion including boh V2G (meaning ha EV sells energy back o he grid) and V2H (meaning ha a porion of he energy sored in EV baery is used o parly cover he household load) opions is considered. The specificaions of a Chevy Vol wih a baery raing of 16 kwh is aken ino accoun. The Chevy Vol is employed wih a charging saion limied o a charging power of 3.3 kw [9]. The same power limi is also assumed o be valid for he discharging operaion in V2G and V2H modes. The charging and discharging efficiencies are considered 0.95. I is also considered ha he iniial EV baery energy is 8 kwh (50% sae-of-energy) while arriving a home and he lower limi of EV sae-of-energy is resriced o 4.8 kwh (30% sae-of-energy) o avoid deep-discharging (a limi around he level proposed by [25], announcing ha he baery users
6 IEEE TRANSACTIONS ON SMART GRID Fig. 4. Time-varying dynamic price signal for DR. Fig. 5. Toal household power demand for consumers willing o charge heir EV immediaely. should no exrac more han 70% 80% of he available capaciy a any ime). The following assumpions hold for he ESS; i consiss of a baery group of 1 kwh capaciy. The charging and discharging rae per hour is assumed o be 0.2 kw. Is iniial sae-ofenergy and charging/discharging efficiencies are 0.5 kwh and 0.95, respecively, and is deep-discharging limi is 0.25 kwh. I should be noed ha in he considered concep, no cos is associaed wih using sorage faciliies such as EV and ESS during he HEM operaion. Inegraing he wo-way energy ransacions beween he end-user and he uiliy, he ne-meering approach is uilized. When he available energy from he household-owned resources is sufficien o cover he oal of he needs, he excess of energy can be sold back o he grid and vice versa. For pricing he bough energy from he grid, a dynamic pricingbased DR scheme is considered. The ime-varying price signal available for he consumer via he smar meer is shown in Fig. 4 [11]. Besides, a fla rae of 3 cens/kwh is paid o he end-user for he energy sold-back o he grid. Paymen of fla raes wih ne meering is an approach also used in pracice such as he case in Turkey. A dynamically changing rae for energy sold can also be easily applied wihin he provided formulaion, as (1) is suiable boh for considering fla and dynamic raes. DR sraegies, especially price-based DR aciviies are mainly considering he preferences of he consumer, and he preferences of he consumers may vary individually. Thus, firs of all, sole consumer preferences-based manual operaion wihou HEM sraegy is analyzed in his paper. Three ypes of consumer preferences, namely consumers willing o charge heir EV immediaely, consumers willing o charge heir EV wih lower prices, and consumers willing o charge heir EV wih lower prices ogeher wih uilizaion of EV V2H opion for peak household power demand periods, are evaluaed. Fig. 5 presens he oal household power demand for consumers willing o charge heir EV immediaely afer arriving home a 6:00 P.M. I can be clearly seen ha he EV load conribues significanly o he available peak period in he load demand given in Fig. 2, and his peak reaches nearly 5 kw insead of he available peak power value of 1.7 kw in he household power demand. Here, he prices where he EV charging significanly conribues o he peak load are a he highes level compared Fig. 6. Toal household power demand for consumers willing o charge heir EV wih lower prices (wihin he period saring from 10 P.M.). o he oher periods of he day: 4 cens/kwh a 6 P.M. and 4.5 cens/kwh a 7 P.M. This issue surely has an impac on he oal daily cos of household power demand supply. The second preference of he consumer for charging EV wih lower prices wihin he period saring from 10 P.M. provides he oal household power demand profile shown in Fig. 6. This ype of operaion leads o 51.7 cens daily elecriciy consumpion cos. I should be noed ha he reason for presening hese moneary values for his case and he cases ha will be discussed below is o beer presen he impac of differen preferences and he proposed mehodology on he cos reducion for he daily operaion of a household. These moneary values were no jus given as numbers; insead, hey will furher be used o give percenages for he cos reducion for each case compared o a base case. To be able o provide a comparaive analysis in order o presen he meris of he proposed mehodology, such percenages will be necessary. Shifing furher he EV charge o even more low-price periods afer midnigh saring from 2 A.M., is also considered as a differen case. This leads o a significanly lower oal cos of 33.3 cens. However, his issue has a serious disadvanage of providing new peaks in normally off-peak periods of uiliy load as seen in Fig. 7 and requires a furher power limiing acion like in [12] [14]. As a furher case sudy, he consumer s will o charge he EV wih lower prices ogeher wih an EV V2H opion o decrease he energy procuremen from he grid during peak price periods is examined. The oal household power demand profile is shown in Fig. 8. I is considered in his case sudy ha as soon as he EV owner arrives home a 6 P.M., he EV is plugged-in and he household power demand is supplied by he
ERDINC e al.: SMART HOUSEHOLD OPERATION CONSIDERING BI-DIRECTIONAL EV AND ESS UTILIZATION 7 Fig. 7. Toal household power demand for consumers willing o charge heir EV wih lower prices (wihin he period saring from 2 A.M.). Fig. 10. Toal household power demand for consumers via proposed HEM sraegy wihou EV V2H opion. Fig. 8. Toal household power demand for consumers willing o charge heir EV wih lower prices ogeher wih EV V2H opion (wihin he period saring from 10 P.M.). Fig. 9. EV baery sae-of-energy variaion for consumers willing o charge heir EV wih lower prices ogeher wih EV V2H opion (wihin he period saring from 10 P.M.). EV unil he baery energy reaches he lower baery energy limi. Afer reaching his limi, procuremen of energy from he grid sars again. Besides, he EV is charged during lower price periods. The oal daily consumpion cos is 49.6 cens if he EV is charged wihin he period saring from 10 P.M. The EV baery sae-of-energy variaion in his period is presened in Fig. 9. As can be seen, he sored energy level of he EV baery reduces while conneced o he household in V2G mode o he predefined lower limi of discharge and he EV baery remains idle afer his period unil he charging process wihin he period saring from 10 P.M. Then, he EV baery is charged wih he maximum allowed charging power unil i is fully charged for he day-ahead uilizaion of he consumer. I should be noed here for Figs. 8 and 9 (and for all oher figures) ha he hour wrien below he figure corresponds o he ime inerval beween he wrien hour and he nex hour. For example, 6 P.M. (he ime EV arrives home) corresponds Fig. 11. Toal household power demand for consumers via proposed HEM sraegy wih EV V2G opion. o he ime inerval from 6 o 7 P.M. infig.9. Thisiswhy he sae-of-energy value is less han he iniial sae-of-energy value of EV baery a 6 P.M. infig.9 as 6 P.M., also includes he uilized energy from EV ill 7 P.M. If furher lower price periods afer midnigh saring from 2 P.M. are considered for he EV charging wih V2H opion in his case sudy, he cos decreases o 23.6 cens. This is he resul of he combined impac of V2H opion in peak periods and EV charging in lower price periods. Since now, manual DR aciviies have been analyzed in erms of he impac of differen consumer preferences on daily elecriciy cos. However, as he main advanages of a smar household are considered o be clearer wih he implemenaion of an auomaic HEM sysem, he impacs of employing such a sysem on coss for differen opions is also evaluaed in his paper. The HEM sysem considers he daily elecriciy prices declared by he eniy ha serves he load, ogeher wih regular load demand paerns of he household o decide he opimum operaing sraegy. Firsly, he EV charging by opimizaion-based HEM sraegy wihou V2H opion is evaluaed. The HEM-based EV charging sraegy resuls in he oal household power demand shown in Fig. 10. As seen in Fig. 10, he HEM sraegy auomaically shifs he EV charging afer 2 A.M. and especially afer 4 A.M., and he EV charging power is a is highes level due o he lowes elecriciy prices hroughou he day. This ype of operaion leads o a oal daily elecriciy cos of 30.5 cens ha is considerably lower han he case where consumers manually decide he charging ime of heir EVs wihou V2H opion. As a furher case sudy, he HEM-based EV operaion wih he V2H opion is examined. The beginning of EV charging is auomaically shifed o 3 A.M. as can be seen in Fig. 11,
8 IEEE TRANSACTIONS ON SMART GRID Fig. 12. EV baery sae-of-energy variaion for consumers via proposed HEM sraegy wih EV V2H opion. Fig. 14. EV and ESS sae-of-energy variaions for consumers wih EV V2H-V2G and ESS2H-ESS2G opions. Fig. 13. Decomposiion of household power demand saisfacion via proposed HEM sraegy for consumers wih EV V2H-V2G and ESS2H-ESS2G opions. Fig. 15. Decomposiion of household power demand saisfacion via proposed HEM sraegy for consumers wih EV V2H-V2G and ESS2H-ESS2G opions furher resriced by peak power limiing DR. similarly o he previous case sudy where he HEM-based EV operaion is evaluaed wihou V2H opion. The sae-of-energy variaion of EV baery in his case is shown in Fig. 12. The EV baery sored energy level reduces unil i reaches 30%, remaining a his level while EV is in idle mode and reaches 100% afer charging, respecively, hroughou he daily operaion. This HEM-based operaion wih addiional V2H opion leads o a oal daily cos of 21.9 cens. As he las case sudy, a household including an addiional ESS ogeher wih he capabiliy of wo-way energy rading wih he grid via V2G and ESS-o-grid (ESS2G) opions of EV and ESS, apar from he regular V2H and ESS-o-home (ESS2H) operaions, is considered. This addiional ESS aids he peak clipping and valley filling by charging using power produced by he PV or bough by he grid and discharging in peak price periods. The relevan resuls concerning he power balance for load supply are presened in Fig. 13. I is obvious ha he ESS and EV supply a varying porion of he load and are charged in off-peak price periods. The corresponding resuls of ESS and EV sae-of-energy are shown in Fig. 14, where he charging/discharging cycles of ESS and EV are seen in deail, and i is eviden ha hey are direcly affeced by he fla selling-back o he grid price. I should be noed ha he lef side y-axis of Fig. 14 corresponds o he EV sae-of-energy inerval while he righ side y-axis corresponds o he ESS sae-of-energy inerval. All he case sudies based on HEM provide new significan peaks in former off-peak periods ill now. As he HEM sysem auomaically shifs he charging of EV o lower price periods, his is likely o happen in real-life condiions. Thus, as an exra evaluaion under he las case sudy, a peak power limiing DR sraegy is also considered in addiion o price-based DR sraegy ha is much likely o be faced in real life as LSE can limi he power ha is drawn from he grid in cerain peak power periods o avoid more sharp peaks similar o [15] [17]. All he operaional possibiliies of PV, EV, and ESS are sill available. This peak power limiing operaion is conduced beween 7 P.M. and 6 A.M. wih a peak power limi of 2 kw in his paper, and he relevan resuls are presened in Fig. 15. Due o he limiaions during periods where EV charge is shifed in Fig. 13, he EV charging has o begin earlier his ime in order o have a fully charged EV baery in he morning, which leads o he uilizaion of more power from he grid in higher price periods. This causes an exra cos which resuls in a oal daily cos of 29.6 cens. However, in erms of his exra cos, he new peak periods faced in afer midnigh periods are prevened, as can be seen from Fig. 15. The comparison of he differen case sudies in his paper is summarized in Table I. I is clear ha differen consumer preferences in a smar household have a significan impac on daily operaing coss. The wors case scenario is considered as consumers having no ESS and willing o charge heir EV immediaely, which is significanly close o our curren daily habis, unforunaely. Compared o his base case, he oal HEM sraegy wih all opporuniies of EV and ESS operaion provides a cos reducion of 65.3%. I is also clear ha he addiional V2G opion of EV ogeher wih he employmen of an exra ESS provides a reducion of nearly 3% compared o he case where HEM
ERDINC e al.: SMART HOUSEHOLD OPERATION CONSIDERING BI-DIRECTIONAL EV AND ESS UTILIZATION 9 TABLE I COMPARISON OF DIFFERENT CASE STUDIES Fig. 16. Energy sold-selling price relaion for consumers wih EV V2H-V2G and ESS2H-ESS2G opions. sraegy wih jus EV V2H opion and wihou ESS is considered. As menioned before, he fla rae paid o he household owner for selling energy back o he grid was considered as 3 cens/kwh. Fig. 16 presens he effec of differen fla raes on he decisions of he HEM sysem for energy exchanges. The fla raes considered in he menioned evaluaion can also be observed from he perspecive of he raio of he selling price o he average buying price, which is 2.93 cens/kwh as derived from Fig. 4. Accordingly, hese raios for he considered fla raes of 1, 2, 3, 4, and 5 cens/kwh are respecively calculaed as 34.1%, 68.2%, 102.3%, 136.4%, and 170.5%. If he resuls in Fig. 16 are examined, he more he fla rae is, he more energy is sold back o he grid. This leads o a more profiable operaion, even leading o a minus cos (profi). The prices used in his paper are assumpion-based and can change from region o region relaed o many facors. For insance, in resrucured power sysems, eniies ha serve he individual loads ha do no paricipae immediaely in he marke (e.g., aggregaors, reailers, ec.) provide a price signal ha allows hem o maximize heir profis in a conex of providing he leas possible prices o he end-users, so ha hey can reain hem as cusomers in a compeiive environmen. Also, he price a which hey are willing o buy elecriciy back from he end-users (fla or dynamic) is deermined by he same raionale. Also, energy pricing can be affeced by he sae policy ha promoes he developmen of specific echnology markes (e.g., like many EU counries have been giving significan economic incenives and subsides in order o promoe small- and large-scale solar energy sysems). This paper included an offline opimizaion ha decided he scheduling of appliances for he 24-h operaion wih he assumpion of a day before noice for he price signaling and perfec knowledge of he user s habis. I should be noed ha for he proposed echnique o be effecive, an esimae of he preferences of differen EV owners by he LSE can also be necessary. Neverheless, several soluions o his problem are already provided in he lieraure by dynamic EV scheduling [26], demographical daa-based esimaion [27], and probabilisic power flow calculaions wih EV uncerainy [28]. Also, forecasing ools can be successfully employed as par of he smar household infrasrucure, which provides deailed daa o he LSE from each end-user premise, and in urn can provide enough informaion o obain a forecas of he general preferences of EV owners. The key challenge for implemening he proposed idea can be he compuaional efficiency. I akes jus 0.11 s o solve he problem for he las case as an example using a Dual Core Lapop wih 2 GHz CPU and 8 GB RAM, which can give an insigh of he compuaion ime required for he mehodology. The developed model ha is uilized in off-line way of applicaion in his paper can also be modified o be employed in online way by using dynamic programming. The uncerainy relaed o he deerminisic PV sysem power producion curve in offline mode can be handled using forecasing ools ha are frequenly used boh in small and big size of applicaions. The uncerainy of knowledge of dynamic pricing daa for upcoming hours can also be solved by shorening he scheduling horizon, considering he horizon of pricing daa sen from LSE via smar meering. Besides, he uncerainies relaed o he sae-of-energy of EV when arriving home can be solved wih a second sage of opimizaion o adop he operaion scheduling via upcoming such as real-ime daa. Tools such as neural neworks could also be fed wih daily daa and herefore such ools could defer he need for mulisage programming. IV. CONCLUSION In his paper, as he main conribuion o he lieraure on smar household operaion, he invesigaion of a collaboraive evaluaion of dynamic-pricing and peak power limiing-based DR sraegies, a disribued small-scale renewable energy generaion sysem, he V2H and V2G capabiliies of an EV ogeher wih wo-way energy rading of EV (using V2G opion) and ESS was provided using a MILP framework-based modeling of a HEM srucure. Two-way energy exchange was allowed hrough ne meering. The energy drawn from he grid has a real-ime cos, while he energy sold back o he grid is
10 IEEE TRANSACTIONS ON SMART GRID considered o be paid a fla rae. This paper makes wo basic assumpions. Firsly, he complee real-ime pricing signal is known perfecly before he beginning of he offline opimizaion horizon. Also, he user preferences and consumpion behavior are assumed o be accuraely known. Real daa from a ypical four-member Poruguese family house and a PV plan were used. Several es cases were examined. The impacs of an exra DR sraegy based on peak power limiing were also invesigaed. A he base case i was assumed ha consumers were willing o charge heir EV as soon as hey arrive home and hey own neiher HEM sysem, nor ESS. Compared o his base case, which is also associaed wih he mos expensive daily operaion, he proposed sraegy provided a more efficien operaion by means of elecriciy cos reducion, reaching abou 65%, which is significan. By adding more smar echnologies, he operaion ha is coordinaed by a HEM sysem offers a more economically efficien use of elecriciy. Indeed, smar echnologies ha will emerge in he fuure will provide more flexibiliy and economic possibiliies for an end-user o paricipae ino he power marke, provided ha he elecriciy marke regulaory framework keeps up wih he echnological advances. Surely insallaion coss should also be considered in order o assess he acual benefis of such invesmens. The proposed mehodology can be easily adaped o larger formulaions including shifable appliances (washing machine, dishwasher, ec.) and oher conrollable appliances (HVAC, ec.) for he exension of he smar household concep. The opimum operaion of a neighborhood consising of muliple smar households is also an easily adapable exension of he proposed mehodology, changing he objecive funcion o be a minimizaion or maximizaion problem from he perspecive of LSE of a muliobjecive problem considering boh he benefis of LSE and end-user household owner. Thus, he model can provide a good basis for expanding is use in larger scales. REFERENCES [1] C. W. Gellings, The Smar Grid: Enabling Energy Efficiency and Demand Response. Boca Raon, FL, USA: CRC Press, 2009. [2] T. Perumal, A. R. Ramli, and C. Y. Leong, Design and implemenaion of SOAP-based residenial managemen for smar home sysems, IEEE Trans. Consum. Elecron., vol. 54, no. 2, pp. 453 459, May 2008. [3] S. Borlease, Smar Grids: Infrasrucure, Technology and Soluions. Boca Raon, FL, USA: CRC Press, 2013. [4] P. Palensky and D. Dierich, Demand side managemen: Demand response, inelligen energy sysems, and smar loads, IEEE Trans. Ind. Informa., vol. 7, no. 3, pp. 381 388, Aug. 2011. [5] A. Khodaei, M. Shahidehpour, and S. Bahramirad, SCUC wih hourly demand response considering ineremporal load characerisics, IEEE Trans. Smar Grid, vol. 2, no. 3, pp. 564 571, Sep. 2011. [6] H. Saele and O. S. Grande, Demand response from household cusomers: Experiences from a pilo sudy in Norway, IEEE Trans. Smar Grid, vol. 2, no. 1, pp. 102 109, Mar. 2011. [7] K. J. Chua, S. K. Chou, W. M. Yang, and J. Yan, Achieving beer energy efficien air condiioning A review of echnologies and sraegies, Appl. Energy, vol. 104, pp. 87 104, Apr. 2013. [8] W. Su, H. R. Eichi, W. Zeng, and M. Y Chow, A survey on he elecrificaion of ransporaion in a smar grid environmen, IEEE Trans. Ind. Informa., vol. 8, no. 1, pp. 1 10, Feb. 2012. [9] (2014, Jul.). GM Chevy Vol Specificaions [Online]. Available: hp://gm-vol.com/full-specificaions/ [10] Z. Chen, L. Wu, and Y. Fu, Real-ime price-based demand response managemen for residenial appliances via sochasic opimizaion and robus opimizaion, IEEE Trans. Smar Grid, vol. 3, no. 4, pp. 1822 1831, Dec. 2012. [11] K. M. Tsui and S. C. Chan, Demand response opimizaion for smar home scheduling under real-ime pricing, IEEE Trans. Smar Grid, vol. 3, no. 4, pp. 1812 1821, Dec. 2012. [12] X. Li and S. Hong, User-expeced price-based demand response algorihm for a home-o-grid sysem, Energy, vol. 64, pp. 437 449, Jan. 2014. [13] J. Zhao, S. Kucuksari, E. Mazhari, and Y. J. Son, Inegraed analysis of high-peneraion PV and PHEV wih energy sorage and demand response, Appl. Energy, vol. 112, pp. 35 51, Dec. 2013. [14] M. Resegar, M. F. Firuzabad, and F. Aminifar, Load commimen in a smar home, Appl. Energy, vol. 96, pp. 45 54, Aug. 2012. [15] M. Pipaanasomporn, M. Kuzlu, and S. Rahman, An algorihm for inelligen home energy managemen and demand response analysis, IEEE Trans. Smar Grid, vol. 3, no. 4, pp. 2166 2173, Dec. 2012. [16] M. Kuzlu, M. Pipaanasomporn, and S. Rahman, Hardware demonsraion of a home energy managemen sysem for demand response applicaions, IEEE Trans. Smar Grid, vol. 3, no. 4, pp. 1704 1711, Dec. 2012. [17] S. Shao, M. Pipaanasomporn, and S. Rahman, Demand response as a load shaping ool in an inelligen grid wih elecric vehicles, IEEE Trans. Smar Grid, vol. 2, no. 4, pp. 624 631, Dec. 2011. [18] E. Maallanas e al., Neural nework conroller for acive demand-side managemen wih PV energy in he residenial secor, Appl. Energy, vol. 91, pp. 90 97, Mar. 2012. [19] F. de Angelis, M. Boaro, S. Squarini, F. Piazza, and Q. Wei, Opimal home energy managemen under dynamic elecrical and hermal consrains, IEEE Trans. Ind. Informa., vol. 9, no. 3, pp. 1518 1527, Aug. 2013. [20] X. Chen, T. Wei, and S. Hu, Uncerainy-aware household appliance scheduling considering dynamic elecriciy pricing in smar home, IEEE Trans. Smar Grid, vol. 4, no. 2, pp. 932 941, Jun. 2013. [21] R. Missaoui, H. Joumaa, S. Ploix, and S. Bacha, Managing energy smar homes according o energy prices: Analysis of a building energy managemen sysem, Energy Build., vol. 71, pp. 155 167, Mar. 2014. [22] Q. Hu and F. Li, Hardware design of smar home energy managemen sysem wih dynamic price response, IEEE Trans. Smar Grid, vol. 4, no. 4, pp. 1878 1887, Dec. 2013. [23] O. Erdinc, Economic impacs of small-scale own generaing and sorage unis, and elecric vehicles under differen demand response sraegies for smar households, Appl. Energy, vol. 126, pp. 142 150, Aug. 2014. [24] CPLEX 12 Solver Descripion [Online]. Available: hp://www.gams.com/dd/docs/solvers/cplex.pdfg [25] A. McEvoy, T. Markvar, and L. Casaner, Pracical Handbook of Phoovolaics: Fundamenals and Applicaions, 2nd ed. Amserdam, The Neherlands: Elsevier, 2012. [26] K.N.Kumar,B.Sivaneasan,P.H.Cheah,P.L.So,andD.Z.W.Wang, V2G capaciy esimaion using dynamic EV scheduling, IEEE Trans. Smar Grid, vol. 5, no. 2, pp. 1051 1060, Mar. 2014. [27] D. Seen, L. A. Tuan, O. Carlson, and L. Berling, Assessmen of elecric vehicle charging scenarios based on demographical daa, IEEE Trans. Smar Grid, vol. 3, no. 3, pp. 1457 1468, Sep. 2012. [28] G. Li and X. P. Zhang, Modeling of plug-in hybrid elecric vehicle charging demand in probabilisic power flow calculaions, IEEE Trans. Smar Grid, vol. 3, no. 1, pp. 492 499, Mar. 2012. Ozan Erdinc (M 14) received he B.Sc., M.Sc., and Ph.D. degrees from Yildiz Technical Universiy, Isanbul, Turkey, in 2007, 2009, and 2012, respecively. He has been a Pos-Docoral Fellow under EU FP7 Projec SiNGULAR a he Universiy of Beira Inerior, Covilhã, Porugal, since 2013. He is currenly an Assisan Professor wih he Elecrical- Elecronics Deparmen, Isanbul Arel Universiy, Isanbul. His curren research ineress include hybrid renewable energy sysems, elecric vehicles, power sysem operaion, and smar grid echnologies.
ERDINC e al.: SMART HOUSEHOLD OPERATION CONSIDERING BI-DIRECTIONAL EV AND ESS UTILIZATION 11 Nikolaos G. Paerakis (S 14) received he Dipl.Eng. degree from he Deparmen of Elecrical and Compuer Engineering, Arisole Universiy of Thessaloniki, Thessaloniki, Greece, in 2013. He is currenly pursuing he Ph.D. degree under EU FP7 Projec SiNGULAR from he Universiy of Beira Inerior, Covilhã, Porugal. His curren research ineress include power sysem operaion and planning, renewable energy inegraion, ancillary services, demand response, and smar grids. Anasasios G. Bakirzis (S 77 M 79 SM 95) received he Dipl.Eng. degree from he Deparmen of Elecrical Engineering, Naional Technical Universiy, Ahens, Greece, in 1979, and he M.S.E.E. and Ph.D. degrees from he Georgia Insiue of Technology, Alana, GA, USA, in 1981 and 1984, respecively. Since 1986, he has been wih he Elecrical Engineering Deparmen, Arisole Universiy of Thessaloniki, Thessaloniki, Greece, where he is currenly a Professor. His curren research ineress include power sysem operaion, planning, and economics. Tiago D. P. Mendes received he M.Sc. degree from he Universiy of Beira Inerior, Covilhã, Porugal, in 2013, where he is currenly pursuing he Ph.D. degree. His curren research ineress include smar homes/buildings, energy managemen, demand response, and smar grids. João P. S. Caalão (M 04 SM 12) received he M.Sc. degree from he Insiuo Superior Técnico, Lisbon, Porugal, in 2003, and he Ph.D. and Habiliaion degrees for Full Professor ( Agregação ) from he Universiy of Beira Inerior (UBI), Covilhã, Porugal, in 2007 and 2013, respecively. He is currenly a Professor wih UBI and a Researcher wih INESC-ID. He is he Primary Coordinaor of he EU-funded FP7 Projec SiNGULAR. His curren research ineress include power sysem operaions and planning, disribued renewable generaion, demand response, and smar grids. He has auhored or co-auhored over 250 papers published in journals, book chapers, and conference proceedings, and has supervised over 20 Pos-Docoral Fellows, Ph.D., and M.Sc. sudens. He is he Edior of he book eniled Elecric Power Sysems: Advanced Forecasing Techniques and Opimal Generaion Scheduling, (CRC Press, 2012), and also ranslaed ino Chinese in 2014. Prof. Caalão is an Edior of he IEEE TRANSACTIONS ON SMART GRID, he IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, and an Associae Edior of IET Renewable Power Generaion.