Natural Aggregation Approach based Home Energy Manage System with User Satisfaction Modelling

Size: px
Start display at page:

Download "Natural Aggregation Approach based Home Energy Manage System with User Satisfaction Modelling"

Transcription

1 IOP Conference Seres: Earth and Envronmental Scence PAPER OPEN ACCESS Natural Aggregaton Approach based Home Energy Manage System wth User Satsfacton Modellng To cte ths artcle: F J Luo et al 2017 IOP Conf. Ser.: Earth Envron. Sc Related content - Savng water n showers R A Alkhaddar, D Phpps, R Morgan et al. - Analyss of Student Satsfacton Toward Qualty of Servce Faclty D Naptupulu, R Rahm, D Abdullah et al. - Evaluaton of User Satsfacton n Publc Resdental Housng - A Case Study n the Outskrts of Naples, Italy Fabana Forte and Yvonne Russo Vew the artcle onlne for updates and enhancements. Ths content was downloaded from IP address on 05/12/2018 at 13:21

2 Internatonal Conference on Sustanable Energy Engneerng Natural Aggregaton Approach based Home Energy Manage System wth User Satsfacton Modellng F J Luo 1,2, G Ranz 1, Z Y Dong 3, J Murata 4 1 School of Cvl Engneerng, Unversty of Sydney, Sydney, NSW 2006, Australa 2 State Key Laboratory of Power Transmsson Equpment & System Securty and New Technology, Chongqng Unversty, Chna 3 School of Electrcal Engneerng and Telecommuncatons, Unversty of New South Wales, Sydney, NSW 2052, Australa 4 Department of Electrcal Engneerng, Faculty of Informaton Scence and Electrcal Engneerng, Kyushu Unversty, Fukuoka , Japan Emal: tracyluofengj@gmal.com; ganluca.ranz@sydney.edu.au; joe.dong@sydney.edu.au; murata@cg.ees.kyushu-u.ac.jp; Abstract. Wth the prevalence of advanced sensng and two-way communcaton technologes, Home Energy Management System (HEMS) has attracted lots of attentons n recent years. Ths paper proposes a HEMS that optmally schedules the controllable Resdental Energy Resources (RERs) n a Tme-of-Use (TOU) prcng and hgh solar power penetrated envronment. The HEMS ams to mnmze the overall operatonal cost of the home, and the user s satsfactons and requrements on the operaton of dfferent household applances are modelled and consdered n the HEMS. Further, a new bologcal self-aggregaton ntellgence based optmzaton technque prevously proposed by the authors,.e., Natural Aggregaton Algorthm (NAA), s appled to solve the proposed HEMS optmzaton model. Smulatons are conducted to valdate the proposed method. 1. Introducton Drven by the developments of two-way communcaton nfrastructure and buldng automaton technologes, Home Energy Management System (HEMS) has attracted lots of attentons n both academa and energy ndustry. HEMS s consdered as a knd of delegaton of end users to cope wth the complexty of varyng electrcty tarff and automatcally manage the home energy resources. Many works have been conducted on the desgn and development of HEMS. In our prevous works [1-3], we desgned HEMSs for managng the resdental ar condtonng systems; [4] optmally schedules the applances under the forecasted real-tme electrcty prcngs; [5] coordnately scheduled the Battery Energy Storage System (BESS) and household applances wth hgh solar penetratons; n [6], a load commtment framework was proposed to mnmze the household operaton costs; n [7], a HEMS was desgned to schedule the applances n each dwellng unt, and based on t the demand of whole communty was forecasted and reported to the utlty. In [8], a two-stage (day-ahead and realtme stages) HEMS was desgned to manage the resdental BESS wth solar penetratons. In our recent work [9], we dscussed the vson of household recommender systems based on servce recommendaton technques. Exstng HEMSs schedule and control the household applances by usng the user pre-specfed allowable operaton tme ranges [4-6] as constrants. However, ths method gnores the fact that n practcal stuatons, the user often has dfferent tolerances on the operaton tme of dfferent applances. For example, for a rce cooker, the user could strctly restrct ts task termnaton to occur before 6pm, Content from ths work may be used under the terms of the Creatve Commons Attrbuton 3.0 lcence. Any further dstrbuton of ths work must mantan attrbuton to the author(s) and the ttle of the work, journal ctaton and DOI. Publshed under lcence by Ltd 1

3 Internatonal Conference on Sustanable Energy Engneerng as the user does not want to have dnner too late. In ths paper, we refer such requrement as the rgd restrcton. For some applances such as washng machne, the user could have a preferred operaton tme range (e.g., 10am to 8pm), but the user could also tolerate runnng t out of the preferred operaton tme range f a consderable amount of electrcty cost can be saved. In ths paper, we refer such requrement as the elastc restrcton. The major contrbuton of ths paper s to propose a new HEMS by consderng above dfferent knds of operatonal requrements of the user on the household applances. The proposed HEMS ams to coordnately schedule the resdental energy resources (a resdental BESS and multple controllable applances) to accommodate the resdental photovoltac solar output and mnmze the 1-day energy cost, whle respectng the user s satsfactons and lfe habt. Further, a metaheurstc optmzaton technque recently proposed by the authors,.e., Natural Aggregaton Algorthm (NAA), s appled to solve the proposed optmzaton model. Ths paper s organzed as follows. Secton 2 formulated the modellng method for the user s psychologcal satsfacton on the applance operaton; Secton 3 presents the HEMS model formulaton; Secton 4 presents the solvng approach for the proposed HEMS model; Secton 5 dscusses the smulaton results; conclusons and future works are drawn n Secton User Satsfacton Modellng As dscussed n the ntroducton part, the user s requrements on the applance s operaton can be categorzed as rgd restrcton and elastc restrcton, respectvely. For the applances wth the rgd restrcton, they must be restrcted to be operated wthn the user-specfed preferred tme ranges; for the applances wth the elastc restrcton, they are allowed to be operated out of the user s preferred tme ranges. But n ths case, the psychologcal satsfacton of the user on the applance operaton needs to be modelled. In [4], the authors assumed the user always preferred to have the applance fnsh ts work as earlest as possble, and measured the dsturbance of an applance schedule to the user by the delay tme rate (DTR) metrc, n whch the later the applance fnshes ts work, the larger of the DTR value. However, ths strategy n [4] may not truly reflect a practcal user s wllngness. To explan ths, smply magng a user who prefers to have the clothes dryer run between the perod of 9am to 5pm (after he/she leaves home to work n the mornng and backs home after work), then the clothes dryer fnshes the work on 11am or 2pm would probably has no dfference to the user. What actually dsturbs the user s to run the clothes dryer out of the user s preferred tme range (e.g., run from 4pm to 5:30pm). Based on ths realzaton, n ths paper we propose the metrc of User Dsturbance Value (UDV) to measure the satsfacton degree of an applance schedule to the user. The smaller value of UDV, the more the user satsfes wth the applance schedule. The UDV of the operaton of th controllable applance s calculated as: UDV T s () t t t t 1 (1) d begn 1 f t l or t l 0 f l t l end t begn end (2) where T s the total number of tme ntervals over the whole schedulng horzon; t s the duraton of d each tme nterval (hour); s the duraton of a full operaton cycle of the th controllable applance begn l end l (hour); and are the begnnng and end tme of the user s preferred operaton tme ranges of s () the th applance; t s the state of the th controllable applance (0-OFF, 1-ON). The meanng of models (1) and (2) s that f a controllable applance s scheduled wthn the user s preferred operaton tme range, then t s consdered that the user s fully satsfed wth the schedule; otherwse, the psychologcal dsturbance degree of the applance schedule to the user s proporton to the operaton tme outsde the user s preferred tme range. Models (1) and (2) wll be appled on the applances wth the elastc restrctons, and wll be ncorporated nto the HEMS model presented n the next secton. 2

4 Internatonal Conference on Sustanable Energy Engneerng Tarff Sgnals Resdental BESS Model Optmal BESS Schedule Solar Output Predcton NAA Solver Evoluton Module Must-Run Home Load Predcton User Satsfacton Evaluaton Ftness Evaluaton Optmal Applance Schedules Controllable Applances Rgd Restrcted Applance Models Elastc Restrcted Applance Models Fgure 1. Schematc of the proposed HEMS 3. HEMS Model Formulaton In ths paper, we study a smart home envronment wth roof photovoltac source nstalled. The Tmeof-Use (TOU) tarff s consdered. A resdental BESS s consdered to accommodate the ntermttent solar power output. The resdental energy resources (RERs) consdered n ths study nclude the BESS and multple controllable applances. The schematc of the smart home can be llustrated n Fgure 1, and the HEMS model s formulated as below. N mn? F C UDV (3) cost 1 Ccos t Ctarff Cbess Cpv (4) where N represents the number of controllable applances; represents the overall home energy cost; s the penalty factor of the user s dssatsfacton. C s calculated as equaton (1). cost Ctarff conssts of three tems: electrcty tarff cost ( ), BESS deprecaton cost ( bess ), and the dscounted C pv C roof solar panel nstallaton cost ( ). tarff s calculated as: T t 1 where pr() t s the electrcty prce at tme t ($/kwh); tme t, calculated as: Ccost net Ctarff pr( t) t max( P ( t),0) (5) N 1 net P ( t) ndcates the net load of the house at net bess pv P ( t) s( t) P L( t) P ( t) P ( t) (6) P where s the rated power of the th controllable applance (kw); Lt () s the overall power bess P () t consumpton of the non-controllable home energy facltes at tme t (kw); s the pv P () t chargng/dschargng power of the BESS at tme t (negatve-dschargng, postve-chargng); s the forecasted solar power at tme t (kw). The operatonal cost of the BESS s calculated as: C P bess () t t (7) bess bess bess bess E ( t+1) E ( t)+ t P ( t) (8) 3

5 Internatonal Conference on Sustanable Energy Engneerng where s the cost factor of the BESS ($/kwh); E bess () t s the energy stored n the BESS at tme t (kwh). The daly dscounted photovoltac source nvestment cost s calculated by ts nstallaton fee pad by the user and the guarantee years provded by the vendor: Cnstall C pv GY (9) N 365 Cnstall GY where denotes the nstallaton fee of the solar panel ($); N s the number of the guaranteed year of the solar panel. The decson varables of the HEMS model nclude the commtments of the controllable applances bess s () ( t P () t ) and the chargng/dschargng power of the BESS ( ). And the model s subjected to followng operaton constrants: (a) Resdental BESS operaton constrants. In ths study, we assume that power output of the resdental BESS s only allowed to serve the local load, and not allowed to nject nto the man grd. Ths restrcton s based on the secure and relable operaton consderatons of the grd, and s appled n some countres such as Japan. The resdental BESS operaton s therefore subjected to followng constrants: lower SOC upper SOC bess rbess P ( t) P t 1: T (10) P bess ( t) P net ( t) f P bess ( t) 0 and P net ( t) 0 (11) lower upper SOC SOC( t) SOC (12) bess rbess SOC( t) E ( t) E (13) nt SOC( T) SOC (14) nt SOC where and are allowable lower and upper SOC lmts of the BESS; s the ntal SOC of the BESS at the begnnng of the schedulng horzon; SOC() t s the state-of-charge of rbess P rbess the BESS at tme t; s the rated power capacty of the BESS (kw); E s the rated energy capacty of the BESS (kwh). The meanng of constrant (14) s to ensure the BESS have suffcent energy at the end of the schedulng horzon, so as to contnuously provde servces for the house n the ncomng day. (b) Operaton cycle constrant of the controllable applance. The tme duraton of the operaton cycle of the controllable applances must be ensured: T ( s( t) t) d 1: N (15) t1 (c) Operaton constrant of the non-nterruptble controllable applances. Controllable applances n a smart home can be categores as non-nterruptble applance (NIAs) and nterruptble applances (IAs). The operaton of NIAs cannot be nterrupted untl t fnshes ts work: * t * t d/ t NIA s ( t) 1 1: N (16) * t t where represents the tme nterval when the th NIA s frst tme to be turned on; number of NIAs. (d) Mnmum onlne/offlne tme constrants of IAs. For the IAs, the mnmum onlne and offlne tme constrant s appled to protect ther mechancal devces, shown as: ( t) s ( t) 0, 1: N on on IA mn, off off IA ( t) mn, s ( t) 1, 1: N NIA N s the (17) 4

6 Internatonal Conference on Sustanable Energy Engneerng on off () where t () and t IA are accumulated onlne and offlne tme of th IA at tme t (hour); N s the on off mn, mn, number of IAs; and are mnmum onlne and offlne duraton lmts of the th IA (hour). NIA IA There s N N N. (e) Operaton tme constrant of the controllable applances wth rgd restrctons: begn end rg s ( t) 0, f t l or t l, 1: N (18) where rg N s the number of rgd restrcted controllable applances. 4. Solvng Approach The proposed HEMS model (3) s a mx-nteger, combnatoral optmzaton problem over a fnte horzon. In ths paper, a new metaheurstc optmzaton algorthm prevously nvented by the authors,.e., NAA [10], [11], s appled to solve the proposed model. Start RES models Electrcty tarff System model confguratons Intal the populaton Forecasted house load Forecasted solar power Indvdual mgraton Located & generalzed searches Constrant handlng NO Ftness evaluaton Termnaton condton satsfed? YES RES schedule output End Fgure 2. Workflow of the NAA-based solvng approach NAA s a bologcal ntellgence nspred optmzaton algorthm. Specfcally, NAA s motvated by the self-aggregaton ntellgence of the group lvng anmals. In NAA, the whole populaton of the ndvduals are organzed as multple sub-populatons, and a stochastc mgraton model s utlzed to mgrate the ndvduals among the sub-populatons. Meanwhle, the located and generalzed search strateges are used n NAA at each generaton. Due to the page lmtaton, we do not gve detaled ntroducton for NAA n ths paper, whch can be found n [10], [11]. 5

7 Internatonal Conference on Sustanable Energy Engneerng 4.1. Encodng scheme By applyng NAA on the proposed HEMS model, each ndvdual s encoded as a potental RER schedulng soluton, wth the dmenson of ( N 1) T. The frst NT dmensons are bnary varables, representng the commtment status of the N controllable applances, where every T dmensons represent the status of an applance over the T tme ntervals. The last T dmensons are contnuous varables, representng the chargng/dschargng power of the BESS Solvng procedures The overall workflow of the NAA-based HEMS model solvng approach s shown n Fgure 2. Frstly, the system models are nputted, whch nclude the profles of forecasted 1-day solar output and noncontrollable home energy consumptons, and the operatonal models of the RERs. Then, the ndvduals are encoded n NAA and the evoluton process s started. In each generaton, for each ndvdual, the constrant handlng s performed to make t feasble, and the ndvdual s ftness value s evaluated by calculatng equaton (3). Fnally, the schedulng results are outputted. 5. Smulaton study In ths secton, smulatons are conducted to valdate the proposed HEMS model. All programs are mplemented on Matlab and executed on a DELL workstaton wth 128-Gegabyte memory and 2 Intel Xeon processors Smulaton setup We smulate a smart home wth fve knds of controllable household applances and one resdental BESS. The schedulng nterval s set to 15 mnutes, the confguratons of the controllable applances and BESS are shown n tables 1 and 2, respectvely. 1-day Australan solar power data wth 1-mnte samplng frequency s used, shown n Fgure 3. The data s then averaged n 15-mnute bass for smulaton. One-day non-controllable house energy consumpton profle s generated from the Australan Smart Grd, Smart Cty data [12], shown n Fgure 4. The TOU tarff publshed by the Energy Australa s used, shown n table 3. It s assumed that the crtcal peak tme prce s actvated. on off mn, mn, The values of both and are set to 15 mnutes. The parameter settngs of NAA are as S follows: populaton sze s set to be 200, maxmum generaton tme s set to be 600, 8 2, Crlocal 0.8, 1.5, and Crglobal 0.2. Table 1. Resdental BESS Confguratons Power Capacty 3kW Energy Capacty 6kWh SOC Upper Lmt 90% SOC Lower Lmt 10% Intal SOC 30% Table 2. TOU Tarff Structure Tme-of-Use Rate ($/kwh) Peak: 2pm-8pm Shoulder: 7am-2pm, 8pm-10pm Off-peak: 10pm-7am Crtcal Peak Prce Rate ($/kwh) 5pm-8pm N S, Cp 25, 6

8 Internatonal Conference on Sustanable Energy Engneerng Fgure 3. 1-Day solar power output Table 3. Controllable Applance Confguratons Operaton Cycle Interruptble Restrcton Preferred/Requred Operaton Rated Applance Duraton Tme Range Power Dsh NO Rgd [7:30pm, 9am] 0.8kW 2 hours Washer Rce NO Rgd [10:30am, 12am] 0.5kW 0.75 hours Cooker1 Rce NO Rgd [5pm, 6:30pm] 0.5kW 0.75 hours Cooker2 Washng NO Elastc [5pm, 7pm] 1.0kW 1 hour Machne Clothes YES Elastc [7:30pm, 10pm] 1.0kW 1.5 hour Dryer a Note: the superscrpts of rce cooker mean ts frst and second tme runnng. Fgure 4. Profle of 1-day must-run load of the house 5.2. Resdental energy resource schedulng results We set the value of the user satsfacton penalty factor to be 0.4 and run the smulaton. And we consder two cases: n case 1, the cost factor of the BESS ( ) s set to a low value,.e., 0.06$/kWh. Ths means the battery operaton cost s lower than the electrcty prce even n the off-peak perod. In case 2, s set to a hgh value,.e., 0.15$/kWh, whch means the battery operaton cost s hgher than the electrcty prce n off-peak and shoulder perods, but lower n the peak and crtcal peak tme perods. We then nvestgate the RER schedulng results under the two dfferent BESS operatonal cost settngs as below Case 1: Low BESS Cost Factor 7

9 Power (kw) Internatonal Conference on Sustanable Energy Engneerng Fgure 5 shows the fnal schedulng of the fve knds of controllable applances n case 1. The TOU sgnals are also plotted n Fgure 5. It can be clearly seen that most of the applances are well scheduled to avod the crtcal peak perod (5pm-8pm). The two rgd restrcted applances,.e., dsh washer and rce cooker, are scheduled to operate wthn the user-specfed tme ranges. It s also notceable that although the preferred operaton tme range of the user on washng machne s [5pm, 7pm], t s scheduled to operate n the perod of [11:am, 12:00am], so as to avod the peak and crtcal peak prcngs. Fgure 6 shows the optmzaton results of the resdental BESS. The profles of the solar power output and total power load of the house are also plotted. The negatve value of the BESS output ndcates dschargng whle the postve value ndcates chargng. The results clearly show that the BESS chargng/dschargng profle follows a desred shape. That s, durng the perod of [0am, 8am] when there s no solar power, the BESS s dscharged to serve the house load; at the noon tme, the solar power s surplus, and the BESS s therefore charged by the solar power. In the evenng tme, the BESS s dscharged agan to fully serve the house load. Fgure 7 shows the correspondng SOC profle of the BESS under the optmzed schedule. It can be clearly seen that the BESS s well controlled wthn ts safe SOC lower and upper lmts. And at the end of the schedulng horzon, the BESS s charged back to larger than ts ntal SOC level (30%), whch makes t can contnuously serve the house load n the ncomng day. Table 4 summarzes the cost tems and the UDV value under the fnal schedule. Fgure 5. Controllable applance schedulng results of case BESS Output Solar Power House Load am 4am 8am 12am 16pm 20pm Tme Fgure 6. Resdental BESS schedulng results of case 1 8

10 SOC (%) Internatonal Conference on Sustanable Energy Engneerng am 4am 8am 12am 16pm 20pm Tme Fgure 7. SOC profle of the resdental BESS of case 1 Table 4. Cost Items and UDV of Case 1 Item Value Electrcty Cost $0.54 BESS Operaton Cost $0.28 Dscounted Daly Solar Generaton Cost $0.66 Total Cost $1.48 Total UDV Case 2: Hgh BESS Cost Factor The schedulng results of the controllable applances are shown n Fgure 8. The applance schedules follow the smlar trend wth those of case 1. That s, all the applances wth elastc restrctons are properly scheduled to avod peak or crtcal peak perods. The resdental BESS schedulng results are shown n Fgure 9 and Fgure 10. Comparng wth the BESS schedulng results of case 1, t can be clearly seen that n the crtcal peak perod, the BESS s well dscharged to serve the house load. However, after 20pm n the evenng, when the house load s hgh and electrcty prce s low, the BESS s not dscharged to serve the load. Ths s because the cost factor of the BESS n case 2 s hgher than the electrcty prce of the off-peak perod. Table 5 reports the house operaton cost and total UDV of case 2. Wth the same settng of the penalty factor (0.4), the total UDV of both cases 1 and 2 are the same, whch s equal to 1.0. Meanwhle, both the electrcty cost and BESS operaton cost of case 2 are larger than those of case 1. Fgure 8. Controllable applance schedulng results of case 2 9

11 SOC (%) Power (kw) Internatonal Conference on Sustanable Energy Engneerng BESS Output Solar Output House Load am 4am 8am 12am 16pm 20pm Tme Fgure 9. Resdental BESS schedulng results of case Senstvty Study of Penalty Factor Fnally, we conduct the senstvty study for the penalty factor, so as to nvstgate the relatonshp of user satsfacton and the house operaton cost. For both cases, we run the smulatons under dfferent settngs of, and the results are shown n Fgure 11 and Fgure 12. Wth the ncrease of the value of, the UDV decreases, and the house operaton cost ncreases. Therefore, dfferent settngs of represent dfferent compromses of the user satsfcaton and household energy cost am 4am 8am 12am 16pm 20pm Tme Fgure 10. SOC profle of the resdental BESS of case 2 Table 5. Cost Items and UDV of Case 2 Item Value Electrcty Cost $0.75 BESS Operaton Cost $0.81 Dscounted Daly Solar Generaton Cost $0.66 Total Cost $2.22 Total UDV

12 House Operaton Cost ($) House Operaton Cost ($) Internatonal Conference on Sustanable Energy Engneerng penalty factor=3 penalty factor=1 penalty factor=0.6 penalty factor=0.2 2 penalty factor= UDV (a) case 1 (b) case 2 penalty factor= UDV Fgure 11. Senstvty study of penalty factor of case 1 & 2 6. Conclusons and future works In ths paper, we propose a new HEMS wth the consderatons of dfferent operatonal requrements of the user on dfferent knds of applances. The penetratons of resdental photovoltac source and BESS are smulated. And a recently proposed metaheurstc optmzaton algorthm,.e., NAA, s appled to solve the proposed HEMS model. The results clearly show that the controllable applances and BESS can be coordnately scheduled by takng account nto the user satsfacton. The smulatons also show that dfferent operaton cost factors of the resdental BESS wll sgnfcantly affect the decson-makng of the HEMS. Future researches can be conducted n dfferent drectons. The stochastc nature of the renewable energy output can be modelled and ncorporated nto the HEMS. The vehcle-to-home (V2H) technology can be also ntegrated nto the smart home energy management. Whle many exstng HEMS researches assume the renewable energy source s prvate and can only serve the load of a sngle house, how to coordnately perform the energy management for multple unts n a shared renewable energy envronment s also a problem deservng of study. Acknowledgment Ths work s supported n part by the Australan Research Councl through ts Future Fellowshp scheme (FT ), n part by the Vstng Scholarshp of State Key Laboratory of Power Transmsson Equpment & System Securty and New Technology (Chongqng Unversty, Chna) (2007DA ), and n part by the Early Career Research Program of Faculty of Engneerng and Informaton Technology, The Unversty of Sydney, Australa. References [1] Luo F, Zhao J, Dong Z, Tong X, Chen Y, Yang H and Zhang H. Optmal dspatch of ar condtoner loads n southern Chna regon by drect load control, IEEE Transactons on Smart Grd, vol. 7, no. 1, [2] Wang H, Meng K, Dong Z, Xu Z, Luo F and Wong K. Effcent real-tme resdental energy management through MILP based rollng horzon optmzaton, n Proc. IEEE Power and Energy Socety General Meetng, [3] Wang H, Meng K, Luo F, Dong Z, Verbc G, Xu Z and Wong K. Demand response through smart home energy management usng thermal nerta, n Proc. Australasan Unverstes Power Engneerng Conference, [4] Zhao Z, Lee W, Shn Y and Song K. An optmal power schedulng method for demand response n home energy management system, IEEE Transactons on Smart Grd, vol. 4, no. 3, [5] Pedrasa M A, Spooner T and MacGll I. Coordnated schedulng of resdental dstrbuted energy resources to optmze smart home energy servces, IEEE Transactons on Smart Grd, vol. 1, no. 2, penalty factor=3 penalty factor=1 penalty factor=0.6 penalty factor=0.2 11

13 Internatonal Conference on Sustanable Energy Engneerng [6] Rastergar M, Fotuh-Fruzabad M and Amnfar F. Load commtment n a smart home, Appled Energy, vol. 96, [7] Ozturk Y, Senthlkumar D, Kumar S and Lee G. An ntellgent home energy management system to mprove demand response, IEEE Transactons on Smart Grd, vol. 4, no. 2, [8] Iwafune Y, Ikegam T, Oozek T and Ogmoto K, Cooperatve home energy management usng batteres for a photovoltac system consderng the dversty of households, Energy Converson and Management, vol. 96, [9] Luo F, Ranz G, Wang X and Dong Z. Servce recommendaton n smart grd: vson, technologes, and applcatons, n Proc. 9 th Internatonal Conference on Servce Scence, [10] Luo F, Zhao J, and Dong Z. A new metaheurstc algorthm for real-parameter optmzaton: Natural aggregaton algorthm, n Proc. IEEE Congress on Evolutonary Computaton, [11] Luo F, Dong Z, Chen Y and Zhao J., Natural aggregaton algorthm: a new effcent metaheurstc tool for power system optmzatons, n Proc. IEEE Internatonal Conference on Smart Grd Communcatons, Sydney, Nov [12] Smart Grd, Smart Cty [Onlne]. Avalable at: efault.aspx 12