Managng the Mcrogrd Net Load Varablty Abdulazz Alanaz, Mohana Alanaz, Amn Khodae Department of Electrcal and Computer Engneerng Unversty of Denver Denver, CO USA Alanaz.Alanaz@du.edu, Mohana.Alanaz@du.edu, Amn.Khodae@du.edu Abstract Ths paper nvestgates and compares two optons to control the mcrogrd net load varablty resulted from hgh penetraton renewable generaton. The proposed optons nclude 1) Local management, whch lmts the mcrogrd net load varablty n the dstrbuton level by enforcng a cap constrant, and 2) Central management, whch recommends on buldng a new fast response generaton unt to lmt aggregated mcrogrd net load varablty n the dstrbuton level. In ths paper, the power flow changes between two consecutve hours,.e., power rampng s consdered as the major representatve of varablty. A mcrogrd optmal schedulng model s developed usng mxednteger programmng. Numercal smulatons demonstrate the effectveness of the proposed approach n dentfyng the more vable opton. Index Terms mcrogrd net load varablty, mcrogrd optmal schedulng, renewable generaton. NOMENCLATURE Indces: d Index for day Index for dspatchable unts t Index for tme arameters: C Generaton cost coeffcent CSD Shut down cost CSU Startup cost D Load demand FC Fxed O&M cost GC Gas power plant capacty H Number of hours k Varablty cap MD Mnmum down tme MF Number of hours the unt can be OFF MN Number of hours the unt can be ON MU Mnmum up tme NL No load cost OCC Overnght captal cost Maxmum generaton capacty mn Mnmum generaton capacty B ayback perod M Capacty of transmsson lne between the utlty and the mcrogrd RD Ramp down rate RU Ramp up rate VC Varable O&M cost W Wnd power Market prce Varables: I Commtment state of the dspatchable unts DER generated power M Mcrogrd net load sd Number of successve OFF hours su Number of successve ON hours TC Total operaton cost y Startup ndcator z Shut down ndcator I. INTRODUCTION Mcrogrds are small-scale power systems whch consst of at least one dstrbuted energy resource (DER) and one load that are connected to the man dstrbuton grd. The mcrogrd s an autonomous system; so t can sland tself from the utlty grd durng outage events and reconnect tself when the dsturbance s removed. The slandng capablty makes the mcrogrd an mportant technologcal development n modern power systems as t can consderably ncrease the power system reslence and relablty [1] [4]. Moreover, mcrogrds facltate the control and operaton of a large number of DERs by utlzng a local controller. Renewable energy resources, such as wnd and solar, can also be effcently ntegrated to the power system va mcrogrds. Renewable generaton, and n partcular wnd energy, s rapdly growng n power systems, prmarly due to the fallng cost of the technology and strct envronmental mandates. In 212, around 283.2 GW of total wnd energy resources were nstalled worldwde, whch s expected to reach 416.4 GW by the end of 215 [5] [11]. The wnd generaton varablty, however, has presented a sgnfcant challenge n ensurng a relable supply-demand balance when utlzng ths technology. In [11], a study of ntegratng renewable generaton wthn a mcrogrd s conducted. The study proposes operatonal controls to help wth the renewable ntegraton and managng the renewable generaton varablty. In [12], t s dscussed that to ensure power system stablty and supply relablty, the sudden wnd power varablty must be compensated by fast response generaton unts, such as natural gas or hydro. As the penetraton of wnd generaton ncreases n the mcrogrd and there s a hgh mcrogrd penetraton n the utlty grd, the wnd power varablty may
cause a severe negatve mpact. Studes conducted by the Calforna ISO suggests that the rapd changes n renewable generaton can cause sgnfcant challenges n supply-demand balance, result n over-generaton rsk especally n nghttme hours when the power demand s low, and requre ncreased flexblty n the system n terms of fast rampng (see Fg. 1) [13]. be controlled by addng operaton constrants to the optmal schedulng problem dependng on the unt type such as generaton lmts, mnmum on/off tme lmts, thermal lmts, and rampng rate lmts. Nondspatchable unts are typcally renewable energy resources such as wnd turbnes and solar photovoltac whch cannot be controlled by the mcrogrd due to the uncontrollable nature of the prmary source of energy. Mcrogrd optmal schedulng problem Opton 1 Addng the varablty cap Cost before ncludng varablty cap (TC1) Opton 2 Buldng a new gas generaton Fg. 1. The Calforna ISO duck curve [13] A relable coordnaton of renewable generaton wthn the mcrogrds requres a vable mcrogrd schedulng model. The mcrogrd optmal schedulng problem determnes the leastcost schedule of local loads and DERs as well as the transferred power whle consderng prevalng operatonal constrants. The mcrogrd optmal schedulng problem and ts formulaton can be found n [14] [17]. Ths paper bulds upon the avalable studes n the lterature to develop a mcrogrd optmal schedulng model that ncorporates mcrogrd net load varablty lmts. Ths model, furthermore, wll be used to analyze the local management opton for lmtng mcrogrd net load varablty. The soluton wll be compared wth the central varablty management opton of nstallng a centralzed power plant from an economc perspectve. The levelzed cost of energy (LCOE) wll be moreover used as an alternatve measure to ensure that the decson s made correctly. LCOE s a convenent measure that ntegrates the captal cost, fuel costs, fxed and varable operatons and mantenance (O&M) costs, and fnancng costs to obtan one fxed number representng the energy cost of any specfc generaton type [18]. The rest of ths paper s organzed as follows. The proposed model s outlned n Secton and formulated n Secton III. Numercal smulatons are presented n Secton IV and the paper s concluded n Secton V. II. MODEL OUTLINE A. Mcrogrd Components The mcrogrd components that are modeled n the proposed mcrogrd optmal schedulng problem nclude local generaton unts and loads. The local generaton unts can be ether dspatchable or nondspatchable. Dspatchable unts can Cost after ncludng varablty cap (TC2) Cost of addng varablty cap LCOE of addng varablty cap TC2- TC1 Whch one s smaller? Cost of buldng a new gas generaton Fg. 2. roposed mcrogrd net load varablty-lmtng model LCOE of gas generaton The optmal soluton of the mcrogrd net load varablty B. Mcrogrd net load varablty management model Fg. 2 depcts the flowchart of the proposed model. The man objectve of ths model s to fnd the optmal soluton to lmt the mcrogrd net load varablty between two consecutve hours (.e., a rampng constrant). The model conssts of an optmal schedulng problem and two cost calculaton problems. The optmal schedulng problem determnes the unts schedule, the utlty transferred power wth the mcrogrd, and the total operaton cost of the mcrogrd before addng the mcrogrd net load varablty constrant. In the local management opton, a varablty constrant (.e., a cap) wll be added to the problem to restrct the net load varablty between any two consecutve hours. A
new utlty transferred power flow wll be compared wth the old one and the mpact of addng the constrant s observed. A new total operaton cost wll be obtaned. When the mcrogrd net load varablty s forced to be small between two consecutve hours, the total operaton cost wll be ncreased dependng. The dfference between the new and the old operaton costs s calculated to fnd the cost of addng the cap. In the central management opton, a new fast response generaton unt (here a gas unt) s consdered to be bult to deal wth the aggregated mcrogrd net load varablty n the dstrbuton level. The plannng cost of buldng the new unt s calculated and annualzed. After calculatng the cost of both optons, a comparson between them wll be conducted to fnd the more economcal soluton. Alternatvely, the LCOE of each opton wll be calculated n order to enable further comparson. The opton that has the smallest LCOE s consdered to be the optmal soluton of lmtng the mcrogrd net load varablty. III. MODEL FORMULATION A. Mcrogrd optmal schedulng problem formulaton The mcrogrd optmal schedulng problem s modeled by mxed-nteger programmng. The objectve of the optmal schedulng problem s to mnmze the total operaton cost of the mcrogrd (1) subject to operatonal constrants (2)-(8). The frst term n the objectve represents the generaton cost of the dspatchable unts, no-load cost, and startup and shut down costs. The second term s the cost of purchasng power from the utlty grd. The mcrogrd net load (also known as the transferred utlty power) s the transferred power from or to the mcrogrd through the pont of common couplng (CC). The transferred power cannot exceed the capacty of the transmsson lne connectng the utlty grd to the mcrogrd as modeled n (2). The mcrogrd net load mght be postve (.e., mcrogrd mports power from the utlty where the transferred power s less expensve than local generaton). On the other hand, when the mcrogrd net load s negatve, mcrogrd delvers power to the utlty grd snce the local generaton s less expensve than the transferred power. The power balance equaton (3) guarantees that the summaton of local generaton and transferred power equals the hourly mcrogrd net load. The nondspatchable unt generaton (here the wnd generaton) s represented as a negatve load n (3). The mcrogrd components are modeled n (4)-(8). The mum and mnmum generaton capacty lmts for each dspatchable unt are modeled by (4). The rampng up and down rate lmts between two consecutve hours are represented by (5)-(6). The mnmum number of successve hours that the unt can be up or down s shown by (7)-(8). The commtment state of a dspatchable unt, the startup state and the shut down state are bnary varables. The commtment state I wll be one when the unt s ON, otherwse t s zero. The startup ndcator y s one when the unt s started up, otherwse t s zero. The shut down ndcator z wll be one when the unt s shut down, otherwse t s zero. mn [C ( ) NL I CSU y CSD z ] t d ρ M, (1) M M, M (2) M D W, (3) mn I I (4) (t1 )d RU (5) (t1 )d RD su sd MU z (t1)d MD y (t1)d The startup and shut down ndcators are determned as n (9)-(1). The startup and shut down counters are modeled as n (11)-(14). I I,t1,d y z (9) y z 1 (1) su MN I (11) (MN 1 )I MN su su,t1,d 1 (12) sd MF (1 I ) (13) 1 (MF 1)I sd sd,t1,d 1 (14) B. Addng varablty cap The local management opton adds a varablty cap to the mcrogrd net load,.e., the power transferred wth the utlty grd. The varablty cap s modeled n ths paper for the nterhour varablty (15) and the nter-day varablty (16). M, M,( t 1) d k t 1, d (6) (7) (8) (15) M,1d M,24( d1) k 1 (16) The optmal schedulng problem wll be used agan to fnd the optmal schedulng of mcrogrd unts after addng the varablty lmt constrants (15) and (16). A new mcrogrd unts schedule and a new total operaton cost (TC2) wll be obtaned. The cost of the local management opton can be found by calculatng the cost ncrease after addng the varablty cap as n (17). Opton 1: Cost TC2 TC1 (17) The varablty cap cost ($/yr) wll be levelzed to obtan the LCOE n $/MWh for the cap value. The LCOE of the varablty cap wll be compared wth the LCOE of gas generaton for makng the decson on optmal soluton.
C. Buldng a new gas generaton Buldng a new gas generaton s another opton to deal wth the ncreasng varablty n the mcrogrd net load. The cost of buldng a new gas power generaton s dvded nto captal and operaton and mantenance (O&M) costs. The operaton cost s also dvded nto fxed O&M cost and varable O&M cost. The cost of the central management opton can be calculated as n (18). GC*OCC B Opton 2: Cost GC*FC GC*VC*H (18) The LCOE for gas generaton s determned n order to compare t wth the LCOE for the addng varablty cap opton. IV. NUMERICAL SIMULATIONS The proposed mcrogrd net load varablty-lmtng model s appled to a test mcrogrd wth four dspatchable unts and one nondspatchable unt (wnd turbne). The characterstc of generatng unts and nondspatchable unt are gven n Table I. One-year tme horzon of forecasted wnd, load and market prce s used n the studes. Mxed nteger programmng s used to model and solve the mcrogrd optmal schedulng problem. The followng cases are studed: Case 1: Addng a varablty cap (local management opton) Case 2: Buldng a new gas generaton (central management opton) Case 1: Addng a varablty cap s the frst opton to lmt the mcrogrd net load varablty. The solved optmal schedulng problem s used as a base case to determne the total operaton cost before lmtng the mcrogrd net load varablty. Dfferent values of varablty cap are added as a constrant to the optmal schedulng problem. The values of varablty cap are rangng from 32 to 14 MW, as the mum power ramp between two consecutve hours s 32 MW. The mpact of addng varablty cap on the total operaton cost s shown n Table II for each reducton value of the varablty cap. Fgs. 3 and 4 show the cost curve and the LCOE curve of each reducton value of the varablty cap, respectvely. TABLE I: CHARACTERISTIC OF GENERATING UNITS (D: DISATCHABLE, ND: NONDISATCHABLE) Unt Type Ramp Cost Mn.-Max. Mn. Up/Down Coeffcent Capacty Up/Down Rate ($/MWh) (MW) Tme (h) (MW/h) G1 D 27.7 4-1 3 5 G2 D 39.1 4-1 3 5 G3 D 61.3 2-6 1 3 G4 D 65.6 2-6 1 3 G5 ND -4.16 - - Cost ($) Fg. 3. The cost ($) of each reducton value of the varablty cap LCOE ($/MWh) 1 9 8 7 6 5 4 3 2 1 2.5 2 1.5 1.5 1 2 3 4 5 6 7 8 9 1 11 12 13 14 15 16 17 18 Reducton value of varablty cap (MW) Fg. 4. The LCOE ($/MWh) of each reducton value of the varablty cap TABLE II: THE IMACT OF ADDING VARIABILITY CA ON THE TOTAL OERATION COST The reducton value of the varablty cap (MW) Reducton value of varablty cap (MW) 1 2 3 4 5 6 7 8 9 1 11 12 13 14 15 16 17 18 The Total Operaton Cost ($/yr) Varablty Cap Impact ($/yr) Increased percentage of the total cost (%) 3,298,764.81.. 1 3,2988,28.28 63.47.2 2 3,298,,951.71 186.9.6 3 3,299,157.72 392.91.12 4 3,299,522.56 757.75.23 5 3,3,12.17 1,355.36.41 6 3,3,79.51 1,944.7.59 7 3,31,531.67 2,766.86.84 8 3,32,98.66 4,143.86.126 9 3,34,56.97 5,292.16.16 1 3,35,888.4 7,123.24.216 11 3,39,6.74 1,241.93.31 12 3,311,828.3 13,63.22.396 13 3,317,997.43 19,232.62.583 14 3,325,759.41 26,994.6.818 15 3,335,443.35 36,678.54 1.112 16 3,346,11.91 47,346.1 1.435 17 3,363,98.34 65,143.54 1.975 18 3,,377,189.12 78,424.31 2.377 Case 2: The second opton s buldng a new gas generaton unt n the dstrbuton network to address the mcrogrd net load varablty. The capacty of the gas generaton unt should be equal to the varablty cap value. The annualzed cost of buldng a 1MW gas generaton, whch s only for 1 MW/h varablty cap, s around $8,/yr. So, the cost of
buldng a new gas generaton s sgnfcantly greater than the cost of addng a 1 MW varablty cap. Smlarly, for the rest of the varablty cap values, addng varablty cap s more economcal than buldng a new gas generaton unt. Another measure (.e., the LCOE) s used to decde the more economcally vable opton. The average LCOE of gas generaton n the Unted States s $66.3/MWh [18]. Fg. 5 depcts the LCOE for each varablty cap along wth the LCOE of gas generaton. It s obvous that the gas LCOE s much greater than the LCOE of all varablty caps. So, addng a varablty cap s always a more vable decson than buldng a new gas generaton unt. LCOE ($/MWh) 7 6 5 4 3 2 1 1 2 3 4 5 6 7 8 9 1 11 12 13 14 15 16 17 18 Reducton value of varablty cap (MW) Fg. 5. The LCOE of both reducton value of the varablty cap and gas generaton V. CONCLUSION Cap LCOE Gas LCOF An effcent model for lmtng the mcrogrd net load varablty was proposed. Two optons were consdered, were the opton nvestgates the addton of a varablty cap to lmt the mcrogrd net load varablty wthn two successve hours and the second opton nvestgated the addton of a new gas generaton unt to the dstrbuton system. The mpact of addng the cap on the total operaton cost n the frst opton was notced by comparng the mcrogrd total operaton cost n both cases (.e., the orgnal soluton and the soluton after addng the varablty cap). The dfference was consdered to be the cost of addng varablty cap. The cost of buld a new gas generaton and the LCOE of gas generaton were further calculated for comparson purposes. The model was tested and analyzed on a mcrogrd test system. The numercal smulatons were shown that addng a varablty cap on the mcrogrd net load was always the more economcal soluton for addressng the mcrogrd net load varablty. REFERENCES [1] Department of Energy Offce of Electrcty Delvery and Energy Relablty, Summary Report : 212 DOE Mcrogrd Workshop, 212. [Onlne]. Avalable: http://energy.gov/stes/prod/fles/212 Mcrogrd Workshop Report 91212.pdf. [Accessed: 13-Nov-214]. [2] S. arhz, H. Lotf, A. Khodae, and S. Bahramrad, State of the Art n Research on Mcrogrds: A Revew, IEEE Access, vol. 3, 215. [3] M. Shahdehpour, Role of smart mcrogrd n a perfect power system, n IEEE ES General Meetng, 21. [4] S. Bahramrad, A. Khodae, J. Svachula, and J. R. Aguero, Buldng Reslent Integrated Grds: One neghborhood at a tme, IEEE Electrf. Mag., vol. 3, no. 1, pp. 48 55, Mar. 215. [5] GWEC Global Wnd Report Annual market update 213 - GWEC- Global-Wnd-Report_9-Aprl-214.pdf. [Onlne]. Avalable: http://www.gwec.net/wp-content/uploads/214/4/gwec-global- Wnd-Report_9-Aprl-214.pdf. [Accessed: 5-Aug-215]. [6] H. Lu, H.-Q. Tan, C. Chen, and Y. L, A hybrd statstcal method to predct wnd speed and wnd power, Renew. Energy, vol. 35, no. 8, pp. 1857 1861, Aug. 21. [7] X. Wang,. Guo, and X. Huang, A Revew of Wnd ower Forecastng Models, Energy roceda, vol. 12, pp. 77 778, Jan. 211. [8] W. Zhang, J. Wu, J. Wang, W. Zhao, and L. Shen, erformance analyss of four modfed approaches for wnd speed forecastng, Appl. Energy, vol. 99, pp. 324 333, Nov. 212. [9] N. Hatzargyrou, H. Asano, R. Iravan, and C. Marnay, Mcrogrds, ower Energy Mag. IEEE, vol. 5, no. 4, pp. 78 94, 27. [1] E. Sortomme and M. A. El-Sharkaw, Optmal power flow for a system of mcrogrds wth controllable loads and battery storage, n ower Systems Conference and Exposton, 29. SCE 9. IEEE/ES, 29, pp. 1 5. [11] K. Strunz, E. Abbas, and D. N. Huu, DC Mcrogrd for Wnd and Solar ower Integraton, IEEE J. Emerg. Sel. Top. ower Electron., vol. 2, no. 1, pp. 115 126, Mar. 214. [12] C. Wu, H. Mohsenan-Rad, J. Huang, and A. Y. Wang, Demand sde management for wnd power ntegraton n mcrogrd usng dynamc potental game theory, n GLOBECOM Workshops (GC Wkshps), 211 IEEE, 211, pp. 1199 124. [13] FlexbleResourcesHelpRenewables_FastFacts.pdf. [Onlne]. Avalable: http://www.caso.com/documents/flexbleresourceshelprenewables_ FastFacts.pdf. [Accessed: 8-May-215]. [14] T. Logenthran and D. Srnvasan, Formulaton of unt commtment (UC) problems and analyss of avalable methodologes used for solvng the problems, n Sustanable Energy Technologes (ICSET), 21 IEEE Internatonal Conference on, 21, pp. 1 6. [15] A. Khodae, Mcrogrd Optmal Schedulng Wth Mult-erod Islandng Constrants, IEEE Trans. ower Syst., vol. 29, no. 3, pp. 1383 1392, May 214. [16] A. Khodae, rovsonal Mcrogrds, IEEE Trans. Smart Grd, vol. 6, no. 3, pp. 117 1115, May 215. [17] A. Khodae, Reslency-Orented Mcrogrd Optmal Schedulng, IEEE Trans. Smart Grd, vol. 5, no. 4, pp. 1584 1591, Jul. 214. [18] Levelzed cost and levelzed avoded cost of new generaton resources n the Annual Energy Outlook 214 - electrcty_generaton.pdf. [Onlne]. Avalable: http://www.ea.gov/forecasts/aeo/pdf/electrcty_generaton.pdf. [Accessed: 15-May-215].