2-1 集成選址與庫存優化 - 創新電動汽車服務網絡設計.pdf

1、集成選址與庫存優化創新電動汽車服務網絡設計Scaling Up Electric-Vehicle Battery Swapping Services in Cities:A Joint Location and Repairable-Inventory Model張玉利 博士北京理工大學管理與經濟學院Joint work with Wei Qi and Ningwei ZhangYuli Zhang2BIT個人介紹張玉利 博士北京理工大學管理與經濟學院副教授、特別研究員、博士生導師2014年于清華大學自動化系獲工學博士學位2011-2012年在加州大學伯克利分校IEOR系分校訪學北京運籌學會副

2、秘書長、理事；中國運籌學會不確定系統分會常務理事國際運籌與管理學會(INFORMS)、國際生產與運營管理學會(POMS)會員中國物流與采購聯合會-采購與供應鏈專家委員會委員工信部工業互聯網產業聯盟-供應鏈特設組副主席Yuli Zhang3BIT主要研究方向復雜系統建模復雜系統建模&優化算法設計優化算法設計數學規劃模型數學規劃模型線性/非線性規劃混合整數規劃隨機/魯棒/動態規劃機器統計模型機器統計模型監督學習模型無監督模型強化學習模型精確優化算法精確優化算法基于梯度的算法基于分支的算法基于分解的算法智能優化算法智能優化算法啟發式算法隨機搜索算法機器學習增強算法系統設計與優化問題系統設計與優化問題

3、需求預測網絡布局庫存控制供需平衡生產計劃運輸物流產線排程資源調度歡迎對運籌優化與供應鏈管理、大數據決策感興趣的同學，聯系攻讀碩士、博士或者開展博士后工作。歡迎對運籌優化與供應鏈管理、大數據決策感興趣的同學，聯系攻讀碩士、博士或者開展博士后工作。Yuli Zhang4BIT主要研究方向復雜系統建模復雜系統建模&優化算法設計優化算法設計混流復雜生產系統混流復雜生產系統調度排程調度排程鋼鐵生產煉鋼-連鑄-熱軋聯合調度問題主持：科技創新2030“新一代人工智能”項目子課題復雜制造環境下人機物三元協同決策優化方法合作企業：寶鋼、新鋼一體化供應鏈物流一體化供應鏈物流網絡優化網絡優化面向運營與中斷風險的物流

4、網絡布局與決策主持：國家自然科學基金面上項目-一體化供應鏈彈性物流網絡設計與優化合作企業：京東物流電動汽車服務網絡服電動汽車服務網絡服務運營務運營電動汽車充換電設施網絡規劃與調度主持：國家自然科學基金面上項目-電動汽車光伏充換電站網絡隨機魯棒運營優化合作企業：北汽、奧動新能源Yuli Zhang5BIT戰略層研究：電動汽車充換電系統網絡設計研究城市基礎設施網絡下換電站和充電站設施部署，以滿足快速充換電需求，緩解電網壓力，減少用戶等待時間。1）提出了“本地本地換電，集中換電，集中充電”充電”的基礎設施電池交換服務的配置2）提出一種數據驅動的聯合選址-庫存模型3）開發了一種新的算法框架，結合約束生

5、成約束生成(CG)(CG)和參數搜索參數搜索(PS)(PS)技術Yuli Zhang6BIT戰略層研究：城市內公交車充電設施選址考慮交通網和電網整合的情況下，多階段的電動公交車充電設施選址規劃問題。MP目標：最小化總成本（充電站、充電樁、電網線路建設成本，公交車交通成本、充電耗電成本）約束：Special case:single-stage planning(SP)modelModel extension:Budgeted multistage planning(BMP)model；Multistage planning no waste(MPNW)model；Budgeted multist

6、age planning-no waste(BMPNW)model.SOC轉化線性轉化10,(,)11(1)min*2*11(1)aaaddaddaddjjji ji ji jjmjmjjjimjorimnmnmnmm nfxK Tb htlYlr1)交通網2)電網3)兩網耦合S=+:S,=2 2,=2,0=1+:1S1 11=,=+,0 ,=1,=1,*,1,0,jjlmiijmmjiijm piijmjmijmijmijmsHd WsmWYi j mWi j mWYi j mWi j m,22,2mnmnmnmmnmPQlvlv非線性項Yuli Zhang7BIT運營層研究：用戶行為數據分

7、析根據出行鏈時空矩陣投影模型，預測城市各個空間的車輛時空分布，基于車輛的充電需求畫像，預測車輛充電需求預測發生的時空分布。結合充電樁的布局，建立優化車輛充電調度模型，得到充電車樁匹配推薦結果。7營運類車輛充電需求預測及充電場站匹配推薦Yuli Zhang8BIT運營層研究：基于價格激勵的車輛共享再平衡Yuli Zhang9BIT調度層研究：基于區塊鏈的智能有序充電引導針對充電站服務商單樁收益率低、用戶充電難等問題，提出基于區塊鏈技術的充電平臺，開發有序充電智能合約算法，通過有序充電引導保障用戶和運營商的雙邊利益。9運營商2運營商1政府運營商2運營商1政府區塊鏈系統Yuli Zhang10BIT

8、調度層研究：Vehicle-to-Grid充放電調度Yuli Zhang11BITEV Service Network DesignScaling Up EV Battery Swapping Services in Cities Motivation Operations of Swapping and Charing Stations Model and Algorithms Case Studies&InsightsYuli Zhang12BITElectric Vehicles are BoomingScalingScaling Up EV Battery Swapping Servi

9、ces in CitiesUp EV Battery Swapping Services in CitiesSoure:BNEFEV share of new car sales worldwide:EV share of new car sales worldwide:500M EVs will be on the road by 2040!500M EVs will be on the road by 2040!Yuli Zhang13BITScalingScaling Up EV Battery Swapping Services in CitiesUp EV Battery Swapp

10、ing Services in CitiesElectric Vehicles are BoomingNational Electric Vehicle Supervision and Management Center 新能源汽車產業發展新能源汽車產業發展規劃（規劃（2021202120352035年）年）Yuli Zhang14BITCharging Mode vs Swapping Mode-UsersSwapping ModeSwiftness3-5 minutesCompactnessSame service level,less spaceSafetyProperly charge

11、,test and maintain batteriesScaling Up EV Battery Swapping Services in CitiesCharging ModeRange anxiety30 minutes 58 hoursResale anxietyUsed values of EVs may deteriorate quicklyOther issuesInconvenience,unsafety,battery degradationYuli Zhang15BITCharging Mode vs Swapping Mode-GovernmentScaling Up E

12、V Battery Swapping Services in Cities“碳達峰、碳中和”純電動乘用車平均生命周期碳排放比傳統汽油車降低26%，但存在6年左右減排滯后期（動力蓄電池制造碳排放占50%，10年/15萬公里）換電模式具有更高的動力蓄電池利用率換電模式具有更高的動力蓄電池利用率 換電模式使純電動汽車生命周期碳排放較傳統汽油車降低36%。換電模式將A級純電動汽車減排滯后期降低到2年4年。來源：中汽數據，換電模式加速汽車產業碳中和變革，https:/newenergy.in- Zhang16BITBattery Swapping is RevivingScalingScaling Up

13、 EV Battery Swapping Services in CitiesUp EV Battery Swapping Services in CitiesNIO(蔚來)offers its car owners battery swapping service free of charge,and is expected to deploy 1,100 swapping stations by 2020.BAIC(北汽)plans to invest$1.4 billion in building 3,000 swapping stations by 2022.Motivation:Pr

14、oject Optimus Prime,BAIC BJEV-Aulton（奧動新能源）-BTC,10000 EVsYuli Zhang17BITBattery Swapping is RevivingScalingScaling Up EV Battery Swapping Services in CitiesUp EV Battery Swapping Services in CitiesMotivation:Project Optimus Prime,BAIC BJEV-Aulton-BTC,10000 EVsYuli Zhang18BITChallenges of Scaling Up

15、Battery SwappingScaling Up EV Battery Swapping Services in CitiesMismatching between Transportation network and Power gridSwapping demandsEnergy supplyYuli Zhang19BITHow to Address Phase II ChallengesScaling Up EV Battery Swapping Services in Cities Swap locally,charge centrally Locally swapped at d

16、ecentralized swapping stations(DSS)Transported to and charged at centralized charging stations(CCS)Strategies of the State Grid Corporation of China Currently piloted in cities such as HangzhouYuli Zhang20BITSetup(Configuration-macro level)Scaling Up EV Battery Swapping Services in CitiesCharging st

17、ations are colocatedwith grid nodes with sufficient capacityEach swapping station independently follows an policy(,)r QSwapping demand:a general renewal process with given mean and varianceYuli Zhang21BITScaling Up EV Battery Swapping Services in CitiesBattery deficitProportion of unfulfilled orders

18、 is less than Battery stock of charging station satisfies()()()EIiiiD tntn tProviding batteries to before timeitFully charged batteries from till time Model the counting process of“battery deficit at a charging station with respect to one swapping station iitProb()hCiiD tRCRSetup(Configuration-micro

19、 level)Yuli Zhang22BITScalingScaling Up EV Battery Swapping Services in CitiesUp EV Battery Swapping Services in CitiesSetup(Configuration-micro level)Battery expenditure:()=Battery income:()=:the transportation time:the charging time:=+,the effective charging timek(t):the number of orders that the

20、swapping station has placed before tBattery deficit:=:the number of customers in the queueYuli Zhang23BITOur ObjectiveScaling Up EV Battery Swapping Services in CitiesWe address the phase-II challenges by studying the Swap locally,charge centrally system:Analysis and Models:Analytical results for a

21、special two echelon service network DSS:Battery swapping with(r,Q)policies CCS:Battery charging with R initial fully charged batteries.Non-convex mixed-integer programming model The optimal charging station location Joint optimization of(r,Q)and RSolution algorithm Submodularity-based constraint gen

22、eration algorithm Parametric search algorithm for the non-convex sub-problemInsights:Cost and environmental implications as swap demand scales upYuli Zhang24BITRelated LiteratureScaling Up EV Battery Swapping Services in CitiesMulti-Echelon Inventory for Repairable Items:Sherbrook(1968,OR)Grave(1985

23、,MS)Lee(1987,IIE)Gerard(2001,OR)Closed-Loop Supply Chain Management:Savaskan et al.(2004 MS)Abbey et al.(2014 POM)Calmon et al.(2020 MSOM)Metric methodExact computation method based on numerical integrationNo closed-form formula Moment matching methodClosed-form formula based onRestrictive assumptio

24、ns,e.g.,symmetric batch size and transportation timeProblem settings:Reverse flow channel;Consumer perceptions;Large time-scaleMethodology:Game theory;Empirical method;Dynamic programmingNo location-inventory analysis Yuli Zhang25BITRelated LiteratureScaling Up EV Battery Swapping Services in Cities

25、Battery swapping and phase-I challenges:Mak et al.(2013,MS)Avci et al.(2015,MS)Lim(2015,MSOM)Swap-locally&charge-centrally:Zhang and Wang(2016 IEEE PS)Liu et al.(2019 IEEE SG)Meta-heuristic algorithm to dispatch batteries between DSS and CSSDeterministic mixed-integer linear program to schedule batt

26、eries and inventoryOur models provide closed-form analysis,which preserves tractability while taking into account demand uncertainty and networked operations.Exact algorithm for the non-convex mixed-integer programming problem by exploiting sub-modularity and concavity.Yuli Zhang26BITOutlineScaling

27、Up EV Battery Swapping Services in Cities Motivation Operations of Swapping and Charing Stations Non-passion Swap Demands Operations of a Swapping Station Operations of a Charging Station Model and Algorithms Case Studies&InsightsYuli Zhang27BITNon-Poisson Swap DemandsScaling Up EV Battery Swapping

28、Services in CitiesData:National Electric Vehicle Supervision and Management Center of China,include more than 20.97 million real-time records.Identified 8,333 swaps in Beijing in December 2019.Aggregate Hourly Demand Profile:Yuli Zhang28BITNon-Poisson Swap DemandsScaling Up EV Battery Swapping Servi

29、ces in CitiesFinding:Non-Poisson swap demand arrivalsPeriodNum.of Non-Poisson DaysPeriodNum.of Non-Poisson Days00:00-00:59 212:00-12:59 301:00-01:59 313:00-13:59 202:00-01:59 114:00-14:59 003:00-03:59 215:00-15:59 104:00-04:59 016:00-16:59 205:00-05:59 017:00-17:59 006:00-06:59 118:00-18:59 007:00-0

30、7:59 119:00-19:59 008:00-08:59 020:00-10:59 009:00-09:59 321:00-21:59 110:00-10:59 222:00-22:59 211:00-11:59 223:00-23:59 0Yuli Zhang29BITOperations of a Swapping StationScaling Up EV Battery Swapping Services in CitiesRenewal process counts the EVs arrivals up to time ,Stockout probability is no gr

31、eater than Re-order level satisfiesProb()1STSNTr(),0SNtt tSr11(1)TSTrTT 212TTrTT Result 1.The reorder point of a swapping station is given byand are mean and variance of swap requirements;is the one-way transit time;is the inverse of the cumulative standard normal distribution function;Result 2.The

32、reorder point under the worst-case is given by1()2(1)2SS 2TTYuli Zhang30BITOperations of a Charging StationScaling Up EV Battery Swapping Services in CitiesBattery deficitProportion of unfulfilled orders is less than Battery stock of charging station satisfies()()()EIiiiD tntn tProviding batteries t

33、o before timeitFully charged batteries from till time Model the counting process of“battery deficit at a charging station with respect to one swapping station iitProb()hCiiD tRCRYuli Zhang31BITOperations of a Charging Station-DeterministicScaling Up EV Battery Swapping Services in CitiesMean of Batt

34、ery DeficitVariance of Battery DeficitThe inventory position of an inventory system following the(r,Q)policy is uniformly distributed over r+1,r+Q in the steady state when the discrete-valued demandforms a renewal process(Sivazlian 1974).Yuli Zhang32BITOperations of a Charging Station-DeterministicS

35、caling Up EV Battery Swapping Services in CitiesLemma 1.Ifis uniformly distributed over ,thenmod(,)Z Q0,1,1QLemma 2.If ,then()0ZQLemma 3.The variance of is bounded asLemma 4.X22Var()Var()2 D tXWVWW22WWVWVar()Var()Var()Var()Var()XWD tXW2221,if 0,()6(),if.QQQQQ 2233()61 Yuli Zhang33BITOperations of a

36、Charging Station-DeterministicScaling Up EV Battery Swapping Services in CitiesResult 3.The mean of the battery deficit at a charging station incurred by a swapping station is given byResult 4.The piecewise expression also approximatesVar()D t222(),if ,Var()1f.,i 06QQD tQQ ()QMain result:We analytic

37、ally characterize the mean and the variance of the battery deficit by Lemma 1-Lemma 4Yuli Zhang34BITOperations of a Charging Station-DeterministicScaling Up EV Battery Swapping Services in CitiesResult 5.The number of batteries to stock at a charging station needs to satisfyhResult 7.The stock level

38、 for a swapping station with on-site charging needs to satisfy111BCBCrTT Battery stock level at a charging station:Benchmark:on-site charging at decentralized swapping stationsYuli Zhang35BITOperations of a Charging Station-RandomScaling Up EV Battery Swapping Services in Cities :the number of custo

39、mers in the queueRandom discrete(DR)service timeYuli Zhang36BITOperations of a Charging Station-RandomScaling Up EV Battery Swapping Services in Cities :the number of customers in the queueRandom discrete(DR)service timeYuli Zhang37BITOperations of a Charging Station-RandomScaling Up EV Battery Swap

40、ping Services in CitiesFor each sub-queue lFor the considered queue Yuli Zhang38BITOperations of a Charging Station-RandomScaling Up EV Battery Swapping Services in CitiesFor the considered queue Yuli Zhang39BITOperations of a Charging Station-RandomScaling Up EV Battery Swapping Services in CitiesF

41、or the considered queue deterministic effective charging timerandom effective charging timeYuli Zhang40BITOutlineScaling Up EV Battery Swapping Services in Cities Motivation Operations of Swapping and Charing Stations Model and Algorithms Integrated Location Model Constraint Generation Algorithm Par

42、ameter Search Algorithm for Subproblem Case Studies&InsightsYuli Zhang41BITIntegrated Location Model Scaling Up EV Battery Swapping Services in Cities2(,)()()()TihtitihSCBtihhihihhihihhihihSCBihihihhihihihhihihihhihihiihcyC z y Qcc zcr yRQycc zyc QQQ Swapping station depreciation costCharging statio

43、n depreciation costBattery depreciation costcost of trucking batteriesNon-convex,piecewisewhether or not build a charging stationwhether or not charging station serves swapping station swapping station reorder quantity from charging station 0 1hz，hh0 1ihy，i0ihQ sih Decision variables:Yuli Zhang42BIT

44、Integrated Location Model Scaling Up EV Battery Swapping Services in Cities,)(P)min(,)s.t.,1,(1)(11)(1)(1,0,0,.z y QihhihhihhihihC z y QyzihyiyzihQQyihabcd min,(,)=+()s.t.(1)(1).()=min0,()+whereLocation and inventory decisionsMulti-variable non-convex concave optimizationYuli Zhang43BITAlgorithm Sum

45、maryScaling Up EV Battery Swapping Services in Cities,.(MP)min s.t.(1)(1),Ckhhihihhhz y whhihkhi hi hhic zywadwgyh To solve the original non-convex,piecewise problem:Constraint Generation Module:We iteratively solve a relaxed mixed 0-1 master problem(P)()min()+:0 ,=0,.Parameter Search Module:We solv

46、e the following subproblem by iteratively solving a parameterized problem.Yuli Zhang44BITAlgorithm SummaryScalingScaling Up EV Battery Swapping Services in CitiesUp EV Battery Swapping Services in CitiesB&B:Branch-and-bound;B&C:Branch-and-cut;BD:Benders decomposition;B&P:Branch-and-pricing;CG:Constr

47、aint generation;CQMIP:Conic quadratic mixed integer program;CQP:Conic quadratic program;EPI:Extended polymatroid inequality;LA:Lagrangian relaxation;MIP:Mixed integer program;PS:Parametric search;UFLP:Uncapacitated facility location problemYuli Zhang45BITScaling Up EV Battery Swapping Services in Ci

48、ties()min()+:0 ,=0,+()+()Constraint Generation AlgorithmYuli Zhang46BITConstraint Generation AlgorithmScaling Up EV Battery Swapping Services in CitiesFor any charging station ,is a submodular function over .h()hg|0,1Lemma 5.For any ,the submodular function satisfies(P)()min()+:0 ,=0,Let be set of p

49、ermutations of elements in ,be the j-th element in permutation ,and be the position of element in jS ii10()(),()0.jjjgg Sg Sg S|0,1yg()max.iiig yg y12|,.()iiig yg yyyy(i)We can construct g(y)by adding up the marginal cost of including swapping stations following a sequence(ii)If we knew y,we could e

50、asily find the optimal sequence by sortingYuli Zhang47BITConstraint Generation AlgorithmScaling Up EV Battery Swapping Services in Cities(P)can be further reformulated as a mixed 0-1 linear program(MIP)min,+s.t.(1)(1),CG:Iteratively solve the following relaxedmaster problem with a subset of constrai

51、nts:(MP)min,+s.t.(1)(1),.,CGSolve(,)Solve MP,obtain,and Inputstop gap CGGenerate by sorting NoOutput,and YesYuli Zhang48BITParameter Search Algorithm for SubproblemScaling Up EV Battery Swapping Services in CitiesConstraint generation algorithm requires solving the subproblem (P)()min()+:0 ,=0,.This

52、 is a multi-variable non-convex problem,not easy to solveYuli Zhang,Zuo-Jun Max Shen,Shiji Song,Exact Algorithms for Distributionally Robust Machine Scheduling with Uncertain Processing Times.INFORMS Journal on Computing,2018,30(4):662-676.Yuli Zhang,ZuoJun Max Shen,Shiji Song,Parametric Search for

53、the Bi-attribute Concave Shortest Path Problem,Transportation Research Part B:Methodological,2016,94:150168.Parametric Search Algorithm for discrete concave minimizationYuli Zhang49BITParameter Search Algorithm for SubproblemScaling Up EV Battery Swapping Services in CitiesThis parameterized problem

54、 is a single-variable convex problem,which is easy to solve(P)SFor given ,can be reformulated to()()()+S(P)Constraint generation algorithm requires solving the subproblem (P)()min()+:0 ,=0,.This is a multi-variable non-convex problem,not easy to solveYuli Zhang50BIT(ii)Given any ,let be the intersec

55、tion point of lines .Let then,any satisfiesParameter Search Algorithm for SubproblemScaling Up EV Battery Swapping Services in CitiesWe can efficiently search for the optimal using a branch and bound procedure based on the following proposition.(P)S*Q*1 2()u Q*Q*(P)(P)S120(,)u v:()(1,2),jjjjlvuuvj12

56、1122,min,guvuvuv 12,Result 6.(i)Let be the optimal solution to and ,then any optimal solution to is also optimal12,uvg(,)(,)f z yf zy(,)zy(,)zy(,)zyuvslope:slope:conv(H)slope:Yuli Zhang51BITParameter Search Algorithm for SubproblemScaling Up EV Battery Swapping Services in Cities,(P)min():0,.iiiiiii

57、bQaQQQiSQ 12,22,(P)min:0,6,(P)min():,iiiiiiiiiiiiiiiibQaQQQbaQQQQ Closed-form solution existsThe remaining task is to solve the easier problem(P)Finally,we also develop routines for algorithm initialization and speedupYuli Zhang52BITParameter Search Algorithm for SubproblemScaling Up EV Battery Swap

58、ping Services in CitiesLet ,and be the optimal solution to ,and .Closed-form solutions()iQ,1()iQ,2()iQ,(P)i1,(P)i2,(P)iSolving(P)Yuli Zhang53BITParameter Search Algorithm for SubproblemScaling Up EV Battery Swapping Services in CitiesLet ,and be the optimal solution to ,and ,thenLemma 6.,(i),where i

59、s unique nonnegative real root of the cubic equation ;(ii);(iii)if or ,where ,otherwise,;(iv),and are non-increasing in .()iQ,1()iQ,2()iQ,(P)i1,(P)i2,(P)i0iS,1()min,iiiQQiQ3230iQaQb,2()min max/(),iiiiQbaQ()iiQQ/iiba,1i,123iiiiba,2()()iiQQ()iQ,1()iQ,2()iQSolving(P)Yuli Zhang54BITParameter Search Algo

60、rithm for SubproblemScaling Up EV Battery Swapping Services in CitiesLemma 7.For any,let be the unique nonnegative real root of equation ,theniSInitializationip32(1/6(3)6iipp bap min,min/,12()12().SpeedupConsider ,let for ,and be the optimal solution and the optimal parameter for(1,2,|)1,2,kSkkkQkPk

61、SResult 7.,(i);and(ii).1,|1k 1kkiiQQ1kkYuli Zhang55BITOutlineScaling Up EV Battery Swapping Services in Cities Motivation Operations of Swapping and Charging Stations Model and Algorithms Case Studies&Insights Calibration Computational Efficiency Managerial Insights:Scalability,FlexibilityYuli Zhang

62、56BITCalibrationScaling Up EV Battery Swapping Services in CitiesParameters about Battery41kWh capacity battery,weighted 330kg7,062.15$with a eight-year lifespan20%remaining SOC of swapped-off batteriesStandard 42kW direct-current fast chargerDelivered by 10-ton deadweight truck 1.13$/kmWe calibrate

63、 the model with real data of EV battery swaps from NEVSMCempirical setting in Beijing consultation with practitioners from BAIC BJEV,Aulton and manufacturerSwapping RequirementsPeriodMeanStandard DeviationMinMax0:00-8:595.683.710229:00-23:5914.513.90328Data From National Electric Vehicle Supervision

64、 and Management Center of ChinaYuli Zhang57BITCalibrationScaling Up EV Battery Swapping Services in CitiesDifferent stations Cost ComponentsEquipmentUnit CostQuantityLifeCharging StationPlant33.90$/Year/m2500 m2-Worker9,604.52$/Year/Person5-Charger(42kW)1,129.94$20015Battery Conveyer14,124.29$110Tra

65、nsformer(2500kVA)35,310.73$230Power Grid Expansion Costs11,299.44$-10Construction Project21,186.44$-10Swapping StationPlant33.90$/Year/m2100 m2-Worker9,604.52$/Year/Person2-Swapping Operators112,994.35$215Construction Project14,124.29$-10Swapping Station with ChargersPlant33.90$/Year/m2300 m2-Worker

66、9,604.52$/Year/Person4-Charger(42kW)1,129.94$3015Transformer(630kVA)11,299.43$130Swapping Operators112,994.35$215Power Grid Expansion Costs11,299.44$-10Construction Project21,186.44$-10Yuli Zhang58BITComputational EfficiencyScaling Up EV Battery Swapping Services in CitiesEffectiveness of the parame

67、ter search algorithm forSince is non-convex,we force to ensure that Gurobi can solve it,the same operation is used in following parts.All solving methods stop gaps are .For each problem size(30,60,90)we solve 100 instances.22()16,0.QQQ()Q710Yuli Zhang59BITComputational EfficiencyScaling Up EV Batter

68、y Swapping Services in CitiesEffectiveness of the CG&PS algorithmCG&PS:the proposed methodGurobi:Gurobi 9.1.0 solverBP:Teo&Shu(2005,OR),Ni et al.(2021,JOC)cannot solve root problem within 1800 secondsHeuristic:Shen et al.(2011,JOC)Force()=2+(21)/6,0 to ensure that Gurobi can solve model.Yuli Zhang60

69、BITManagerial Insights-ScalabilityScaling Up EV Battery Swapping Services in Cities12040608010001234567MainBenchmarkBenchmark,SlowMain,60%-Q1204060801000510152025303540MainBenchmarkBenchmark,SlowMain,60%-QMain:Swap-locally,charge-centrally.Benchmark:On-site charging.Observation 1.Centralized chargin

70、g is LESS scalable than decentralized on-site charging(if all else being equal).Yuli Zhang61BITManagerial Insights-ScalabilityScaling Up EV Battery Swapping Services in CitiesObservation 1.Centralized charging is LESS scalable than decentralized on-site charging(if all else being equal).This is cont

71、rary to the wisdom in traditional supply chains in which pooling demands is economicalReason:The favorable pooling effect is dominated by two unfavorable effects with centralized charging:Order batching effect Transportation lead time effect Yuli Zhang62BITManagerial Insights-ScalabilityScaling Up E

72、V Battery Swapping Services in Cities12040608010001234567MainBenchmarkBenchmark,SlowMain,60%-Q1204060801000510152025303540MainBenchmarkBenchmark,SlowMain,60%-QSlow AC charging:7kW vs.Standard DC charging:42kWObservation 2.Centralized charging is MORE scalable than decentralized on-site charging(if o

73、nly slow charging permits).Yuli Zhang63BITFlexibilitiesScaling Up EV Battery Swapping Services in Cities01020304050600246810-80%-40%0+40%+80%0246Slow AC ChargeOptimalBaseline DC ChargeObservation 3.Centralized charging allows remarkable flexibilities in:-abating battery quantities-adjusting charging

74、 station deployment.Yuli Zhang64BITFlexibilitiesScaling Up EV Battery Swapping Services in CitiesObservation 4.Order splitting is only advantageous when the demand is scaled up by reducing fixed ordering cost.Yuli Zhang65BITSummaryScaling Up EV Battery Swapping Services in CitiesWe study swap-locall

75、y,charge-centrally service networks to address the phase-II infrastructural challenges Developed models of battery stocking,charging and circulating operationsProposed an effective algorithm for solving this special joint location-inventory optimization problemManagerial insights:Scalability depends on grid capacity Flexibilities in operations and design