logoGlobal Energy Interconnection




      Global Energy Interconnection

      Volume 3, Issue 5, Oct 2020, Pages 453-463

      Aggregator-based demand response mechanism for electric vehicles participating in peak regulation in valley time of receiving-end power grid

      Chen Fang1,2 ,Xiaojin Zhao3 ,Qin Xu1 ,Donghan Feng3 ,Haojing Wang1 ,Yun Zhou3
      ( 1.Electric Power Research Institute,State Grid Shanghai Municipal Electric Power Company,Shanghai 200437,P.R.China , 2.East China Electric Power Test and Research Institute Co.,Ltd,Shanghai 200437,P.R.China , 3.Department of Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,P.R.China )


      With the increase in the power receiving proportion and an insufficient peak regulation capacity of the local units,the receiving-end power grid struggles to achieve peak regulation in valley time.To solve this problem while considering the potential of the large-scale charge load of electric vehicles (EVs),an aggregator-based demand response (DR) mechanism for EVs that are participating in the peak regulation in valley time is proposed in this study.In this aggregator-based DR mechanism,the profits for the power grid’s operation and the participation willingness of the EV owners are considered.Based on the characteristics of the EV charging process and the day-ahead unit generation scheduling,a rolling unit commitment model with the DR is established to maximize the social welfare.In addition,to improve the efficiency of the optimization problem solving process and to achieve communication between the independent system operator (ISO) and the aggregators,the clustering algorithm is utilized to extract typical EV charging patterns.Finally,the feasibility and benefits of the aggregator-based DR mechanism for saving the costs and reducing the peak-valley difference of the receiving-end power grid are verified through case studies.

      1 Introduction

      In China,electricity transmission project from the west to the east is being developed at a high speed,and the power receiving proportion of the East China Grid is increasing each year [1].Taking Shanghai City in east China as an example,the proportion of the external power to the total electricity consumption in 2018 has already exceeded 50%.During July and August of the same year,this percentage can even reach approximately 60%.As the main receivingend power grid,the East China Grid is confronted with the difficulty of peak regulation in valley time during an insufficient local power generation.In addition,inflexible external power does not participate in the peak regulation during the actual operation.Therefore,the pressure of the peak regulation in valley time of the receiving-end power grid further increases,which severely affects its economy and reliability.

      The demand response (DR) is an effective method to conduct valley filling for the receiving-end power grid [2].The DR can motivate customers to reduce or shift their electricity consumption through financial incentives [3].One of the most potential DR participator is the electric vehicle (EV) as the impact of EV charging on a large scale cannot be underestimated [4].Around 61% of EVs are privately owned,and they are used for commuting [5].These owners usually charge their EVs after work,and they are unplugged before they leave home [6].As a result,the EV plug-in duration largely overlaps with the valley time.If the EVs participate in peak regulation,valley filling can be achieved and the load fluctuation,typically caused by disordered charging,can be avoided.

      The EV charging control methods have been widely studied.Sortomme et al.[7]coordinated the charging of the EVs to minimize the losses in the distribution network.Although centralized control can fully coordinate the grid operation and EV charging,it is difficult to be implemented because of ignoring owners’ willingness.Therefore,a priced-based DR is proposed to encourage EV owners to change their charging behaviors actively and fill the valley [8].However,time-based and price-based control strategies run the risk of causing another load peak during the middle of the night [9].To implement the perfect valley-filling charging strategy,the price of electricity and the noncooperative game theory were used in [10]to guide the decentralized EV resources.However,it is impossible for the independent system operator (ISO) to collect the optimal charging strategies of all the EVs and differences in the individual charging times,which are not considered.

      In this context,the involvement of the aggregators,whose function is aggregating the EV charging resources and trading with the ISO as intermediaries,is particularly important [11-15].As the mediator between the EVs and the power system,the aggregator plays a significant role in the peak regulation [16-20].In a study by Papadopoulos et al.[16],a three-layer multi-aggregator system was used to coordinate the EV charging and it satisfied the EV owners’ preferences under normal and emergency situations.When considering the network operation constraints,a demand reduction service was provided and EV charging was optimized via the aggregators in [17].While ensuring the efficient operation of the distribution network,the communication between the different aggregators and the privacy of the EV information was also investigated by Karfopoulos et al.[18].To integrate the uncertainty of photovoltaic output,an energy management strategy for the aggregators to control the EV charging and discharging power was presented by Hu et al.[19].To minimize the load variance of the Duck Curve,an efficient and optimal charging algorithm to compute iterative process between the EVs and the aggregators is proposed by Le Floch et al.[20].

      As previously mentioned,for the receiving-end power grid,increasing the external power and the decline of the local power generation can lead to an insufficient local peak regulation capacity in valley time.Aiming to analyze the feasibility and economic potential for the EVs that are participating in the peak regulation in valley time,in this study,an aggregator-based DR mechanism for the EVs to participate in the peak regulation in valley time of the receiving-end power grid is proposed.Unlike the aggregators that are described in the other references that directly control the charging,the aggregators in this DR mechanism will send invitations to the owners before taking control of the EVs to guarantee the owners’ decisionmaking priority over the EV charging.To tackle the mismatch of the EV charging period (e.g.,charging period may across 0:00 AM) and the optimal dispatch period of the day-ahead unit commitment (i.e.,0:00 AM-24:00 PM),a rolling optimization process is included.Therefore,the DR mechanism not only aggregates the EV charging resources with full consideration of the owners’ willingness but also schedules the unit generation reasonably for the power grid by rolling the optimization process.

      The rest of the paper is organized as follows.The details of the aggregator-based DR mechanism for the EVs that provides the peak regulation in valley time of the receivingend power grid are presented in Section 2.The rolling unit commitment model considering EVs’ participating in DR is established in Section 3.The case studies to verify the effectiveness of the proposed mechanism are included in Section 4.Finally,Section 5 concludes the paper.

      2 Aggregator-based DR mechanism for EVs providing peak regulation in valley time

      2.1 Framework of aggregator-based DR mechanism for EVs

      The main participants in this aggregator-based DR mechanism for the EVs include the ISO,the aggregators,and the EVs,in which the aggregators are usually car companies or charging pile operators.The capacity of the individual EV is too insufficient to take part in the day-ahead market [21].Compared with the EVs’ direct transactions with the ISO,the necessity of the aggregators has the following reasons.(1) It aggregates distributed EV charging resources into sufficient capacity to participate in the DR.(2) It collects information and predicts the EV charging load more accurately using historical data.(3) It improves the efficiency and the convenience of the DR programs because the EV owners have no need to operate themselves and it is conducive to unified management for the ISO.(4) It respects the willingness of the EV owners.(5) They have specific online technology platforms and matched smart charging piles for their users.

      The structure of the aggregator-based DR mechanism for the EVs is shown in Fig.1.In most studies,the aggregators are considered to have direct control over the EVs during the whole optimization period.In fact,the aggregators do not have this permission due to privacy and security issues.This is why it is important to send DR invitations to the EV owners before the EVs participate in the DR.If an EV agrees to participate in the DR program and allows the aggregator to control the charging process,it is a flexible EV load.Otherwise,it is an inflexible EV load,for which the aggregator is just an electricity provider [22].

      Fig.1 Structure of aggregator-based DR mechanism for EVs

      The flow chart of the aggregator-based DR mechanism for the EVs is shown in Fig.2.The detailed steps of this DR mechanism are as follows.

      Step 1:The ISO sends DR invitations to aggregators a few days or a few hours in advance,which includes the DR periods (e.g.,from 23:00 PM to 07:00 AM) and the feedback deadline of invitation (tddl).

      Step 2:The aggregator circularizes the DR invitations to the managed EVs.If the EV owners decide to participate in the DR program,they need to upload the desired state of charge (SOC) at the time of departure to the responsible aggregators.

      Step 3:The aggregators predict and update the flexible EV load and the inflexible EV load according to the EV owners’ decisions on the participation and their historical charging data until the feedback deadline.

      Fig.2 Flow chart of aggregator-based DR mechanism for EVs

      Step 4:The aggregators send the total EV load prediction to the ISO.

      Step 5:Based on the EV load and the non-EV load predictions,the ISO schedules the unit generation and the DR resources.The ISO sends generation scheduling to the power plants and the DR resource dispatch to the aggregators.

      Step 6:The aggregators receive the DR resource dispatch and notify the EV owners in advance.

      Step 7:The aggregators are allowed to dispatch and control the EV charging process within the DR time.

      Through the process from Step 1 to Step 7,the aggregator-based DR mechanism for the EVs can be implemented for real applications.

      2.2 Clustering of typical EV charging patterns

      During Step 3 in the above DR mechanism,the DR participation decisions for a large number of EVs are obtained by the aggregator.However,submitting the participation information of every EV to the ISO is not realistic,and accurately predicting the charging load of every EV is difficult.Furthermore,when considering the complexity of the optimization problem,multitudinous conditions generate multitudinous constraints,which dramatically decreases the solution efficiency.As a mature method to tackle the curse of dimensionality,a relatively small quantity of EV charging patterns (e.g.,“typical EV charging patterns”) can be extracted from the original ones by the clustering algorithm [23].Subsequently,the typical EV charging patterns can be utilized in Step 3 to estimate the flexible EV load and the inflexible EV load of all the EVs that are managed by the aggregators.

      Each EV charging pattern is described by a vector (t p lug-in,tp lug-out,Ech),which consists of the plug-in time tplug-in,the plug-out time tplug-out,and the charging energy needed Ech (determined by the EV user’s desired SOC,initial SOC,and battery capacity).This vector can be predicted by the aggregators based on the historical charging data.With the clustering procedure,the predicted EV charging patterns that share the most similar property are classified into the same cluster,which are represented by the typical EV charging pattern of this cluster.Each typical EV charging pattern is described by a vector (t p lug-in,t p lug-out,E ch,v),where v is the number of EVs in this cluster.To distinguish the different aggregators and one aggregator’s different clusters,the clustered EV charging patterns are presented as (t p lug-in, s,π,t p lug-out, s,π,Ec h, s,π,vs,π),where the subscript s and π represent the π-th cluster of the typical EV charging pattern that is managed by the s-th aggregator.

      Equation (1) represents the relationship between the total EV number NEV and the number of EVs in each cluster vs,π.NA is the number of aggregators (s =1,2,...,NA) and Πs is the number of clusters that are managed by the s-th aggregator (π=Π1,2,...,s).

      Several clustering algorithms are widely applied in many studies,such as the K-means clustering algorithm,the Gaussian mixture algorithm,and the mean shift algorithm.Technically,all these clustering algorithms could be applied for generating typical EV charging patterns.Since the definite number of clusters is unknown,clustering algorithms that do not require presupposed cluster numbers are more convincing and practical for aggregators to acquire typical EV charging patterns in this study.The mean shift algorithm is adopted as the clustering algorithm because of its applicability to large amounts of data and the characteristic of no pre-set cluster number [24].

      The clustering procedure for Step 3 in the DR mechanism that is proposed in Section 2.1 replaces numerous individual EV charging patterns with typical EV charging patterns while retaining the main characteristics.Fewer charging patterns mean fewer variables in the optimization model,which can significantly decrease the complexity of the optimization problem.Predicting the general charging pattern of a cluster is less inaccurate than predicting every EV charging pattern in this cluster.Therefore,aggregators only need to send typical EV charging patterns to the ISO rather than the detailed information of every EV,which simplifies the communication between the aggregators and the ISO.

      2.3 Rolling optimization process for power system dispatching while considering aggregator -based DR mechanism for EVs

      Day-ahead unit commitment is one of the most important parts of the power system dispatching,whose main objective is to determine the operating states and the power generation of units under a certain load range.Since the charging of the EVs often lasts from night to the next morning,it will affect the twoday unit generation scheduling.To simulate realistically the impact of EV charging on the power grid scheduling for two consecutive days,the rolling optimization process is utilized in this study [25].The rolling optimization process is a control technology.The basic concept for this process is that an optimal control problem in a forward optimization window can be solved during each step,but only the first few intervals of the control sequences are implemented (called implemented intervals).Fig.3 depicts the timeline of the rolling optimization process for power system dispatching while considering the aggregator-based DR mechanism for the EVs.In Fig.3,a 72-h window is chosen as an optimization window to fully demonstrate the effect of the EVs on valley filling and the length of the interval is 15 min.The implemented intervals are 24 h long and the rest are unimplemented intervals.Since the entire EV charging period for the last 12 h of the 72-h window may span two optimization windows,the DR is not considered in the last 12 h in the 72-h optimization window.When it comes to the DR feedback deadline,the ISO obtains the latest prediction information of the EV load and the non-EV load.Subsequently,the rolling optimization process starts to work.The detailed steps are as follows.

      (1) Measure the initial information of the variable at the beginning of the optimization window and predict the other information of the remaining intervals.Prediction data are available for the following day that the EVs and the aggregators have the DR contracts on.For the rest of the periods in the whole window,the previous data would be used as prediction data.

      (2) The ISO schedules the unit generation and the DR resources through a unit commitment model within the optimization window,which is shown by the dotted rectangle in Fig.3.In this 72-h window,only the charging schedules for those EVs that will plug in within the implemented intervals will be executed by the aggregators in the following day.

      (3) The power system keeps updating its status and is waiting to move to the next optimization window.Repeat (a) and (b).

      3 Unit commitment model while considering aggregator-based demand response

      As a mature method,the operation of an electric system is modeled as a unit commitment problem.The unit commitment problem can be formulated as a mixedinteger linear programming (MILP) problem with multiple constraints,and it can be effectively solved by commercial MILP solvers.To quantify the value of the EV’s participation in the peak regulation in valley time of the receiving-end power grid,a unit commitment model that considers the aggregator-based DR that is supplied by the EVs is established with the objective of maximizing the social welfare,or in other words,minimizes the total cost of the generating units in the system.Based on the characteristic of EV charging,the familiar unit commitment model is expanded as a rolling unit commitment model with the DR in this section [26].This is shown in subsections (1)-(4),where subsection (1) is the objective function of the model,and subsections (2) - (4) are the different categories of the constraints.

      Fig.3 Timeline of the rolling optimization process

      (1) Objective function

      The objective function is shown below.

      The objective function in (2) is to minimize the total cost of the generating units in the system,which consists of the fuel costs,startup costs,and the shutdown costs.The cost of the external power is not considered in this study.In Equation (2):NG is the number of units;T is the number of time intervals in the rolling optimization window for the peak regulation;cpit,,cuit,,and cdit, are the generation cost,startup cost,and the shutdown cost of unit i at time t,respectively.The details of cpit,,cuit,,and cdit, are shown in Equations (3),(4),and (5) as follows.

      Where pit, is the power generation of unit i at time t;uit, is the binary variable that is equal to 1/0 if unit i is online/offline at time t;CAi,CBi,and CCi are the fuel cost parameters of unit i;yit,/zit, is the binary variable that is equal to 1 if unit i is starting up/shutting down at time t;and CStarti/CShuti is the startup cost/shutdown cost of unit i for one time.Equation (3) can be rewritten as a linear constraint through the piecewise linearization method [27].Based on the objective function in Equations (2) - (5),the relevant linear constraints are introduced in the rest of this section.These include the commitment constraints in Equations (6) - (12),the system constraints in Equations (13) - (14),as well as the DR constraints in Equations (1) and (15) - (17).(2) Commitment constraints

      Equation (6) ensures that yit, and zit, have correct values when ui,t changes.The startup and shutdown of one unit are restricted from occurring simultaneously in Equation (7).Equations (8) and (9) are the minimum online and offline time constraints,respectively,where Ton,i/Toff,i is the minimum online/offline time of unit i.The upper and lower bound limits of the power generation pi,t are shown in Equation (10).Equations (11) and (12) represent the unit ramping constraints,where Pup,i/Pdown,i is the maximum ramp-up/ramp-down rate of unit i.

      (3) System constraints

      Equation (13) represents the power balance of the system,where PD,t contains the predicted non-EV load and the inflexible EV load;PEV,t is the flexible EV load;and,Ploss,t is the network losses at time t.Equation (14) represents the spinning reserve requirement,where Rt is the spinning reserve capability at time t.

      (4) Aggregator-based demand response constraints

      There are two alternative approaches to model the flexible EV charging load for optimization [22].The global approach describes the EV charging process based on the total values of the EV variables while the divided approach is based on the individual information of each EV.It is indicated that the divided approach is more economical and robust than the global approach [28].In addition,the clustering of the typical EV charging patterns in Section 2 can reduce the presence of high variability in the divided approach.Therefore,the divided approach is adopted to establish the DR model for the flexible EV load, and the corresponding constraints are shown in Equations (1) and (15)-(17).

      is the set of the typical EV charging patterns that belong to the flexible EV load.In accordance with the definitions in Section 2.2,(,)s π represents the π-th cluster of the EV charging pattern that is managed by the s-th aggregator.Equation (15) ensures that the energy requirements of the predicted typical EV charging patterns are satisfied during the plug-in periods.The plug-in period Ts,π is determined by the plug-in time and the plug-out time of pattern (,)s π,η is the charging efficiency,pst, ,π is the total charging power of pattern (,)s π at time t,and Δt is the time step.The upper and lower bound limits of the charging power pst, ,π are shown in Equation (16).Equation (17) aggregates the flexible EV load into PEV,t.The obtained optimal charging powerwill be allocated to every EV in this cluster based on energy buffer factor consensus algorithm proposed by Liu et al.[29].

      4 Case study and results

      The proposed aggregator-based DR mechanism for the EVs was tested on a typical receiving-end power grid in Shanghai Power Gird.Fig.4 shows the historical daily load curve and the external power curve for five consecutive summer days (including the weekdays and the weekend) in Shanghai.It can be observed that there is a load valley from 0:00 AM to 08:00 AM and the peak-valley difference of the daily load is more than 10,000 MW.However,external power has a small fluctuation throughout the day.It accounts for 50-65% of the load in valley time,which leads to a huge pressure of peak regulation at that time depending solely on the local units in Shanghai.In the following paper,the external power has been deducted from the baseload to clearly display the situation of the receiving-end power grid.

      Fig.4 Daily load curve and external power curve of Shanghai

      Table1 Classification of generating units that are considered in Shanghai power grid in the case study

      Unit type Units' number Peak regulation participation Coal-fired units 40 Participation Gas-fired units 25 Participation Heating units 9 Non-participation

      In the case study,the generating units that were considered in the Shanghai Power Grid are presented in Table1,which include 40 coal-fired units,25 gas-fired units,and nine heating units.The coal-fired units can participate in the peak regulation through the basic peak regulation,the deeper peak regulation,and the short-time startup and shutdown regulation [30],which may be costly.Due to the fast startup and shutdown characteristic,the gasfired units usually participate in the peak regulation in the peak time.In addition,the uninterruptible heating demand limits the heating units’ capacities in the peak regulation.Thus,there is an insufficiency of the local peak regulation capacity in valley time.

      Due to the Shanghai government’s promotion of the EV subsidy policies,the quantity of the EVs in Shanghai has an obvious advantage over the other regions.By February 2019,the quantity of the EVs in Shanghai had reached 250,000.In the case study,it is assumed that only domestic charging events (i.e.,slow charging mode) have a potential flexible EV load.An inflexible EV load means that the owner will not or cannot participate in the DR programs and will charge the EV as soon as he or she gets home (i.e.,disordered charging mode).For two categories of the EVs,the maximum charging rate is 7 kW and the charging efficiency η is 90%.

      Table2 Case configuration in the case study

      No.Aggregator-based DR considered Percentage of EVs participating in the DR Case 1 Without DR 0%Case 2 With DR 30%

      To verify the effectiveness of the aggregator-based DR mechanism for the EVs,two cases are designed in the case study,as Table2 shows.Case 1 is the rolling unit commitment problem without the DR while Case 2 considers the aggregator-based DR of the EV charging load.Since the capacity of individual EV is usually insufficient to participate in the day-ahead electricity market,the case study does not consider the EVs directly providing DR without the aggregators as a comparison case.For both cases,the total quantity of EVs is 250,000 and the 2017 domestic charge point data in the UK is taken as the historical charging data for the clustering algorithm in this study [31].In Case 2,the participation rate of the aggregator-based DR mechanism is about 30%,i.e.,75,000 EVs.Since the given problem is a MILP problem,it is solved using Gurobi under the GAMS platform.

      After aggregators implemented the charging schedules for the next 24 h several times,the results of the EV charging load and total load of the grid in two cases are shown in Fig.5 and Fig.6 respectively.The timeline in Fig.5 is from day 2 12:00 to day 5 12:00 so that the entire load shifting process can be displayed.In the absence of the DR in Case 1,the disordered charging behavior causes a large number of EVs to be charged during the peak hours from 18:00 PM to 20:00 PM,leading to the sharpness of the peak load.This is not only detrimental to the stable operation of the power grid but also causes frequent changes in the status of the unit and reduces the lifetime of the unit.In Case 2,the EV charging load is shifted from peak time to valley time through the DR,which contributes to reducing the peak-valley difference as Fig.6 shows.It is indicated that the optimal cost solution is nearly equal to providing a valley filling effect.Due to the limited of number of EVs participating in the DR,the valley filling effect in Case 2 is not that obvious.

      Fig.5 The results of the EV charging load curve in two cases

      Fig.6 The results of the power grid load curve in two cases

      Table3 Optimization results of Case 1 and Case 2

      Rolling window No.Startup/shutdown costs (104 ¥)Total costs (104 ¥)Peak-valley ratio (%)1 Case 1 512.54 15,836.27 61.51 Case 2 295.06 15,626.62 58.59 2 Case 1 418.56 15,724.01 58.73 Case 2 283.12 15,578.89 56.58

      The results in two consecutive 72-h rolling optimization windows,day 2 - day 4 and day 3 - day 5 of Case 1 and Case 2,were selected and are presented in Table3.Compared with the results in Case 1,the aggregatorbased DR of the EVs in Case 2 can reduce the total costs of this receiving-end power grid from 15,836.27 × 104 ¥ to 15,626.62 × 104 ¥ in the first rolling window (day 2-day 4),which is about 209.65 × 104 ¥ (1.32%) reduction.In the second rolling window (day 3-day 5),the reduction is 145.12 × 104 ¥ (0.92%).It can be observed from Table3 that there is an obvious gap in the startup/shutdown costs for two cases,which makes the greatest contribution to the cost-saving.This is because the aggregators can control part of the EVs and charge them in valley time,which can avoid frequent startups and shutdowns of some units and eliminate the corresponding costs.Based on the load curve without the external power,the peak-valley ratio decreases from 61.51% to 58.59% in the first rolling window (day 2-day 4) and from 58.73% to 56.58% in the second rolling window (day 3-day 5),which alleviated the pressure of the peak regulation in the valley time to some extent.

      It is predicted that by 2030,the total number of EVs in Shanghai will reach 1.55 million under normal development and 2.45 million under high-speed development [32].On the one hand,large-scale disordered EV charging will further increase the peak load.On the other hand,the potential DR resources will be more abundant.To further illustrate the validity of the aggregator-based DR mechanism for EVs in the future,the rolling unit commitment problem for the receiving-end power grid with different EV quantities is realized.The participation rate of the DR program is still set to 30%.We assume that the grid capacity and the non-EV load will stay the same in these cases,which may not be consistent with the actual situation in the future.

      Fig.7 Load curves of EV participating in peak regulation in valley time on a different scale

      Table4 Effects of EV's number on peak-valley difference

      EV number (104) 0 25 50 100 155 245 Peak-valley difference (MW) 10,483 10,073 9,831 9,792 9,190 9,137 Peak-valley ratio (%) 61.33 58.59 56.84 55.94 51.79 50.10

      Fig.8 Effects of EV's number on total costs of power grid

      The corresponding 24-h load curves and the total costs are shown in Fig.7,Table4,and Fig.8.As shown in Fig.7,the peak load increases as the quantity of the EVs increases because most of the EV charging behaviors are disordered even with the DR,thus worsening the power grid’s operation state during the peak time.Meanwhile,with the growing quantity of the EVs involved with the DR,the capacity for valley filling improves.Even when the peak load increases gradually,the DR mechanism reduces the peak-valley difference and the peak-valley ratio of the receiving-end power grid as demonstrated in Table4.Fig.8 displays the total costs under the two cases in one rolling optimization period and the cost-saving proportion that is brought by the DR.More EVs that are connected to the grid will inevitably increase the total costs of the power grid irrespective of whether there is a DR.The cost-saving proportion that is related to the right-Y axis in Fig.8 is the proportion of the saved costs to the total costs.It can be seen from Fig.8 that the effect of the proposed DR mechanism on the cost savings is gradually significant as the number of EVs increases.In conclusion,the case study with the rising number of EVs further verifies the great potential and the feasibility of the DR mechanism for EVs in the future.

      5 Conclusion

      In this study,an aggregator-based DR mechanism for the EVs participating in the peak regulation in the valley time of the receiving-end power grid is proposed.First,the aggregators predict the flexible and inflexible EV load based on their participation and the historical charging data.Based on the load prediction and the unit characteristics,the ISO schedules the unit generation and the DR resources for the receiving-end power grid.The case studies that were tested on the Shanghai Power Gird verify the feasibility of the proposed aggregator-based DR mechanism and demonstrate the potential of the EV participating in the peak regulation in the valley time on a different scale at different stages of urban development.By observing the case study results,the following conclusions can be obtained.

      (1) The EV participating in the peak regulation in valley time can effectively reduce the total costs and the peakvalley difference ratio of the receiving-end power grid by avoiding frequent startups and shutdowns of the local coalfired units and load shifting,respectively.

      (2) When the DR participation rate remains constant,the DR potential and cost-saving effects of the EV depend mainly on its quantity,which provides a convictive reference for the popularization of EVs in the future.

      This paper emphasizes a feasible framework for the EVs participating in the DR and quantifies the value of the EVs’ participation in the peak regulation in the valley time of the receiving-end power grid.However,the design of the subsidy strategy for the EV owners that are involved with the DR mechanism has not been included in this paper,which needs to be further studied.


      This work was supported by the Science and Technology Project from the State Grid Shanghai Municipal Electric Power Company of China (52094019006U) and the Shanghai Rising-Star Program (18QB1400200).

      Declaration of Competing Interest

      We declare that we have no conflict of interest.


      1. [1]

        Lu J,Yan L,Zhou B et al (2016) Integration of renewable resources in East China Grid.Paper presented at the IEEE Green Energy and Systems Conference (IGSEC 2016),Long Beach,CA,USA,6-7 November 2016 [百度学术]

      2. [2]

        Logenthiran T,Srinivasan D,Tan S (2012) Demand side management in smart grid using heuristic optimization.IEEE Transactions on Smart Grid,3(3):1244-1252 [百度学术]

      3. [3]

        Upadhaya B,Feng D,Zhou Y et al (2019) Optimal decision making model of battery energy storage-assisted electric vehicle charging station considering incentive demand response.Paper presented at the 8th Renewable Power Generation Conference,Shanghai,China,24-25 October 2019 [百度学术]

      4. [4]

        Kumar K N,Tseng K J (2016) Impact of demand response management on chargeability of electric vehicles.Energy,111:190-196 [百度学术]

      5. [5]

        Qian K,Zhou C,Allan M et al (2011) Modeling of load demand due to EV battery charging in distribution systems.IEEE Transactions on Power Systems,26(2):802-810 [百度学术]

      6. [6]

        Quirós-Tortós J,Ochoa L F,Lees B (2015) A statistical analysis of EV charging behavior in the UK.Paper presented at the IEEE PES Innovative Smart Grid Technologies Latin America,Montevideo,Uruguay,5-7 October 2015 [百度学术]

      7. [7]

        Sortomme E,Hindi M M,MacPherson S D J et al (2011) Coordinated charging of plug-in hybrid electric vehicles to minimize distribution system losses.IEEE Transactions on Smart Grid,2(1):198-205 [百度学术]

      8. [8]

        Xi X,Sioshansi R (2014) Using price-based signals to control plug-in electric vehicle fleet charging.IEEE Transactions on Smart Grid,5(3):1451-1464 [百度学术]

      9. [9]

        Callaway D S,Hiskens I A (2011) Achieving controllability of electric loads.Proceedings of the IEEE,99(1):184-199 [百度学术]

      10. [10]

        Ma Z,Callaway D S,Hiskens I A (2013) Decentralized charging control of large populations of plug-in electric vehicles.IEEE Transactions on Control Systems Technology,21(1):67-78 [百度学术]

      11. [11]

        Tavakoli A,Negnevitsky M,Saha S et al (2019) Self-scheduling of a generating company with an EV load aggregator under an energy exchange strategy.IEEE Transactions on Smart Grid,10(4):4253-4264 [百度学术]

      12. [12]

        Neyestani N,Damavandi M Y,Shafie-Khah M et al (2017) Plugin electric vehicles parking lot equilibria with energy and reserve markets.IEEE Transactions on Power Systems,32(3):2001-2016 [百度学术]

      13. [13]

        Vagropoulos S I,Kyriazidis D K,Bakirtzis A G (2016) Real-time charging management framework for electric vehicle aggregators in a market environment.IEEE Transactions on Smart Grid,7(2):948-957 [百度学术]

      14. [14]

        Xu Z,Hu Z,Song Y (2017) Risk-averse optimal bidding strategy for demand-side resource aggregators in day-ahead electricity markets under uncertainty.IEEE Transactions on Smart Grid,8(1):96-105 [百度学术]

      15. [15]

        Shao C,Wang X,Wang X (2017) Hierarchical charge control of large populations of EVs.IEEE Transactions on Smart Grid,7(2):1147-1155 [百度学术]

      16. [16]

        Papadopoulos P,Jenkins N,L.Cipcigan M et al (2013) Coordination of the charging of electric vehicles using a multiagent system.IEEE Transactions on Smart Grid,4(4):1802-1809 [百度学术]

      17. [17]

        Unda I G,Papadopoulos P,Skarvelis-Kazakos S et al (2014) Management of electric vehicle battery charging in distribution networks with multi-agent systems.Electric Power Systems Research,110:172-179 [百度学术]

      18. [18]

        Karfopoulos E L,Hatziargyriou N D (2013) A multi-agent system for controlled charging of a large population of electric vehicles.IEEE Transactions on Power Systems,28(2):1196-1204 [百度学术]

      19. [19]

        Hu J,Zhou H,Li Y et al (2020) Multi-time scale energy management strategy of aggregator characterized by photovoltaic generation and electric vehicles.Journal of Modern Power Systems and Clean Energy,8(4):727-736 [百度学术]

      20. [20]

        Le Floch C,Belletti F,Moura S (2016) Optimal chargi ng of electric vehicles for load shaping:a dual-splitting framework with explicit convergence bounds.IEEE Transactions on Transportation Electrification,2(2):190-199 [百度学术]

      21. [21]

        Bruninx K,Pandžić H,Le Cadre H et al (2020) On the interaction between aggregators,electricity markets and residential demand response providers.IEEE Transactions on Power Systems,35(2):840-853 [百度学术]

      22. [22]

        Bessa R J,Matos M A (2013) Global against divided optimization for the participation of an EV aggregator in the dayahead electricity market.Part I:theory.Electric Power Systems Research,95:309-318 [百度学术]

      23. [23]

        Zheng J,Wang X,Men K et al (2013) Aggregation modelbased optimization for electric vehicle charging strategy.IEEE Transactions on Smart Grid,4(2):1058-1066 [百度学术]

      24. [24]

        Cheng Y (1995) Mean shift,mode seeking,and clustering.IEEE Transactions on Pattern Analysis and Machine Intelligence,17(8):790-799 [百度学术]

      25. [25]

        Guo Y,Xiong J,Xu S et al (2017) Two-stage economic operation of microgrid-like electric vehicle parking deck.IEEE Transactions on Smart Grid,7(3):1703-1712 [百度学术]

      26. [26]

        Bakirtzis E A,Marneris I G,Vagropoulos S I et al (2018) Demand response management by rolling unit commitment for high renewable energy penetration.Paper presented at the International Conference on Smart Energy Systems and Technologies (SEST 2018),Sevilla,Spain,10-12 September 2018 [百度学术]

      27. [27]

        Carrion M,Arroyo J M (2006) A computationally efficient mixed-integer linear formulation for the thermal unit commitment problem.IEEE Transactions on Power Systems,21(3):1371-1378 [百度学术]

      28. [28]

        Bessa R J,Matos M A (2013) Global against divided optimization for the participation of an EV aggregator in the day-ahead electricity market.Part II:numerical analysis.Electric Power Systems Research,95:319-329 [百度学术]

      29. [29]

        Liu B,Pan Z,Tan Z et al (2018) A real-time schedule optimization of massive electric vehicles and energy storage system based on grey wolf optimizer.Paper presented at the IEEE 8th Annual International Conference on CYBER Technology in Automation,Control,and Intelligent Systems (CYBER),Tianjin,China,19-23 July 2018 [百度学术]

      30. [30]

        Zeng M,Liu X,Peng L (2014) The ancillary services in China:an overview and key issues.Renewable and Sustainable Energy Reviews,36:83-90 [百度学术]

      31. [31]

        Department for Transport (2018) Electric chargepoint analysis 2017:domestics.https://www.gov.uk/government/statistics/electricchargepoint-analysis-2017-domestics.Accessed 8 July 2020 [百度学术]

      32. [32]

        Wang M (2019) Focus on electric vehicles to participate in grid interactive services with nearly zero marginal cost.In:Advanced batteries for xEV/ESS conference,Jintan,Changzhou,China,27-29 November 2019.https://www.e-car home.com/newsDetail/2082225049.html.Accessed 8 July 2020 [百度学术]

      Fund Information

      supported by the Science and Technology Project from the State Grid Shanghai Municipal Electric Power Company of China (52094019006U); the Shanghai Rising-Star Program (18QB1400200);

      supported by the Science and Technology Project from the State Grid Shanghai Municipal Electric Power Company of China (52094019006U); the Shanghai Rising-Star Program (18QB1400200);


      • Chen Fang

        Chen Fang received the B.S.and Ph.D.degrees from the Department of Electrical Engineering,Tsinghua University,Beijing,China,in 2006 and 2011,respectively.He is currently working in State Grid Electric Power Research Institute,SMEPC,Shanghai and East China Electric Power Test and Research Institute Co.,Ltd,Shanghai.His current research interests include renewable energy integration of smart grid and power storage technology.

      • Xiaojin Zhao

        Xiaojin Zhao received the B.S.degree in electrical engineering from Shanghai Jiao Tong University,Shanghai,China,in 2019.She is pursuing the M.S.degree at Shanghai Jiao Tong University,Shanghai.Her current research interests include the modeling and optimization of the energy internet and energy management of electric vehicles.

      • Qin Xu

        Qin Xu received the B.S.and M.S.degrees from the Department of Electrical Engineering,Huazhong University of Science and Technology,Wuhan,China,in 2005 and 2007,respectively.She is currently working in State Grid Electric Power Research Institute,SMEPC,Shanghai.Her current research interests include electric energy green management,grid-generation coordination,and substation automatic debugging.

      • Donghan Feng

        Donghan Feng received the B.S.and Ph.D.degrees from the Department of Electrical Engineering,Zhejiang University,Hangzhou,China,in 2003 and 2008,respectively.He has been with the faculty of Shanghai Jiao Tong University,Shanghai,China since 2008,where he currently is a full professor.He also serves as the Deputy Director of the State Energy Smart Grid Research and Development Center,Shanghai,China.His research interests include spot pricing in smart energy networks.

      • Haojing Wang

        Haojing Wang received the master’s degree in Electrical Engineering &Its Automation from North China Electric Power University.She is an engineer in State Grid Electric Power Research Institute,SMEPC,Shanghai.Her research interests include optimized operation of smart grid,micro-grid and distributed generation.

      • Yun Zhou

        Yun Zhou received the B.S.(Hons.) and Ph.D.degrees in electrical engineering from Shanghai Jiao Tong University,Shanghai,China,in 2012 and 2017,respectively.He has joined the faculty of the Electrical Engineering Department,Shanghai Jiao Tong University as a Lecturer since Dec.2017.His current research interests include power system restoration and energy internet.

      Publish Info




      Reference: Chen Fang,Xiaojin Zhao,Qin Xu,et al.(2020) Aggregator-based demand response mechanism for electric vehicles participating in peak regulation in valley time of receiving-end power grid.Global Energy Interconnection,3(5):453-463.

      (Editor Dawei Wang)
      Share to WeChat friends or circle of friends

      Use the WeChat “Scan” function to share this article with
      your WeChat friends or circle of friends