Autonomous Vehicle-to-Grid Design for Provision of Frequency Control Ancillary Service and Distribution Voltage Regulation
Shota Yumiki, Yoshihiko Susuki, Yuta Oshikubo, Yutaka Ota, Ryo Masegi, Akihiko Kawashima, Atsushi Ishigame, Shinkichi Inagaki, Tatsuya Suzuki
AAutonomous Vehicle-to-Grid Design for Provision of FrequencyControl Ancillary Service and Distribution Voltage Regulation ∗ Shota Yumiki † , Yoshihiko Susuki, Yuta Oshikubo ‡ , Yutaka Ota, Ryo Masegi § ,Akihiko Kawashima, Atsushi Ishigame, Shinkichi Inagaki, and Tatsuya Suzuki Abstract
We develop a system-level design for the provision of Ancillary Service (AS) for controlof electric power grids by in-vehicle batteries, suitably applied to Electric Vehicles (EVs)operated in a sharing service. The provision is called in this paper the multi-objectiveAS: primary frequency control in a transmission grid and voltage amplitude regulationin a distribution grid connected to EVs. The design is based on the ordinary differentialequation model of distribution voltage, which has been recently introduced as a newphysics-based model, and is utilized in this paper for assessing and regulating the impactof spatiotemporal charging/charging of a large population of EVs to a distribution grid.Effectiveness of the autonomous V2G design is evaluated with numerical simulations ofrealistic models for transmission and distribution grids with synthetic operation dataon EVs in a sharing service. In addition, we present a hardware-in-the-loop test forevaluating its feasibility in a situation where inevitable latency is involved due to power,control, and communication equipments.
The coordinated use of batteries in electric vehicles (EVs) has been attracted a lot of interestfor the control of electric power grids. In a large-scale power transmission grid, a largepopulation of in-vehicle batteries is being investigated for the provision of ancillary service(AS) to the so-called transmission system operator (TSO), coined in [1, 2] as the V2G. It aimsto shift the peak load (called valley filling) and to provide reserves for primary, secondary, andtertiary frequency controls of the transmission grid: see, e.g., [3]. The V2G for the reserveof primary frequency control (PFC) is called frequency response in PJM [4] and also calledfrequency-controlled normal operation reserve in the Nordic energy region [5]. It requires theachievement of the fast responsiveness of several seconds to several minutes [1]. In-vehicle ∗ This work was supported in part by Japan Science and Technology Agency, Core Research for EvolutionalScience and Technology (JST-CREST) Program [email protected] ) † During this work, S. Yumiki, Y. Susuki, and A. Ishigame were with Department of Electrical and Infor-mation Systems, Osaka Prefecture University, Japan. ‡ Y. Oshikubo and Y. Ota were with Department of Electrical and Electronics Engineering, Tokyo CityUniversity, Japan. § R. Masegi, A. Kawashima. S. Inagaki, and T. Suzuki were with Department of Mechanical Science andEngineering, Nagoya University, Japan. Abbreviation of a regional transmission organization that coordinates the movement of wholesale electricityin all part of 13 states and the District of Columbia in United States of America a r X i v : . [ ee ss . S Y ] J a n atteries are capable of responding faster than synchronous generators used in large thermalpower plants and are hence suitable for the PFC provision. A large volume of literature existsfor the provision of PFC reserve by in-vehicle batteries: as papers including experiments, see[6, 7, 8, 9]. On the other hand, in a small-scale distribution grid, the so-called distributionsystem operator (DSO) [10] is investigated for procuring AS from EVs [11, 12]. The ASis intended for compensating with in-vehicle batteries some of services provided by DSOincluding congestion prevention and voltage magnitude regulation (see [11] in detail). Thecoordinated use of in-vehicle batteries for DSO has been reported recently, see references inthe comprehensive review [12]. The situation where a large population of EVs is connectedto the distribution grid has motivated a research direction in the control of electric powergrids for past ten years [13].One major concern in the situation is with how the impact of EV’s charging/dischargingto the distribution voltage is assessed and regulated. An EV can move anywhere and thusconduct charging and discharging anywhere (precisely, any location such as households andcharging stations) in the distribution grid. In this, the uncoordinated charging of EVs canlead to larger voltage variations and lower power quality. As one of the early studies, theauthors of [14, 15] studied the impact of EV’s charging to the distribution voltage based onthe power-flow equation and proposed an optimization-based scheduling of EV’s chargingconsidering the voltage impact. The authors of [16] presented a review of the current statusand implementation of V2G technologies on distributed grids until 2012. The assessmentand regulation problems are now still actively studied by many groups of researchers, see,e.g., [17, 18] and references in [12]. As stated in [12], since EVs are mainly connected tothe distribution grid, the provision of AS to TSO may affect the distribution grid, which hasbeen scarcely studied until now. One early study is found in [19] where the PFC and voltageregulation in a small power grid are simultaneously achieved with in-vehicle batteries. To thebest of our survey, there is no comprehensive research on methodology and tools for managingthe charging/discharging of EVs in which they provide the PFC reserve while its physicalimpact on the distribution voltage is regulated. The research is of technological significancesince EVs have been becoming more and more popular as new service providers in recentyears [12].Our research team has developed theory and algorithm for solving this problem by meansof EVs in a sharing service [20, 21, 22, 23, 24]. An EV-sharing system has a function oftracking trajectories of EVs to allocate them upon user’s requests [25, 26] and has attractedinterest as the last mile transportation [27]. It is capable of monitoring and managing thestatus of in-vehicle batteries such as state-of-charge and degree-of-health, and hence it canwork for their effectiveness use for not only its primary concern of transportation but also theAS to exiting power grids. The work in [20] motivating the present paper shows a possibilityof cooperative transportation-energy management, in which EV-sharing system and EMSwork consistently. A computational method was developed in [21, 22, 23] for synthesizing aspatial pattern of EV operation in terms of charging/discharging modes, which was referredto as the synthesis of charging/discharging pattern . The method builds upon a new physics-based model of distribution voltage profile, called the nonlinear ordinary differential equation(ODE) model, which is originally derived in [28] and utilized in [29, 24] for the impactassessment of shared EVs in a distribution grid. The method in [21, 22, 23] determines theamount of charging/discharging power of local charging stations, where individual EVs areconnected to a distribution grid, so that the global objectives of AS for frequency controland of voltage regulation are achieved. This idea is called in [23] as the provision of multi- bjective AS by in-vehicle batteries. Here, it should be mentioned that independently fromus, the authors of [30] studied the use of automated, shared EVs for the V2G and stated“This allows for a direct connection to the high voltage electricity transmission in designatedpoints without overloading the low-voltage distribution network.” Similar arguments can befound in the recent review [12] and the recent paper [31]. These indicate that the cooperativetransportation-energy management, possibly combined with the emergent automated driving,is worth pursuing in the control systems technology.The cooperative management was conceptually reported in [20], its architecture for coop-eration between DSO and EV sharing was reported in [32], but its connection to the system-level design for multi-objective AS was not developed so far. We devote the present paperto propose an autonomous V2G design in the architecture based on the preceding papers[33, 32]. The authors of [33] proposed an autonomous scheme of the provision of PFC reserveby distributed EVs, in which its impact on distribution voltage was not considered. In [32],we developed a computational technique for determining an upper bound for the synthesis ofcharging/discharging patterns in terms of the voltage regulation. The technique is based onthe nonlinear ODE model in the similar manner as in [21, 22, 23]. The proposed design hereis a combination of these techniques in [33, 32] and provides a control system for providingthe multi-objective AS. Precisely, charging or discharging power by EVs distributed for mul-tiple stations is regulated with the local measurement of frequency for the provision of PFCreserve, while the maximum amount of charging or discharging power is determined globally for the voltage regulation. Effectiveness of the design is evaluated with numerical simulationsof realistic models for transmission and distribution grids with synthetic operation data onEVs in a sharing service. In addition to the effectiveness, we use the hardware-in-the loop(HIL) testbed partly developed in [23] in order to evaluate practical feasibility of the design.The main contributions of this paper are two-fold: • The autonomous V2G design considering the voltage regulation is introduced andproven to be effective with numerical simulations. The mechanism for the PFC provi-sion is based on the so-called droop-type characteristic [19, 34, 33, 35] and is thereforenot new. In this paper, by combing the mechanism with the technique in [32], we pro-pose the design for providing the PFC reserve by in-vehicle batteries in a decentralizedmanner while regulating (precisely, guaranteeing by design) their impact of to the dis-tribution voltage. This design has a mathematical foundation based on the nonlinearODE model summarized in Section 2. Our development for control of transmission anddistribution grids is novel to the best of our survey. • The practical feasibility of the autonomous V2G design is established with the Power-HIL testbed. The Power-HIL is utilized for testing the coupling of transportation andenergy systems, especially, the V2G: see, e.g., [6, 7, 8, 36]. The importance of HIL forvalidating the AS provision of power-electronics-interfaced distributed generations likeEV batteries is discussed in [37]. Our Power-HIL testing shows that the dynamics offrequency and voltage under the autonomous V2G design are consistently simulated.The simulation is done in a connection of multiple components occurring in practice,such as hardware including EV batteries, software (digital simulator), communicationlines, and measurement devices. It is then shown that the autonomous V2G design This implies the achievement of multiple control objectives by provision of a single AS and does not implythe use of any technique of multi-objective optimization in the computational method. ank straight-line feeder x x L = 0 = Figure 1: Single-line representation of balanced three-phase distribution feeder that starts ata bank (transformer), through a finite-length line with length L , and ends at a non-loadingterminalworks in a practical situation where inevitable latency due to physical, control, andcommunication equipments is involved. Establishing the feasibility of the design isnovel.It should be noted that the conference proceeding [32] as a preliminary work of this paperdoes not contain the autonomous V2G design in Sections 3, 4, and 6, which is the maincontribution of the present paper.The rest of this paper is organized as follows. Section 2 provides a brief review of [32]used in this paper. In Section 3, we describe the autonomous V2G design for the provisionof multi-objective AS. Effectiveness of the proposed design is evaluated in Section 4. Itsfeasibility testing is presented in Sections 5 and 6. Section 7 is the conclusion of this paperwith a brief summary. This section introduces the ODE model of distribution voltage profile based on [28]. A singlefeeder model is shown in Figure 1 that is straight-line and starts at a bank (transformer),where we introduce the origin of the displacement (location) x ∈ R as x = 0. The voltagephasor at the location x is represented by v ( x ) exp { i θ ( x ) } where i is the imaginary unit, v ( x )the voltage amplitude in volt [V], and θ ( x ) the voltage phase in radian. Then, the followingnonlinear ODE is derived in [28] to determine the functions v ( x ) and θ ( x ): d vdx = v (cid:18) dθdx (cid:19) − Gp ( x ) + Bq ( x ) v ( G + B ) , − ddx (cid:18) v dθdx (cid:19) = Bp ( x ) − Gq ( x ) G + B . (1)The constant parameters G and B stand for the per-unit-length conductance and susceptance[S / km]. Also, the function p ( x ) (or q ( x )) is the active (or reactive) power flowing into thefeeder (note that p ( x ) > x ).We will call p ( x ) and q ( x ) the power density functions whose are in [W / km] and [Var / km],respectively. Also, as the important ancillary function in this paper, the voltage gradient [V / km] is defined as w ( x ) := ddx v ( x ) . (2)The ODE poses a nonlinear boundary-value problem and thus can not be analyticallysolved. The authors of [21, 22] propose an approximate solution of the problem. For this4pproximation, consider again the simple feeder model in Figure 1, where N number ofcharing stations and loads are located at x = ξ i ∈ (0 , L ) ( i = 1 , . . . , N ) satisfying ξ i +1 < ξ i .It is assumed that the in-vehicle batteries and loads are operated under unity-power factor.The assumption can be relaxed as in [22]. Thus, by denoting as P i the active-power discharged( P i >
0) (or charged (consumed) ( P i < x = ξ i , the power density functions p ( x ) and q ( x ) are represented as p ( x ) = N (cid:88) i =1 P i δ ( x − ξ i ) , q ( x ) = 0 , (3)where δ ( x − ξ i ) is the Dirac’s delta function supported at x = ξ i . Then, the followingapproximation of solution for w ( x ) is derived in [21, 22]: w ( x ) ∼ (cid:88) j ∈I x P j GY , (4)where Y := √ G + B , I x ⊆ { , , . . . , N } is the set of all indexes i of the locations of chargingstations and loads satisfying x < ξ i , i ∈ { , . . . , N } , ξ N +1 := 0. This implies that the voltagegradient w ( x ) (and hence the voltage amplitude v ( x ) through (2)) can be controlled withthe regulation of charging/discharging power P j by EVs. Based on the observation, we willintroduce a systematic method of how to determine the values of charging and dischargingpower of in-vehicle batteries. This section is a review of [32] and summarizes the computation of upper bound for thesynthesis of charging/discharging patterns. For this, let us assume that all charging stationsconsume non-negative power (operated at the charging mode), i.e., P i := P EVs ,i ≤ i ∈ { , . . . , N sta } , where N sta is their total number. The associated location ofthe i -th station is denoted as ξ sta ,i ∈ { ξ i , . . . , ξ N } . Then, from (4), the deviation of voltageamplitude at the end of the feeder ( x = L ) due to the charging of EVs is approximatelyestimated in [32] as dv ( L ) = N sta (cid:88) i =1 GY | P EVs ,i | ξ sta ,i . (5)See [32] for its derivation. This shows that dv ( L ) is determined with the amounts of chargingpower and locations of the stations. The evaluation formula (5) can work for the case, whereall the stations are operated at the discharging mode (i.e., P EVs ,i ≥ i ).We now describe the method for computing an upper bound of the synthesis of charging(or discharging) patterns [32]. The method is based on the simple evaluation of (5) andeffective for regulating dv ( L ) at its pre-defined acceptable value, which we denote by dV limit .The method eventually works if dv ( L ) > dV limit . For this, we use a common parameter foreach station, denoted by α ∈ [0 , dv ( L ) α := N sta (cid:88) i =1 GY | αP maxEVs ,i | ξ sta ,i = α × dv ( L ) . (6)This dv ( L ) α represents a modification of the measure of (5) by uniformly regulating (decreas-ing) the maximum power of every station. Then, by taking dv ( L ) α = dV limit as the physical5pecification of voltage, α is determined as α = dV limit dv ( L ) . (7)For computing the upper bound, α cha (or α discha ) is calculated with (7) and the pre-definedvalue dV cha , limit at the charging mode (or dV discha , limit at the discharging mode). Hence, it ispossible to compute the upper bound as {− α cha P maxEVs ,i : i = 1 , . . . , N sta } (or { α discha P maxEVs ,i : i = 1 , . . . , N sta } ) so that the deviation of voltage amplitude at the end point of the feederdue to charging (or discharging) of EVs can be bounded at dV cha , limit (or dV discha , limit ).Here, it should be emphasized that the computation of upper bound becomes feasiblewhen the cooperative management works for DSO and EV-sharing operator. It needs bothinformation of distribution grid and mobility system. From EV-sharing operator, DSO re-ceives prediction data of the number of EVs at charging stations in a distribution feeder. Byusing (5), DSO judges whether the deviation of voltage amplitude at the end point of thefeeder is smaller than dV limit or not. For this, it is firstly supposed that all the number ofin-vehicle batteries connected to every station can charge (or discharge) with the maximumpower − P maxEVs ,i ( <
0) and P maxEVs ,i ( > dv ( L ) ≤ dV limit , then the in-vehicle batteries un-der the sharing service can charge or discharge with their maximum power at every station.Otherwise, namely, dv ( L ) > dV limit , then DSO computes the upper bound as shown aboveand sends it to EV-sharing operator in order to keep the voltage level. This section aims to describe the main idea of this paper: to propose the novel control-system referred to as autonomous V2G design for providing the multi-objective AS: PFCreserve and voltage regulation. In this design, charging or discharging power by EVs dis-tributed for multiple stations is regulated with the local measurement of frequency, while themaximum amount of charging or discharging power is determined globally using the methodin Section 2.2. In this sense, the distributed EVs are cooperative with the voltage regulationof a distribution grid.For the multi-objective AS, in the rest of this paper we consider the conventional hierar-chical structure of transmission and distribution grids: in the high-level transmission grid orthe associated TSO, the frequency deviation is determined in a dynamic manner: and in alow-level distribution feeder or DSO, the voltage profile is determined in a static manner. Forthe distribution feeder, we utilize the ODE model in order to assess and regulate the voltageprofile. For the transmission grid, we introduce the block diagram for frequency dynamics inFigure 2 that is based on [39, 23]. The deviation ∆ ω of the grid’s angular frequency from thenominal, 50 Hz in this paper, is determined with the net imbalance ∆ P of supply and demandin the entire grid, the inertia constant M of the grid, and its damping coefficient D . Thenet imbalance ∆ P is calculated with the output of thermal plant determined by economicdispatch control (EDC) [40], load frequency control (LFC) [40], load in the hierarchical grids,power generation by photo-voltaic (PV) units, and charging/discharging power of EVs. Themodel of thermal power plant with turbine and governor is also based on [39, 23].Now, we are in position to describe the autonomous V2G design for determining activepower output by EVs. Specifically, we adopt from [33] the autonomous scheme based ondistributed EVs for realizing the fast responsiveness to TSO. Here, it is supposed that the6igure 2: Block diagram for frequency dynamics of a transmission gridtarget distribution feeder has N sta charging stations labeled by integer i as introduced inSection 2.2. Then, DSO computes the upper bound of charging (or discharging) power as − α cha P maxEVs ,i (or α discha P maxEVs ,i ) at i -th station connected to the feeder at x = ξ sta ,i . In thispaper, we propose that the charging/discharging power P EVs ,i at i -th station connected tothe feeder at x = ξ sta ,i is regulated with the droop-type characteristics as in [33]. To do this,the frequency deviation ∆ f := ∆ ω/ (2 π ) is locally measured at the connection terminal, and P EVs ,i is then determined as shown in Figure 3, P EVs ,i = − α cha P maxEVs ,i , if ∆ f ≤ ∆ fK cha ∆ f , else if 0 ≤ ∆ f < ∆ f K discha ∆ f , else if − ∆ f ≤ ∆ f < α discha P maxEVs ,i , else , (8)where K cha (or K discha ) is calculated with α cha P maxEVs ,i (or α discha P maxEVs ,i ) and the new parameter∆ f . The regulation scheme (8) indicates that EVs connected to i -th station can charge anddischarge within the pre-defined range of frequency, [50 Hz − ∆ f , 50 Hz + ∆ f ], and thusthey behave in a similar manner as the thermal power plant with PFC [40]. The scheme isbased on the local measurement of the grid’s frequency, while the maximum amount of theactive power output is managed globally using the upper bound in Section 2.2. Thus, theEVs distributed for multiple stations are capable of providing the PFC reserve, while theyare cooperative with the voltage regulation.Here, we discuss the capability of providing the PFC reserve to TSO. The capabilitymeasure in the distribution feeder at the charging (or discharging) mode, denoted as ∆ P cha / Hz (or ∆ P discha / Hz ), is defined in this paper as follows:∆ P cha / Hz := − α cha ∆ f N sta (cid:88) i =1 P maxEVs ,i ∆ P discha / Hz := α discha ∆ f N sta (cid:88) i =1 P maxEVs ,i . (9)7igure 3: Charging/discharging regulation with the droop-type characteristics against thefrequency deviation for autonomous vehicle-to-grid designUnder a fixed value of α cha (or α discha ), the capability measure ∆ P cha / Hz (or ∆ P discha / Hz )is inversely proportional to the parameter ∆ f . The measures (9) indicate the regulationreserve of active power by EVs in the distribution feeder per the unit frequency deviationand are closely related to the frequency characteristic requirement in [4] as a parameter ofPFC, to which is referred as the frequency characteristic of a control area.It is valuable to compare the schemes proposed by [33] and in (8). In [33], distributedEVs at i -th station can charge (or discharge) up to the maximum determined by EV-sharingoperator, i.e., − P maxEVs ,i (or P maxEVs ,i ). In the proposed method, distributed EVs at i -th stationcan charge (or discharge) up to the upper bound determined cooperatively by EV-sharingoperator and DSO, i.e., − α cha P maxEVs ,i (or α discha P maxEVs ,i ). Because of α cha and α discha less thanor equal to unity, the capability measure of providing the PFC reserve for [33] is larger thanthose in (8). However, as described in Section 2.2, the proposed scheme provides a cooperativeoperation of distributed EVs for the voltage regulation. No measure for the multi-objectiveAS is reported to the best of our survey, and its unified definition remains to be solved is inour future research. This is related to the current interest of the cooperation of DSO andTSO: see, e.g., [41].As the end of this section, we summarize the autonomous V2G design for providing themulti-objective AS, consisting of the following three steps T1 to T3 : T1 DSO updates the prediction data on the number of EVs at charging stations from EV-sharing operator every pre-defined period. One example of the period is 15 minutes in[38] where EV-sharing operator receives the information on reservations by users every15 minutes. T2 By using the operational data on EVs and the method described in Section 2.2, DSOcomputes the upper bound of charging/discharging of EVs at each station in terms ofthe voltage management. In this paper, we assume that DSO determines the value of dV cha , limit (or dV discha , limit ) and thereby the upper bound for the synthesis of charging8 ank Switchgear x = 0 x = L
18 7 6 4 3 25
Figure 4: Model of multiple feeders based on a practical distribution grid of residential areain Japan. The 8 charging stations are installed and denoted by the circled number: see [22]for details.
Station No. N u m b e r o f E V s Figure 5: Number of electric vehicles (EVs) at the 8 charging stations in the distributionfeeder of Figure 4.(or discharging) patterns as − α cha P maxEVs ,i ( <
0) ( α discha P maxEVs ,i ( > T3 EVs at each station can charge and discharge within the upper bound by DSO. Thisoperation is conducted only with the local measurement of the grid’s frequency. Noexchange of information between any charging stations is required.
The aim of this section is to evaluate the effectiveness of the proposed design using full-digitalsimulations of both the frequency dynamics and the distribution voltage profile.
First, we describe the simulation setting of power grids and mobility system. The setting oftransmission grid in Figure 2 is based on [23], where we assume that the transmission gridis spanned in a geographical region with a population of approximately 9 million customers,and that the grid’s capacity is about 8 . . .
227 Ω / km (or 0 .
401 Ω / km). See [22] for details. Itis supposed that 600 same feeders of the distribution grid are connected to the transmissiongrid. For the mobility system, we use synthetic operation data on EVs in a sharing servicefor the numerical evaluation: see [32] for details. The number of EVs at the 8 stations for15 minutes from 12 o’clock are shown in Figure 5 and used in the numerical simulations.For the evaluation, direct numerical simulations of the nonlinear ODE (1) are conducted.Using the same manner as in [32], to evaluate dv ( L ) for the model of multiple feeders inFigure 4, we focus on the horizontal straight-line feeder with length L = 4 .
63 km and mea-sures the voltage amplitude at its end point as the rightmost part of the model. The es-timation is done with N sta = 8. In this simulation, we set dV limit at the two conditions:( dV cha , limit , dV discha , limit ) = (80 V ,
50 V) and (80 V ,
30 V). For the frequency control, we set∆ f at 0 . {− α cha P maxEVs ,i : i = 1 , . . . , N sta } (or { α discha P maxEVs ,i : i = 1 , . . . , N sta } ). The other case is based on the method of [33]. In this case,distributed EVs at each station can charge (or discharge) up to their maximum power, towhich we will refer as single-objective AS . Figure 6 shows the numerical evaluation of the proposed design under ∆ f = 0 . blue lines) at each station. The group of EVsat each station charges and discharges in the range ( blue line) so that the deviation of voltageamplitude at the end point of the feeder can be within the upper acceptable limit. In the fig-ures (a) and (b), we set dV limit at the two conditions: ( dV cha , limit , dV discha , limit ) = (80 V ,
50 V)and (80 V ,
30 V). In the figure (a), we use the computed values ( α cha , limit , α discha , limit ) =(0 . , . α cha , limit , α discha , limit ) = (0 . , . red, solid lineshows the active power output for single-objective AS, while the blue, dashed line does the ac-tive power output for multi-objective AS. The bottom two figures show the time responses offrequency of the transmission grid. The black, solid line shows the time response of frequencywithout EVs, while the red, solid (or blue, dashed ) line does the time responses for single-objective (or multi-objective) AS. By comparison of the frequency responses with/withoutEVs, we see that the frequency approaches the nominal value, 50 Hz, by charging/dischargingfrom EVs. Here, we compare the frequency responses with EVs for single- and multi-objectiveAS. The red, solid line (single-objective AS) is closer to the nominal value than the blue,dashed line (multi-objective AS). For this, the charging/discharging from EVs at [0 s , ,
36 s] in the middle of Figure 6(a) (or [39 s ,
53 s] in the middle of Figure 6(b)) is boundedwith the proposed design. This implies that the total amount of supply power by EVs in10ulti-objective AS is smaller than that in single-objective AS, and that the single-objectiveAS shows a better performance for the frequency control.The associated data on distribution voltage for the numerical evaluation are shown inFigure 7. The voltage amplitude was computed with the nonlinear ODE (1). The top twofigures show the time series of distribution voltage at x = L in Figure 4. The two black,dashed lines show the upper/lower limits of voltage determined by dV cha , limit or dV discha , limit .The red, solid (or blue, dashed ) lines shows the time series of distribution voltage for single-objective (or multi-objective) AS. The figure (a) (or (b)) indicates that the voltage deviationat the end point of feeder is mitigated by considering the upper bound with α cha (or α discha ).The bottom two figures show the spatial profiles of distribution voltage along the horizontalstraight-line feeder in Figure 4. The figure (a) shows the voltage profile at time 2 s, andthe figure (b) shows the voltage profile at 40 s. In the figure (a), the solid black line showsthe voltage for the loads, the red, solid (or blue, dashed ) line shows that for both the loadsand EVs at the charging mode with unregulated − P maxEVs ,i (or regulated − α cha P maxEVs ,i ). In thefigure (b), the solid black line shows the voltage drop by the loads, the red, solid (or blue,dashed ) line shows that for both the loads and EVs at discharging mode with unregulated P maxEVs ,i (or regulated α discha P maxEVs ,i ). From the figures (a) and (b), we see that the distributionvoltage profile is regulated to be within determined range by the proposed design. This partlyresults from the current condition of the feeder where large loads and PVs are not consideredin Figure 4, and thus the voltage amplitude monotonically decreases from the bank to the endpoint of feeder. Here, the deviation estimated with the nonlinear ODE (1) were 83 . . dV cha , limit and dV discha , limit . This is mainly because the proposed design is based on the approximatesolution of the nonlinear ODE (1), which will be discussed in the next subsection. In the above subsection we showed the performance of the proposed autonomous V2G designnumerically. The capability for providing the PFC reserve to TSO is evaluated with themeasures in (9). As mentioned above, the capability depends on the choice of the parameter∆ f . Thus, the parameter dependence of the total output power from EVs and associatedfrequency responses are shown in Figure 8. The value of ∆ f varies from 0 . . f . In Figure 8(b), we see that the frequency approaches the nominal valuerapidly as the value of ∆ f becomes smaller. This is clearly characterized by the capabilitymeasure ∆ P discha / Hz in (9), which is inversely proportional to ∆ f . Hence, the frequencycharacteristic requirement for the PFC reserve if posed from TSO can be fulfilled sufficientlyfor a small choice of ∆ f .In the end of the last subsection, we mentioned a small error for the voltage regulation inthe proposed design. The control is parameterized with the parameter dV limit that is assignedby DSO. Here, we discuss about how the control error depends on the choice of dV limit andloading condition of the distribution feeder. The total amount of the loads is set as the twoconditions: 10% and 40% of the bank’s rated capacity in Figure 4. The associated directnumerical results on distribution voltage profiles are presented in Figure 9, where the red,solid (or blue, dashed ) line shows the voltage for 10% loading (or 40% loading). Here wedefine the control error as the absolute value of the difference between the direct simulationat the end of feeder and dV cha , limit . The dV cha , limit -dependence of the control error is shown11n Figure 10. In Figure 10, for each of the loading condition, the control error increases with dV cha , limit . This basically comes from the fact that the approximate solution of the nonlinearODE (1) is derived under the condition that all the voltage amplitudes v on the right-handside of (1) are close to unity. Here, we can see in Figure 10 that the control error behavesin a linear manner with dV cha , limit and the loading condition. This suggests that the controlerror can be anticipated with the parameter used in the application. This section is a brief review of the development of Power-HIL testbed for experimentalvalidation of the autonomous V2G design for multi-objective AS. This is built as a laboratoryfacility of a real-time digital simulator and physical devices including an EV battery.
First, we describe the Power-HIL testing for this study. Figure 11 shows the overview ofPower-HIL testbed for validating the multi-objective AS. The testbed includes a real-timedigital simulator and physical devices of PCS, controller, power amplifier, and EV battery. Inthe real-time digital simulator, the frequency dynamics of a transmission grid and the voltagedynamics of a distribution grid are simulated with mathematical models, denoted by “PowerSystem Model” in Figure 11. That is, the frequency deviation ∆ f and the distribution voltageat the stations are calculated with the models introduced in Section 5.2 and sent to “ChargingStation Model” in Figure 11. The block “Charging Station Model” computes the AS signalsfor the stations according to the control logic (8) in Section 3, which provides the regulationlaw of active power output for the provision of multi-objective AS. In addition to the digitalsimulator, the AS signal is sent from “Charging Station Model ” through a communicationline to “Controller” in Figure 11. The block “Controller” is used for commanding the ASsignal (including the information on ∆ f ) to “PCS,” namely “EV Battery.” The Power-HILtestbed is therefore closed in loop by sending the signal to “PCS,” measuring its actual valueof charging/discharging power, and receiving the measured signal at “Power System Model.”In addition, the frequency deviation ∆ f and distribution voltage in “Power System Model”are sent to “Power Amplifier” that emulates in the physical domain the temporal change ofdistribution voltage at one station where the real “EV Battery” is connected. In this section, we describe the mathematical models of transmission and distribution gridsdenoted by “Power System Model” in Figure 11. The setting of the models and parameters isbasically from [23]. Figure 2 is used for the model of transmission grid, where the block “EV”is replaced with the block model of multiple distribution grids as shown in Figure 12. It ishere assumed that 400 same feeders of the distribution grid are connected to the transmissiongrid. Figure 13 shows the model of one distribution feeder that is a single, straight-linefeeder with length 4 . . P refEV , and the output is the active-power output P EV of PCS. Themodel is adopted in Figure 13 to the three stations built in the digital simulator. In thisstudy, we set the parameter T of time-delay as 0 .
30 s and T of time-lag as 0 .
43 s. Thesevalues are based on the measurement of physical devices, which will be discussed in Figure 18through comparison.Here, it is noted that no voltage regulation device is considered in the following analysisbecause the effectiveness of the proposed autonomous V2G design is evaluated in a relativelysimple setting. Related to this, the time-varying PV generation is considered in the model oftransmission grid not the distribution one.
Figure 16 shows the two pictures of the Power-HIL testbed. The testbed contains the real-timedigital simulator manufactured by OPAL-RT Technologies (Model µ s for calculation of the frequency deviation, the distribution voltage, and the charg-ing/discharging power for the charging station. The setting of the time step is enough to thetractable simulation of the distribution grid that exhibits the fastest transient phenomenonin this testing. The frequency deviation ∆ f is calculated every time step, namely 100 µ s.Figure 16 also shows the configuration of Power-HIL. The power amplifier is manufacturedby AMETEK (Model dV cha , limit and dV discha , limit at 80 V. By using the syntheticoperation data on EVs in Figure 14 and the method in Section 2.2, the upper boundfor charging/discharging patterns is pre-computed as {− α cha P maxEVs ,i : i = 1 , . . . , N sta } (or { α discha P maxEVs ,i : i = 1 , . . . , N sta } ) with N sta = 4 in Figure 13. The parameter ∆ f is set at0 . The aim of this section is to establish the practical feasibility of the proposed autonomousV2G design. For this, we show a series of experimental results and implications on thePower-HIL testing.
First, we show the frequency dynamics under the design and see how the physical deviceaffects the performance. Figure 17 shows the Power-HIL simulation of the AS signal andassociated frequency dynamics. The blue, dashed line shows the time series of the AS signalfor one feeder. The positiveness (or negativeness) of the AS signal implies the discharging(or charging) command. Here, we focus on the difference of time-delay and time-lag for thedigital simulator and PCS. For this, we show in Figure 18 the associated time series on inputAS signal P refEV and output power P EV in Figure 15 at Station 1 and Station 2. The red, solid lines in Figure 18 represent the AS signals, and the blue, dashed lines do the output power.The two black, dashed lines show the upper and lower limits as {− α cha P maxEVs ,i : i = 1 , . . . , N sta } and { α discha P maxEVs ,i : i = 1 , . . . , N sta } . The upper/lower limits in Figure 18 change once atthe onset when the operation data on EVs at the four charging stations exhibit the stepwisechange in Figure 14 (although not clearly shown in the left figures of Figure 18 at about120 s). This is because dV cha , limit and dV discha , limit are fixed during the whole simulation,and the parameters α cha and α discha depend on the number of EVs only. In Figure 18(a)for Station 1 as PCS, we see that the time series of input AS signal and output power arenot the same. This difference is manly due to the time-delay of “Controller” in Figure 11and conversion characteristics of the real PCS. In Figure 18(b) for Station 2 on the digitalsimulator, the time series of input AS signal and output power are also not the same becausethe digital simulator considers time-delay and time-lag as shown in Figure 15. The other twostations (Station 3 and 4) show qualitatively the same input/output responses as those in thefigure (b). From these, we conclude that the quantitatively similar latency to the real one issimulated in the software-based stations, and hence that the modeling in Figure 15 involvesdoing the HIL simulation of the grid dynamics affected by the consistent latency.Finally, we validate the proposed autonomous V2G design for the provision of PFC re-serve. The frequency response in Figure 17 is kept close to the nominal 50 Hz, and thus thePFC reserve by EVs works effectively for the frequency control. For comparison, we con-duct the Power-HIL simulation in a case that EVs are fully distributed and can charge or14ischarge with their maximum power as described in Section 4. Figure 19 shows the Power-HIL simulation of the AS signal and associated frequency response. In this case, becausedistributed EVs at each station can charge or discharge for the frequency control only, thegrid’s frequency approaches to the nominal (50 Hz) better by comparison with Figure 17: see[170 s ,
200 s] in Figures 17 and 19. This is expected in the design stage and indeed shownnumerically in Section 4. From the above simulations, we experimentally show that the au-tonomous V2G design works for the provision of PFC reserve, namely, the frequency controlof the transmission grid.
Figure 20 shows the distribution voltage for the Power-HIL simulation associated with Fig-ures 17 and 18. In this figure, the voltage sampled at the six locations including the end( x = 4 . without considerationof upper limit. Figure 21 presents the input/output responses of the two charging stationsin this case. The red, solid lines in Figure 21 show the AS signals, and the blue, dashed lines do the output power. The two black, dashed lines show the upper and lower limits as {− α cha P maxEVs ,i : i = 1 , . . . , N sta } and { α discha P maxEVs ,i : i = 1 , . . . , N sta } . Unlike Figure 18 withthe consideration of upper limit, we see that the magnitude of output power goes over thelimits in Figure 21: see the responses during [170 s ,
200 s] and [260 s ,
280 s]. The associatedtemporal deviations of voltage are shown in Figure 22. Here, the charging/discharging powerat two charging stations is not kept within the determined range for the voltage regulation inFigure 21. We see that the voltage of Figure 22(b) is largely deviated from the upper/lowerlimits compared to the bottom in Figure 20. This implies that the autonomous V2G designusing the upper/lower limits is effective for reducing the deviation of distribution voltage atthe end. It should be mentioned that the voltage in Figure 20 is over the lower limit becauseof the intrinsic approximation error for the computation of upper bound. This is studiedsubstantially in Section 4 and shows that it can be reduced by appropriately choosing thegrid’s condition. The Power-HIL simulation shows that the autonomous V2G design alsoworks for the voltage regulation.
This paper proposed the new design of autonomous V2G for the provision of multi-objectiveAS. The design is computationally simple (no need of optimization), easily implemented (asdemonstrated in Section 5), and relevant in a physical sense because guided by the physics-based models of power grids (as shown in Sections 2.1). Then, we numerically show that theproposed design effectively works for not only the provision of PFC reserve and the regulationof distribution voltage. It is also experimentally shown that the design works in a practicalsituation where inevitable latency (time-delay and time-lag) due to electrical, control, andcommunication equipments is involved, which shows the practical feasibility of the design.15his study, which built upon on the recent development of coupling of transportationand energy systems, is another step toward this development. Several research directions arepossible. In this study, we did not consider the operational strategy of EV-sharing operator,e.g., reported in [20], for assignment and reallocation of EVs. It is in our future work toexplore a detailed design of the architecture in [32] by taking them into account. Relatedto this, the so-called multi-domain HIL simulation of transportation and energy systems ischallenging but could be tackled.
References [1] W. Kempton and J. Tomic, “Vehicle-to-grid power fundamentals: Calculating capacityand net revenue,” J. Power Sources, vol. 144, no. 1, pp.268–279, 2005.[2] J. Tomi´c and W. Kempton, “Using fleets of electric-drive vehicles for grid support,” J.Power Sources, vol. 168, issue 2, pp. 459–468, 2007.[3] M.S. Galus, M.G. Vay´a, T. Krause, and G. Andersson, “The role of electric vehicles insmart grids,” WIRES Energy Environ., vol. 2, pp. 384–40, 2013.[4] Y.G. Rebours, D.S. Kirschen, M. Trotignon, and S. Rossignol, “A survey of frequencyand voltage control ancillary services—Part I: Technical features,” IEEE Trans. PowerSyst., vol. 22, no. 1, pp. 350–357, 2007.[5] “Ancillary services to be delivered in Denmark: Tender conditions,” 44 pages,https://en.energinet.dk/Electricity/Rules-and-Regulations/Approval-as-supplier-of-ancillary-services—requirements; 2019. [accessed 9 October 2020][6] Y. Ota, H. Taniguchi, J. Baba, and A.Yokoyama, “Implementation of autonomous dis-tributed V2G to electric vehicle and DC charging system,” Electr. Pow. Syst. Res.,vol. 120, pp. 177–183, 2015.[7] M. Marinelli, S. Martinenas, K. Knezovi´c, and P.B. Andersen, “Validating a centralizedapproach to primary frequency control with series-produced electric vehicles,” J. EnergyStorage, vol. 7, pp. 63–73, 2016.[8] K. Knezovi´c, S. Martinenas, P.B. Anderson, A. Zecchino, and M. Marinelli, “Enhancingthe role of electric vehicles in the power grid: Field validation of multiple ancillaryservices,” IEEE Trans. Transport. Electrific., vol. 3, no. 1, pp. 201–209, 2017.[9] N.B. Arias, S. Hashemi, P.B. Andersen, C. Træholt, and R. Romero, “V2G enabledEVs providing frequency containment reserves: Field results,” Proc. IEEE InternationalConference on Industrial Technology, pp.1814–1819, 2018.[10] D. Apostolopoulou, S. Bahramirad, and A. Khodaei, “The interface of power: Movingtoward distribution system operators,” IEEE Power Energy Mag., vol. 14, no. 3, pp. 46–51, 2016.[11] K. Knezovi´c, M. Marinelli, A. Zecchino, P.B. Andersen, and C. Træholt, “Supportinginvolvement of electric vehicles in distribution grids: Lowering the barriers for a proactiveintegration,” Energy, vol. 134, pp. 458–468, 2017.1612] N.B. Arias, S. Hashemi, P.B. Andersen, C. Træholt, and R. Romero, “Distributionsystem services provided by electric vehicles: Recent status, challenges, and futureprospects,” IEEE Trans. Intell. Transp. Syst., vol. 20, no. 12, pp. 4277–4296, 2019[13] A. Ipakchi and F. Albuyeh, “Grid of the future,” IEEE Power Energy Mag., vol. 7,issue 2, pp. 52–62, 2009.[14] K. Clement-Nyns, E. Haesen, and J. Driesen, “The impact of charging plug-in hybridelectric vehicles on residential distribution grid,” IEEE Trans. Intell. Transp. Syst.,vol. 25, no. 1, pp. 371–380, 2010.[15] K. Clement-Nyns, E. Haesen, and J. Driesen, “The impact of vehicle-to-grid on thedistribution grid,” Electr. Pow. Syst. Res., vol. 81, pp. 185–192, 2011.[16] M. Yilmaz and P.T. Krein, “Review of the impact of vehicle-to-grid technologies ondistribution system and utility functions,” IEEE Trans. Power Electr., vol. 28, no. 12,pp. 5673–5689, 2013[17] J. Wang, G.R. Bharati, S. Paudyal, O. Ceylan, B.P. Bhattarai, and K.S. Myers, “Co-ordinated electric vehicle charging with reactive power support to distribution grids,”IEEE Trans. Ind. Info., vol. 15, no. 1, pp. 54–63, 2019.[18] J. Dixon and K. Bell, “Electric vehicles: Battery capacity, charger power, access to charg-ing and the impacts on distribution networks,” eTransportation, vol. 4, article no. 100059,2020.[19] M. Tokudome, K. Tanaka, T. Senju, A. Yona, T. Funabashi, and C.-H. Kim, “Frequencyand voltage control of small power systems by decentralized controllable loads,” Proc.Int. Conf. Power Electron. Drive Syst. (PEDS), pp. 666–671, 2009.[20] A. Kawashima, N. Makino, S. Inagaki, T. Suzuki, and O. Shimizu, “Simultaneous opti-mization of assignment, reallocation and charging of electric vehicles in sharing services,”Proc. IEEE 1st Conference on Control Technology and Applications, pp. 1070–1076,2017.[21] N. Mizuta, Y. Susuki, Y. Ota, and A. Ishigame, “An ODE-based design of spatialcharging/discharging patterns of in-vehicle batteries for provision of ancillary service,”Proc. IEEE 1st Conference on Control Technology and Applications, pp. 193–198, 2017.[22] N. Mizuta, Y. Susuki, Y. Ota, and A. Ishigame, “Synthesis of spatial charg-ing/discharging patterns of in-vehicle batteries for provision of ancillary service andmitigation of voltage impact,” IEEE Syst. J., vol. 13, no. 3, pp. 3443–3453, 2019.[23] N. Mizuta, S. Kamo, H. Toda, Y. Susuki, Y. Ota, A. Ishigame, “A Hardware-in-the-looptest on the multi-objective ancillary service by in-vehicle batteries: Primary frequencycontrol and distribution voltage support,” IEEE Access, vol. 7, no. 1, pp. 161246–161254,2019.[24] T. Suzuki, S. Inagaki, Y. Susuki and A.T. Tran (editors), “Design and Analysis ofDistributed Energy Management Systems: Integration of EMS, EV, and ICT,” SpringerNature, Switzerland, 2020. 1725] P. Fairley, “Car sharing could be the EV’s killer app,” IEEE Spectrum, vol. 50, issue 9,pp. 14–15, 2013.[26] D. F. Bignami, A. C. Vitale, A. Lu´e, R. Nocerino, M. Rossi, and S. M. Savaresi (editors),“Electric Vehicle Sharing Services for Smarter Cities: The Green Move Project for Milan:From Service Design to Technological Deployment,” Springer International Publishing,2017.[27] K.K. Tan, K.K.K. Htet, and A.S. Narayanan, “Mitigation of vehicle distribution inan EV sharing scheme for last mile transportation,” IEEE Trans. Intell. Transp. Syst.,vol. 16, no. 5, pp. 2631–2641, 2015.[28] M. Chertkov, S. Backhaus, K. Turtisyn, V. Chernyak, and V. Lebedev, “Voltage collapseand ODE approach to power flows: Analysis of a feeder line with static disorder inconsumption/production,” Preprint arXiv:1106.5003, 2011.[29] Y. Susuki, N. Mizuta, A. Kawashima, Y. Ota, A. Ishigame, S. Inagaki, and T. Suzuki,“A continuum approach to assessing the impact of spatio-temporal EV charging to powerdistribution grids,” Proc. IEEE 20th International Conference on Intelligent Transporta-tion Systems, pp. 2372–2377, 2017.[30] R. Iacobucci, B. McLellan, and T. Tezuka, “Modeling shared autonomous electric vehi-cles: Potential for transport and power grid integration,” Energy, vol. 158, pp. 148–163,2018.[31] T. Chen, K.-F. Chu, A.Y.S. Lan, D.J. Hill, and V.O.K. Li, “Electric autonomous vehiclecharging and parking coordination for vehicle-to-grid voltage regulation with renewableenergy,” Proc. IEEE PES General Meeting, 2020.[32] S. Yumiki, Y. Susuki, R. Masegi, A. Kawashima, A. Ishigame, S. Inagaki, and T. Suzuki,“Computing an upper bound for charging/discharging patterns of in-vehicle batteriestowards cooperative transportation-energy management,” Proc. IEEE 22th InternationalConference on Intelligent Transportation Systems, pp. 655–660, 2019.[33] Y. Ota, H. Taniguchi, T. Nakajima, K.M. Liyanage, J. Baba, and A.Yokoyama, “Au-tonomous distributed V2G (vehicle-to-grid) satisfying scheduled charging,” IEEE Trans.Smart Grid, vol. 3, no. 1, pp. 559–564, 2012.[34] J.A.P. Lopes, F.J. Soares, and P.M.R. Almeida, “Integration of electric vehicles in theelectric power system,” Proc. IEEE, vol. 99, no. 1, pp.168–183, 2011.[35] H. Liu, Z. Hu, Y. Song, and J. Lin, “Decentralized vehicle-to-grid control for primaryfrequency regulation considering charging demands,” IEEE Trans. Power Syst., vol. 28,no. 3, pp.3480–3489, 2013.[36] S. Martinenas, K. Knezovi´c, and M. Marinelli, “Management of power quality issuesin low voltage networks using electric vehicles: Experimental validation,” IEEE Trans.Power Del., vol. 32, no. 2, pp. 971–979, 2017.1837] P.C. Kotsampopoulos, F. Lehfuss, G.F. Lauss, B. Bletterie, and N.D. Hatziargyriou,“The limitations of digital simulations and the advantages of PHIL testing in study-ing distributed generation provision of ancillary services,” IEEE Trans. Ind. Electron.,vol. 62, no. 9, pp. 5502–5515, 2015.[38] R. Masegi, A. Kawashima, N. Mizuta, Y. Susuki, S. Inagaki, and T. Suzuki, “Integratedoperation planning for EV sharing service and distribution voltage support,” Proc. 62ndAnnual Conference of the Institute of Systems, Control, and Information Engineers, 6pages, May 2018 (in Japanese).[39] H. Toda, Y. Ota, T. Nakajima, K. Kawabe, and A. Yokoyama, “HIL test of power systemfrequency control by electric vehicles,” Proc. 1st E-Mobility Power System IntegrationSymposium, pp. 1–6, 2017.[40] G. Andersson, “Dynamics and Control of Electric Power Systems,” Lecture 227-0528-00,ITET, ETH, 2012.[41] P.D. Martini, “Operational coordination architecture: New models and approaches,”IEEE Power Energy Mag., vol. 17, no. 5, pp. 29–39, 2019.[42] S. Kamo, Y. Ota, K. Kawabe, A. Yokoyama, C. Kashiwa, and T. Nakajima, “Au-tonomous voltage and frequency control by smart inverters of photovoltaic generationand electric vehicle,” Proc. 2nd E-Mobility Power System Integration Symposium, pp. 1–5, 2018. 19
10 20 30 40 50 60
Time [s] -0.8-0.400.40.8 E V ou t pu t [ G W ] Single-Objective ASMulti-Objective AS
Time [s] F r e qu e n c y [ H z ] Without EVSingle-Objective ASMulti-Objective AS (a) ( dV cha , limit , dV discha , limit ) = (80 V ,
50 V)
Time [s] -0.8-0.400.40.8 E V ou t pu t [ G W ] Single-Objective ASMulti-Objective AS
Time [s] F r e qu e n c y [ H z ] Without EVSingle-Objective ASMulti-Objective AS (b) ( dV cha , limit , dV discha , limit ) = (80 V ,
30 V)
Figure 6: Numerical evaluation of autonomous vehicle-to-grid design for multi-objective an-cillary service–I: (top) synthesis results of upper bounds for charging/discharging patternsof in-vehicle batteries; (middle) time responses of total output power from electric vehicles(EVs); (bottom) time responses of the grid’s frequency with/without EVs.20
10 20 30 40 50 60
Time [s] V o lt a g e a m p lit ud e [ V ] Single-Objective ASMulti-Objective ASUpper LimitLower Limit
Location [km] V o lt a g e a m p lit ud e [ V ] Without EVSingle-Objective ASMulti-Objective ASUpper LimitLower Limit (a) ( dV cha , limit , dV discha , limit ) = (80 V ,
50 V)
Time [s] V o lt a g e a m p lit ud e [ V ] Single-Objective ASMulti-Objective ASUpper LimitLower Limit
Location [km] V o lt a g e a m p lit ud e [ V ] Without EVSingle-Objective ASMulti-Objective ASUpper LimitLower Limit (b) ( dV cha , limit , dV discha , limit ) = (80 V ,
30 V)
Figure 7: Numerical evaluation of autonomous V2G design for multi-objective ancillaryservice–II: (top) time series of distribution voltage at x = L associated with Figure 4 and(bottom) spatial profile of distribution voltage at time 2 s for (a) and at 40 s for (b).21
10 20 30 40 50 60
Time [s] -0.6-0.4-0.200.20.40.6 E V ou t pu t [ G W ] f =0.2Hz f =0.4Hz f =0.6Hz f =0.8Hz Time [s] F r e qu e n c y [ H z ] f =0.2Hz f =0.4Hz f =0.6Hz f =0.8Hz (a) Whole
27 28 29
Time [s] -0.200.20.4 E V ou t pu t [ G W ] f =0.2Hz f =0.4Hz f =0.6Hz f =0.8Hz
27 28 29
Time [s] F r e qu e n c y [ H z ] f =0.2Hz f =0.4Hz f =0.6Hz f =0.8Hz (b) Zoom-up of [27 s ,
29 s] in (a)
Figure 8: Parameter (∆ f ) dependence of total output power from EVs and associatedfrequency responses under ( dV cha , limit , dV discha , limit ) = (80 V ,
50 V).22
Location [km] V o lt a g e a m p lit ud e [ V ]
10% loading40% loading
Figure 9: Numerical results on distribution voltage profiles for 10% and 40% of the loadingin Figure 4. C on t r o l E rr o r [ V ] Figure 10: Numerical results on control error for the direct numerical simulations in Figure 4.Figure 11: Overview of Power-HIL (Hardware-In-the-Loop) testbed built in this paper. Digi-tal and analog components in this testbed are connected via power and communication lines.23igure 12: Change of the block diagram in Figure 2 for Power-HIL testing. It containsmultiple distribution grids connected to electric vehicles that provide the multi-objectiveancillary service.
BankLoad 5 Station 2Load 2
Load 4 Load 3 Load 1
Station 3 Station 1Station 4 PCS(physical)Digital Simulator
Figure 13: Model of one distribution feeder used for experimental validation. The firstcharging station “Station 1” is emulated with Power Conditioning System (PCS), and theother stations with the real-time digital simulator.
Time [s] N u m b e r o f E V s Station 1Station 2Station 3Station 4
Figure 14: Temporal change of the number of electric vehicles (EVs) at the 4 charging stationsin the distribution feeder of Figure 13. 24igure 15: Dynamic model of input/output response for power conversion. The input is theancillary-service (AS) signal P refEV , and the output is the active-power output P EV of PCS.The model is adopted in Figure 13 to the three stations built in the digital simulator. PCSPowerAmplifier Real-TimeDigital Simulator EV BatteryController
Figure 16: Configuration of Power-HIL (Hardware-In-the-Loop) testbed. It contains thereal-time digital simulator OPAR-RT Technologies (Model
Time [s] F r e qu e n c y [ H z ] -1500-1000-500050010001500 R e gu l a ti on [ k W ] FrequencyRegulation
Figure 17: Generated ancillary-service (AS) signals to one distribution feeder as the primaryfrequency control (PFC) reserve and associated frequency responses of the transmission grid.The blue, dashed lines show the AS signals, and the red, solid lines do the grid’s frequency.25
80 160 240 320
Time [s] -500-400-300-200-1000100200300400500 P o w e r [ k W ] Input SignalOutput PowerUpper LimitLower Limit
160 170 180 190 200
Time [s] -500-400-300-200-1000100200300400500 P o w e r [ k W ] Input SignalOutput PowerUpper LimitLower Limit (a)
Station 1 (built on Power Conditioning System (PCS))
Time [s] -500-400-300-200-1000100200300400500 P o w e r [ k W ] Input SignalOutput PowerUpper LimitLower Limit
160 170 180 190 200
Time [s] -500-400-300-200-1000100200300400500 P o w e r [ k W ] Input SignalOutput PowerUpper LimitLower Limit (b)
Station 2 (built on the digital simulator)
Figure 18: Input/output responses of the two charging stations associated with Figure 17.The red, solid lines show the input ancillary-service (AS) signals, and the blue, dashed linesshow the output power, and the two black, dashed lines show determined the upper/lowerlimits. The left and right figures show the whole time-series during [0 s ,
320 s] and theirzoom-up during [160 s ,
200 s], respectively. 26
80 160 240 320
Time [s] F r e qu e n c y [ H z ] -1500-1000-500050010001500 R e gu l a ti on [ k W ] FrequencyRegulation
Figure 19: Generated ancillary-service (AS) signal and associated frequency response whendistributed electric vehicles can charge/discharge up to the maximum power without consid-eration of upper/lower limits. The result is derived by the Power-HIL simulation.
Time [s] V o lt a g e a m p lit ud e [ V ] Time [s] V o lt a g e a m p lit ud e [ V ] voltage (at 4.5km)Upper LimitLower Limit Figure 20: Time series of sampled distribution voltage that are associated with Figure 17.(Upper) The voltage sampled at the six locations x = 0 km (bank), 1 . . . . . .
80 160 240 320
Time [s] -500-400-300-200-1000100200300400500 P o w e r [ k W ] Input SignalOutput PowerUpper LimitLower Limit (a)
Station 1
Time [s] -500-400-300-200-1000100200300400500 P o w e r [ k W ] Input SignalOutput PowerUpper LimitLower Limit (b)
Station 2
Figure 21: Input/output responses of the two charging stations when distributed electricvehicles can charge/discharge up to the maximum power without consideration of upperlimit as in Figure 19. The red, solid lines show the input signals, the blue, dashed lines showthe output power, and the black, dashed lines show the determined (but not used here) limitsfor the voltage regulation.
Time [s] V o lt a g e a m p lit ud e [ V ] (a) x = 4 . Time [s] V o lt a g e a m p lit ud e [ V ] voltage (at 4.5km)Upper LimitLower Limit (b) End ( x = 4 .5 km) with upper/lower limits