Aggregated Electric Vehicles load profiles with Fast Charging Stations

Transcription

1 Aggregated Electric Vehicles load profiles with Fast Charging Stations G. Celli, G. G. Soma, F. Pilo, F. Lacu, S. Mocci, N. Natale Department of Electrical and Electronic Engineering University of Cagliari Piazza d Armi, Cagliari ITALY celli@diee.unica.it, ggsoma@diee.unica.it, pilo@diee.unica.it, susanna.mocci@diee.unica.it, nicola.natale@diee.unica.it Abstract--The concept of electrical-mobility in opposition to the present oil-mobility is becoming even more attractive worldwide. Fast Charging Station (FCS) refers charging stations with nominal power equal or higher than 50 kw. The impacts of EV charging on electricity grids is becoming an increasingly important subject of study, but detailed knowledge about the future charging profiles of EVs appears to be missing in Literature. The FCS requires high power and they must be connected to MV networks. For that reasons, it is crucial to analyze these situations and to model the FCS consumption, in order to correctly plan the expansion of the MV system. In the paper a Monte Carlo simulation methodology is proposed to model the aspects that influence the request of fast charge for EVs. The EV charge profiles should be used in the planning and in the EV impact analysis of the future networks. Keywords Fast Charging Station; Electric Vehicles; Monte Carlo simulation; Distribution Planning. I. INTRODUCTION Under the pressure of even more severe environmental constraints, the electrification of the transportation sector is becoming even more attractive worldwide, due to the dependency reduction from liquid fuels in this sector and the increment of primary energy sources diversity used in countries energy mixes. Currently, the transport sector relies on fossil fuels, causing a significant part of greenhouse gas emissions. The car for passengers is the major consumer of energy, accounting for more than half of the total transportation energy. The electrical mobility is based on the usage of battery powered electric vehicle (EV) and Plug-in Hybrid Electric Vehicle (PHEV) as the main future technology to combat greenhouse gas emissions [1]. If the number of electric vehicles grows substantially, the impacts on the power system will increase as well. Since this is recognized widely, one observes an increasing number of studies on the impacts of EVs on the power system. From the point of view of the electrical infrastructures, the key questions are when and where drivers would recharge their vehicles. The primary source of charging will rely on normal charging boxes, located at home or in the parking at work and operated manually by the driver or, preferably, managed remotely by a suitable control system (owned by the local distributor or by an independent aggregator) [2]. In both cases, 3 kw AC slow chargers will be spread in the LV network (home chargers) or concentrated in some parking lots and connected to the LV or MV network. Alternative to the slow charges, fast charges will occur when previous charging options are not available or when, in the middle of a trip, the battery approaches minimum SoC (State of Charge). Fast charging refers to DC charging stations with nominal power equal to or higher than 50 kw. Some European Original Equipment Manufacturers (OEMs) expect a charging rate of up to kw for a typical EV battery as a realistic target for DC fast charging in year 2020 [3]. Consequently, a Fast Charging Station (FCS) will be characterised by high momentary peak power absorptions and it must be connected to MV networks. Since in a future scenario FCS will be the equivalent of today's fuel stations, the prior knowledge of the profiles of charging to be met is a fundamental aspect to evaluate the optimal allocation of these resources, the impact these will have on distribution networks and, last but not least, to have a clearer view on the investment statement relating to these infrastructures. For these reasons, it is crucial to model the FCS consumption, in order to correctly plan the expansion of the MV system [4]-[5]. The first step of this representation is the definition of a daily profile of the power absorbed by a FCS. In fact, if some assumptions used to build the FCS demand profile are incorrect, the network investment would be overestimated or underestimated, resulting cost-ineffective or causing power quality deterioration. Therefore, in-depth studies should be performed on modelling the future behaviour of EV s drivers. In the recent literature, some models were proposed starting from real data of current travel behaviour of ICE (Internal Combustion Engine) vehicle drivers: mobility profile, arrival time distribution at refuelling station, departure time and daily average distance [3], [6]. Given the uncertainties that characterize the problem, the majority of these approaches is based on probabilistic calculation with the goal to obtain an average load profile. However, they are often excessively dependent on the current behaviour of ICE vehicle drivers and do not take into due account the probable changing in the behaviour of the future EV drivers. In the paper, a Monte Carlo simulation methodology is proposed to model, with suitable probability distributions, several aspects that may influence the request of fast charge for EVs: the driven distance of each journey, the speed and the gas consumption (that depend on the driving style), the departure time, the initial SoC (taking account also of the existence of home charging facilities), the SoC threshold that induces the driver to recharge its EV, and some others. The proposed procedure is also able to perform the calculation considering always the availability of fast charging facilities that allows

2 recharging the EV immediately when it is necessary (ideal assumption), or assuming a prefixed number of charging points and the consequent possible occurrence of a queue in the FCS during the peak recharging hours. At the end of this stochastic process, the proposed methodology gives a daily demand profile with per minute or per hour granularity, with an expected demand value and a standard deviation for each interval. This representation is fundamental for the expansion planning of Active Distribution Systems that requires probabilistic models for a better representation of the uncertainties in the planning data, and the introduction of the risk concept in the selection of planning alternatives [7], [8]. In Section II the reasons that lead to the choice of a probabilistic approach are presented. In Section III the Monte Carlo simulation methodology is illustrated, whereas in Section IV the results and discussion of the proposed approach are carried out. II. PROBABILISTIC VS DETERMINISTIC PLANNING A. Inadequacy of Traditional Planning Distribution networks are, in general, sized to cope with the worst-case scenario of a given load forecast and in a way that minimum or no operation is required. This approach, known as fit and forget, is carried out in a deterministic way, i.e., without considering uncertainties. Once a planning study is defined, different alternatives might be considered. The most cost-effective solution is finally the planning alternative likely to be adopted. While this passive way of planning and operating distribution networks has proven cost-effective in the last decades, it might in the future become a barrier for increasing penetrations of renewable generators and nonconventional loads, like the EVs. Indeed, with the increment of uncertainties brought by these new distribution system s customers, the fit and forget approach results in massive network investments only motivated to deal with worst-case scenarios that may have an extremely low probability of occurrence. B. Probabilistic Models In order to overcome the limitations previously stated, the use of a more active approach for managing distribution networks has been proposed by academia and industry. This goal suggests both the use of probabilistic models, to better represent planning data, and the introduction of the risk concept in the selection of planning alternatives. In the Literature the studies about the definition of the FCS charging profile face the problem of determining the time at which the charging takes place. Several authors are facing this problem with different approaches, although with the same goal, namely to determine when and in what quantity the energy will be required from the network [3], [9]-[11]. It is evident that the objective is a function of the behaviour of the individual driver and the composition and characteristics of the vehicle fleet that the network will host in the future. In the proposed methodologies there is therefore a common guideline: the highly stochastic nature of the parameters in the game. The state of the art presents mostly probabilistic approaches, based on parameters characterized by suitable probability distributions and appropriate assumptions. To represent the random variables, the most suitable probability distribution is the normal or Gaussian pdf (probability density function). This distribution, in fact, better reflects the current behaviour of consumers and the general randomness of the variables under consideration. For the highly uncertain nature of the variables that characterize the problem, the Monte Carlo simulation is used in almost all studies. In [9] a comparison with a deterministic approach has been proposed, going to consider the unique values for the input variables, under certain assumptions. The FCS charging profile obtained with the deterministic approach presents values much greater compared to the profile obtained with the probabilistic approach. It is clear that the assumptions made by the latter approach are the most compelling and better reflects the real scenario. However, among different probabilistic methodologies the most significant differences regard the estimation of the time when the recharge is required. Some of the studies consider the charging times proportional to the current mobility charts [3], in [9] they have been considered equivalent to the current arrival times to the traditional gas stations, while in [10] the actual departure times of vehicles are taken into account, considering more departures per day for the same vehicle (for example, starting in the morning and return in the evening). The hypothesis assumed in [3], acceptable from a probabilistic point of view (the more cars circulate, the greater is the possibility that a number of vehicles needs to recharge), however does not consider the real need of charging, related to the path travelled, to the driver habits and to the vehicle characteristics. In [9] the possible variation of drivers' behaviour or need has not been considered, which could involve a totally different use of an electric vehicle, and therefore different consumption and autonomy, compared to traditional combustion vehicles. Instead, by considering the departure times, the distance covered and the average speed it is possible to interrelate in a more effective way the charging time to the actual charging needs of the vehicle [10]. In conclusion, for the definition of the FCS load profiles the most suitable calculation method is absolutely the probabilistic one. In this way an average value of the outputs is obtained, characterized by appropriate variables, and calculated on the basis of as many as possible input values and of the uncertainty associated with these values. In addition, to determine the time at which the charging will be necessary it is not reasonable to use the diagrams of mobility, which provide no information on the actual need of recharging, but rather it is better to consider each single vehicle, with its own characteristics and mobility habits, and then to analyse the set of all the vehicles that are going to use the FCS. The proposed methodology for the definition of load profiles will be deeply described in the next Section.

3 III. METHODOLOGY FOR THE DEFINITION OF FCS LOAD PROFILES In the near future the FCSs will overlap and start to replace the traditional gas stations. Thus, an a priori knowledge of the daily load profiles is crucial to understand their interaction with the electricity grid and to estimate their impact on power distribution network investments. Indeed, in every power system planning study it is essential the characterization of all customers in terms of electric power consumed or generated. For new loads, like the FCS, where historical statistics are not available, this could become a critical issue because an incorrect representation can easily take to underestimate or overestimate their impact on network planning and operation. For this particular load, a pivotal aspect is represented by the day-to-day behaviour of the EV driver that is influenced by several factors (e.g., average covered distance, driving style, daily variation of recharging price, etc.) and can drastically change the EV recharging moment and, consequently, the overall daily load profile of the FCS. Therefore, the first goal assumed in the paper has been the development of a general automatic procedure versatile enough to generate different daily load profiles for FCS suited to different scenarios. In the definition of an FCS load profile it must be taken into due account the high level of randomness of the factors that influence the EV driver behaviour. Accordingly, the representation of this new load has to accent these uncertainties so as to be correctly treated by modern probabilistic planning tools [8]. In order to achieve these two goals for the FCS model (flexibility and probabilistic representation), the proposed procedure is based on a Monte Carlo simulation algorithm, where the movements of a fleet of EVs in a typical day have been simulated several times, detecting when the fast recharging of each vehicle becomes necessary. By so doing, it is found the probability distribution of the number of EVs that, in each hour, are connected to the FCS to recharge their batteries. Known the power associated to the fast charging and the energy required by each vehicle, the probabilistic daily load profile of the fleet of EVs is immediately obtained and, if no hypotheses are made on the vehicles flows (EVs and FCSs uniformly distributed over a specific geographical area), it can be easily scaled to the load profile of a single FCS of a given number of recharging poles. The results of this procedure could be used also to estimate the number of fast charging stations that are needed for a specific penetration of electric vehicles assumed in the study. The definition of the FCS load profile starts inevitably from the description of the mobility scenario. Three aspects have to be outlined: the characteristic of the area of study, the habits of the driver and the features of the EVs fleet and of the FCSs. The main data used are collected in Table I. The first two categories of information can be obtained from the analysis of the mobility data of the geographical area under investigation. An important factor to consider when dealing with these data is represented by the presence of slow charging facilities used during long parking time (e.g. at home, at work or in a shopping centre car park). Recent surveys have shown that domestic slow charge will be the preferred charging option, most probably to exploit convenient electric energy tariffs during the night [12]. Therefore, the drivers that prevalently use the FCS will be commuters or those that does not have domestic slow charging facilities. The extension of the region from which commuters come from allows identifying an average covered distance and its variability. TABLE I MAIN DATA USED FOR THE DEFINITION OF THE MOBILITY SCENARIO. Area characteristic Driver s habits EVs fleet and FCSs features Total number of ICE vehicles (N ICE) % of commuters Covered distances Departure hour Parking interval Average speed EV penetration (% of N ICE) type of EVs (small, medium and large) and their number EV s battery capacities EV s consumptions Home charging option (%) Fast Charging power N of poles per FCS For commuters the habits are regular during the workdays since they always start at the same hour, stay at work generally for the same period and drive similar distances. During weekends or for different categories of drivers, shopping or leisure trips may happen at different times and vary significantly in distances. Regarding the current drivers preferences for refuelling, they refuel their vehicle only when the fuel tank is nearly empty. It is not immediate the application of this habit to the future EV drivers, due to the longer time for refuel (recharge the battery), the possibility to do it in facilities different from dedicated fuel/charging stations and the application of special prices. It is evident that the introduction of specific incentives and tariffs for slow and fast charging may influence drastically this behaviour, by changing the time, the location and the technology used for recharging the electric vehicles. In this paper, these economic aspects have been disregarded, assuming that the drivers do not change appreciably their current behaviour with the introduction of the electric vehicles, and they still resort to FCSs when the SoC of the EV s battery is approaching a certain limit. However, the developed procedure is enough flexible to be easily adapted also for different scenarios, as it will be shown in the following. In the definition of the fleet of EVs, three categories have been considered according to the technical characteristics, mainly depending on the capacity of the batteries (Table II). TABLE II TECHNICAL CHARACTERISTICS OF THE THREE EV CATEGORIES. EV category Battery capacity range Consumption range [kwh] [kwh/km] Small ± 0.03 Medium ± 0.03 Large ± 0.03 The whole procedure is summarized in the flowchart of Fig. 1. The definition of the mobility scenario is built firstly by assigning some parameters that are not subject to change during the calculation and that characterize the specific study, like the foreseen penetration of EVs, the total percentage of EV owners that have the domestic slow charge availability, the number of commuters and the covered distance (given by a normal distribution with mean value and standard deviation).

4 outward and return trip) that identifies the need of recharge the battery are the typical data used in each Monte Carlo run. Once all these data have been assigned, the SoC variation during the day can be calculated for the battery of each EV involved (from the departure to the arrive at home) and FCS. The SoC decrement has been approximated with the following linear equation: SoC arrive = SoC departure D C 100 (1) BC i where and the SoC departure SoC arrive are the states of charge of the battery (in %) of the i th EV in the j th day at the departure and the arrive of the trip (e.g. from home to work), D i,j is the distance covered in the specific trip (in km) by the i th EV in the j th day, C i,j is the average consumption of the i th EV during the j th day (in kwh/km), and BC i is total battery capacity of the i th EV (in kwh). For each trip it is extracted the SoC threshold from a normal distribution, SoC thr. If this value is between and SoC departure SoC arrive, the EV has to be recharged during the trip. SoC thr is then reduced by a factor α to take account of the energy consumed to reach the FCS: ( ) = α SoCthr (2) SoC FCS The developed procedure has the possibility to specify different thresholds for the outwards and the return trips (Fig. 2). This allows increasing the versatility of the procedure and has been used in the paper to represent the probable EV driver disposition to recharge his vehicle during the return trip in order to have it ready in the next morning and avoid to waste time going to work for recharging. By using this option, the procedure can be used to represent also different scenarios, like the presence of different tariffs for the fast charging, moving the disposition to recharge of the EV driver in different periods of the day. Fig. 1 Flowchart of the FCS load profile definition procedure. After this assignment, the EV fleet is defined car by car by extracting randomly the battery capacity of the vehicle, the characteristics of the driver (commuter or non regular driver) and, for each commuter, the departure hour and the kind of employment (full-time or part-time) from which depends the parking interval (i.e., 8 or 4 hours) and the departure hour of the return trip. Then, for each day simulated with the Monte Carlo algorithm all the uncertain data are extracted from normal or uniform distributions. The minute of departure (from home or work), the average consumption during the trip (within the range of the vehicle category), the average speed, the distance covered, the initial SoC of the vehicle s battery for those drivers that do not have the availability of domestic slow recharge or that have forgotten to do that (low probability), and the SoC threshold (eventually different for Fig. 2 Probabilistic calculation of the expected FCS recharge time. Known the state of charge when the EV enters into the FCS (SoC FCS ), the exact distance covered from the departure place is derived from eq. (1): SoC ( ) ( ( ) departure SoC ) FCS = ( ) BC i (3) D FCS 100 C and the minute at which the fast recharge starts, ( ) t FCS, is:

5 ( ) = tdeparture t FCS + 60 D FCS (4) υ where t departure is the departure minute of the trip and υ i,j is the average speed of the i th EV in the j th day (in km/h). In addition to the time of the recharge, the second parameter needed to derive the FCS load profile is the duration of the recharge. The energy required by the recharging EV, E FCS, can be easily derived considering that fast charging is applied to a maximum limit of the 80% of the battery capacity, to avoid quick degradation of the battery (the remaining 20% has to be recharged slowly): ( ) ( ) SoC = 0.8 FCS BC 100 i (5) E FCS Then, the duration in minutes of the fast recharge is: E Δt FCS = 60 FCS (6) P FCS where P FCS is the fast charging power (assumed 50 kw). By summing all the charging profiles of the EV fleet, the FCS load profile for a specific day is determined. Two kinds of days are considered in the Monte Carlo calculations, workdays and weekends by adapting where necessary the probability distributions of some parameters (e.g. the covered distances). The calculation is made on a minute scale in order to avoid boundary issues (all the vehicles that recharge simultaneously at the beginning of the hour), but the results are given in per hour basis, more treatable for post calculation analyses (e.g. planning studies). This load profile is given in terms of an average daily curve and a standard deviation different for each hour of the day (Fig. 3). : average hourly power demand for fast charging : band of uncertanty Fig. 3 - Example of the daily power demand profile of EVs for fast charging (4000 vehicles without home charging option). The probability distribution of the FCS demand in each hour is assumed normally distributed. This assumption is confirmed by the analysis of the results that show a good accordance with the Gaussian distribution (Fig. 4). The maximum value of the power demand profile, registered in the peak hour (e.g. the 18 th hour of Fig. 3), can be used to estimate the ideal number of FCSs in the geographical area of interest that assures always the availability of a fast recharging pole. It is sufficient to divide this maximum power request by the nominal rate of the FCS (e.g. 300 kw for an FCS with 6 poles of 50 kw). Obviously, this is a rough approximation because it does not consider the real paths of the vehicles flows, but it can be used as a good starting point for deeper analyses. Probability of occurrance Fig. 4 - Frequency of the power demand of the whole EV fleet for fast charging in the 18 th hour of the day. The proposed procedure is also able to work with a prefixed number of FCSs. In this case, if this number is insufficient to satisfy the whole maximum power demand, the daily load profile is modified to take account of the queue of EVs. From the ideal calculation (without limit on the FCS) it is extracted an average fast charging duration that is used to estimate the waiting time the vehicles have to spend on the queue before starting their fast recharging. In this way the ideal daily load profile is shaved to the maximum power that the FCSs can overall provide and shifted to the right due to this delay. IV. RESULTS AND DISCUSSION The developed procedure has been tested on the area of Cagliari (Sardinia) and its environs. The parameters used to define the mobility scenario are collected in Table III. The values have been derived from existing reports of the local government, from Italian and European statistics or rationally chosen. In particular, the 50 km/h of average speed has been assumed by considering the specific traffic conditions of the Cagliari s district and a mix of extra-urban and urban paths for the daily trip of each commuter. Moreover, the assumption of the initial SoC for those vehicles without the home charging option (normal distribution - mean of 60% - standard deviation of 6%) has been derived by simulating several times the behaviour of these vehicles during an interval of a week and recording the SoC at the beginning of each day (on Monday the initial SoC considered is 80%, the limit for the fast recharge). Additional investigation has to be done to recognize how stationary these distributions of SoC are during the day (not having the availability of home recharge, the distribution is the same at the beginning and the end of the day).

6 TABLE III DEFINITION OF THE MOBILITY SCENARIO Parameter Value Total number of ICE vehicles N ICE = EV penetration 20% (N EV = 20000) Home slow charging opportunity Percentage of who forgets to recharge at home Category of drivers Parking time 80% (16000 EVs) μ = 10% σ = 2% 80% commuters (83% full-time, 17% par-time) 20% non regular drivers Full-time commuter Par-time commuter workdays 8 h 4 h weekends Average speed [km/h] Covered distance [km] Initial SoC (at home) SoC threshold workdays weekends recharge at home no recharge at home outward return Non regular driver 1 h 3 h (uniform distribution) 1 h 5 h (uniform distribution) μ = 50 σ = 7 μ = 30 σ = 5 μ = 40 σ = 7 100% μ = 60% σ = 6% μ = 30% σ = 1.5% μ = 40% σ = 2% Reduction coefficient (α) 0.8 Fast charging power (kw) 50 N of poles per FCS 6 The general load profile obtained for the whole EV fleet is the one previously depicted in Fig. 3. It shows a power demand mainly concentrated in the afternoon and in the evening, as expected, due to the natural discharge of the batteries during the day and the simulated disposition of the drivers to recharge during the return trip. From a deeper analysis of the results, no driver that has recharged at home manifests the need to recharge in an FCS during the workdays, showing that with the covered distances considered (typical of the region under investigation) a regular domestic slow charge practice is sufficient. Only in the weekends, with longer distances covered for country outings, a small percentage of these vehicles resort to the fast recharge. Indeed, their bearing in the average daily energy requirement for fast recharging is limited to little more than 3%. The maximum number of EVs simultaneously recharged in an FCS is 208 that leads to an ideal number of 35 FCSs. The average fast recharge duration results in about 12 minutes. A second simulation has been performed increasing the distances covered by the commuters in the workdays (average distance of 100 km and a standard deviation of 10 km). In this case, the need to recharge increases in the morning and the evening peak is exalted due to the higher number of EVs that arrives at the end of the day with low SoC values (Fig. 5). Moreover, the overall daily energy requirement is largely greater (almost quintupled), meaning that the domestic slow recharge is no more sufficient to avoid the resort to the fast charging and a larger number of EVs uses the FCSs. Indeed, more than 70% of the total daily energy demand for fast recharge is requested by EVs with the availability of domestic slow recharge. The maximum number of EVs simultaneously recharged in a FCS becomes 537 (almost tripled) that leads to an ideal number of 90 FCSs. x Fig. 5 Daily power demand profile of the whole EV fleet for fast charging obtained considering longer covered distances (μ = 100 km). Maintaining these longer distances but increasing the number of drivers with the availability of domestic slow charging causes a shift of the fast charging time towards the evening hours with an additional growth of the peak, because more drivers depart from home with the battery fully charged. The result of the simulation with a total availability of the domestic slow charging (100%) is shown in Fig. 6. x Fig. 6 - Daily power demand profile of the whole EV fleet for fast charging obtained considering longer covered distances (μ = 100 km) and 100% of domestic slow charging availability. On the contrary, if the opportunity of home slow charging is drastically reduced (for instance, 20% of the EV fleet), the peak of the demand appears in the morning, because an insufficient initial SoC characterizes many EVs (Fig. 7). Almost the entirety of the EV fleet resort to the fast recharge during the workdays, increasing even more the whole daily energy requirement and the maximum peak of power demand:

7 711 EVs simultaneously recharged in a FCS, leading to an ideal number of 125 FCSs. x Fig. 9 - Daily power demand profile of the whole EV fleet for fast charging obtained considering longer covered distances (μ = 100 km) and lower domestic slow charging availability (20% of the fleet). The final cases examined use again all the data of Table III, but a fix number of FCSs lower than the ideal value (35) is imposed. In this scenario, the maximum values of the demand in each hour of the load profile is bounded to the maximum power available, and some saturation effects start to appear shifting the time of fast recharge also in the night. In Fig. 8 and Fig. 9 this behavior is shown for the case with 15 FCSs and with 10 FCSs. Obviously, in these cases the average duration of the fast recharge increases due to the waiting time spent on the queue (about 24 minutes with 15 FCSs and more than 36 minutes with only 10 FCSs). V. CONCLUSIONS The paper presented an original Monte Carlo methodology for assessing the power demand of a Fast Charging Station. The methodology allows assessing the average consumption and its variance taking into account several stochastic quantities and the availability of FCS nearby. The analysis of the results allows, for instance, identifying the impact of FCS number and position on queue effect. The FCS load consumption is the first step of a more complex study that aims at identifying the optimal number and position of FCS in given area taking account of the expected consumption for fast recharge. Since the FCS power demand is directly influenced by the number of other FCSs as well as traffic situation, an integrated probabilistic planning algorithm will be used for the simultaneous placement of FCS. VI. REFERENCES [1] International Transport Forum, 2010: Reducing transport greenhouse gas emissions Trends & Data, OECD/ITF, available on-line at: [2] G. Celli, E. Ghiani, F. Pilo, G. Pisano, G. G. Soma, 2012, Particle Swarm Optimization for Minimizing the Burden of Electric Vehicles in Active Distribution Networks, Proceedings IEEE PES General Meeting, S. Diego (USA). [3] G. Mauri, A. Valsecchi, 2012, Fast charging stations for electric vehicle: The impact on the MV distribution grids of the Milan metropolitan area, Proceedings IEEE International Energy Conference and Exhibition (ENERGYCON), Florence (Italy), 9-12 Sept. 2012, pp Fig. 7 - Daily power demand profile of the whole EV fleet for fast charging with 15 FCSs available. Fig. 8 - Daily power demand profile of the whole EV fleet for fast charging with 10 FCSs available. [4] G. Celli, S. Mocci, G. G. Soma, F. Pilo, R. Cicoria, G. Mauri, E. Fasciolo, G. Fogliata, Distribution network planning in presence of fast charging stations for EV, in Proc CIRED conference. [5] R. Green II, L. Wang, and M. Alam, "The impact of plug-in hybrid electric vehicles on distribution networks: A review and outlook," Renewable and Sustainable Energy Reviews, [6] M. Simpson, 2012, Mitigation of Vehicle Fast Charge grid impacts with renewable and energy storage, presentation at the 26 th international Electric Vehicles symposium, Los Angeles (USA). [7] C. Abbey, A. Baitch, C. Carter-Brown, G. Celli, K. El Bakari, S. Jupe, F. Pilo, F. Silvestro, J. Taylor, Planning and optimisation of active distribution systems An overview of CIGRE Working Group C6.19 activities, Proc. of CIRED Workshop, Lisbon, May, [8] G. Celli, E. Ghiani, F. Pilo, G. G. Soma, New electricity distribution network planning approaches for integrating renewable, Wiley Interdisciplinary Reviews: Energy And Environment, vol. 2, pp , 2013; ISSN X. [9] K. Yunus, H. Z. De La Parra, and M. Reza, "Distribution Grid Impact of Plug-In Electric Vehicles Charging at Fast Charging Stations Using Stochastic Charging Model," Proc. of European Power Electronics Conference, EPE 2011, Birmingham, UK. [10] M. Simpson and T. Markel, "Plug-in Electric Vehicle Fast Charge Station Operational Analysis with Integrated Renewables," in Conference Paper of National Renewable Energy Laboratory, Los Angeles, [11] H. Hoimoja, M. Vasilidiotis, and A. Rufer, "Power Interfaces and Storage Selection for an Ultrafast EV Charging Station," in Power Electronics, Machines and Drives, 6th IET International Conference PEMD [12] Deliverable 1.1: Specification for an enabling smart technology, European MERGE project: Mobile Energy Resources in Grids of Electricity, 24 August 2010, available on-line at: