This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON SMART GRID 1

Investigating the Impacts of Plug-in Hybrid Electric Vehicles on Power Distribution Systems
Soroush Shaee, Mahmud Fotuhi-Firuzabad, Senior Member, IEEE, and Mohammad Rastegar, Student Member, IEEE

AbstractDespite the economic and environmental advantages of plug-in hybrid electric vehicles (PHEVs), the increased utilization of PHEVs brings up new concerns for power distribution system decision makers. Impacts of PHEVs on distribution networks, although have been proven to be noticeable, have not been thoroughly investigated for future years. In this paper, a comprehensive model is proposed to study the PHEV impacts on residential distribution systems. In so doing, PHEV fundamental characteristics, i.e., PHEV battery capacity, PHEV state of charge (SOC), and PHEV energy consumption in daily trips, are accurately modeled. As some of these effective characteristics depend on vehicle owners behavior, their behavior and interests are considered in the proposed model. Also, to get a more practical model of PHEVs, the number of vehicles in a residential distribution network, the PHEV penetration level for upcoming years, distribution of PHEVs in the network, and estimation of household load growth for upcoming years are extracted from related published reports. The proposed model is applied to the IEEE 34-node test feeder, and PHEV impacts on residential distribution network are studied in different time horizons. A sensitivity analysis is also performed to demonstrate the effects of PHEV operation modes on the network load prole. Index TermsLoad growth, PHEV characteristics, PHEV impacts, plug-in hybrid electric vehicle (PHEV).

I. INTRODUCTION NTERNAL combustion engine vehicles are one of the major sources of air pollution and global warming [1]. Smart grid, as a state-of-the-art technology, realizes electrical vehicles, i.e., hybrid electric vehicle (HEV), plug-in hybrid electric vehicle (PHEV), and battery electric vehicle (BEV), as a promising solution to this issue. PHEVs are equipped with rather large batteries and are connected into the grid to be charged [2]. From the PHEV owner facet, batteries of the PHEV have to be charged overnight, so the owner can drive in the morning with a fully-charged battery. These vehicles require a large amount of electrical energy for charging their batteries and this demand of electrical power would likely have negative impacts on the distribution grid specications [3]. In order to plan an electric power delivery system, the transmission and distribution planner must know how much
Manuscript received July 29, 2012; revised October 30, 2012; accepted February 23, 2013. Paper no. TSG-00469-2012. The authors are with the Center of Excellence in Power System Management and Control, Electrical Engineering Department, Sharif University of Technology, Tehran, Iran (e-mail: soroush.shaee@gmail.com, fotuhi@sharif.edu, rastegar_m@ee.sharif.edu). Digital Object Identier 10.1109/TSG.2013.2251483

power it is expected to be served. Many distribution planners view voltage, along with loading, as the factors they must engineer in designing an acceptable distribution system. Not only voltage prole, but also power losses are essential to the distribution grid operator as well as grid customers. So, peak load, total losses, and voltage deviation are the most important specications of distribution system which should be investigated in the widespread presence of PHEVs. To evaluate the impacts of PHEVs on the distribution system, it is preliminary necessary to thoroughly explore the PHEV characteristics including battery capacity, state of charge (SOC), the amount of energy required for charging the battery, and charging level [4]. The capacity of PHEV battery depends on the type of vehicle and the driving range of vehicle in the electrical mode designated as all electric range (AER). SOC is dened as percentage of the charge remained in the battery. The amount of energy required for charging PHEV depends on the energy consumed by the vehicle in daily trips subsequently associated with daily miles driven and operation mode of PHEV, i.e., electric motor or combustion engine. Also, charging level directly affects the duration time of charging such that lower charging level increases the duration time of charging a PHEV. In conclusion, PHEV characteristics can be generally divided into two categories. The rst category relates to those characteristics of PHEV which are known based on the car manufacturer data or power system structure. The features such as battery capacity are placed in this category. The second class comprises of those properties that depend on travelling habits of the vehicle owner. Features such as daily miles driven and starting time of charging PHEV reside in this category. Thus, investigating PHEV owners behavior, along with manufacturer data, is an important key to study the effects of PHEV deployment in distribution systems. In recent years, considerable efforts have been concentrated on the investigation of the impacts of PHEV on the distribution system. In almost all of the literatures, different assumptions have been made to simplify the investigations. The authors of [5] assumed different penetration levels of PHEV for three charging periods, i.e., hour 21 until hour 6, hour 18 until hour 21, and hour 10 until hour 16. References [6] and [7] evaluated the impacts of charging PHEVs on a residential distribution network with different charging strategies and various PHEV penetration levels. The starting time of charging was assumed to be between 5 and 7 P.M. without any precise study on the vehicle owners behavior. A smart load management to control PHEV charging in a residential network was proposed in [8], [9]. They investigated the impacts of uncontrolled and

1949-3053/$31.00 2013 IEEE

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE TRANSACTIONS ON SMART GRID

controlled PHEV charging on distribution systems. Three time zones were determined for charging PHEVs, without considering vehicle owners behavior as one of the most important factors to extract substantial parameters of PHEVs. In [10], a large-scale distribution network planning model has been used for the calculation of the required network investment with different future levels of PHEV penetration. Reference [10] assumed that 85% of the PHEVs charge at off-peak hours. Another studies in [11] presented the effects of different penetration levels of PHEVs on the aging of distribution transformer and feeder losses. It considered last trip end time as charging start time while the vehicles type, AER, and the amount of energy required for charging PHEVs are not taken into account. Moreover, many researchers considered various assumptions for the size of PHEVs battery. For example, [5], [7], and [11][14] used identical battery sizes for all PHEVs. In [15], the battery capacity of each vehicle is dened as a continuous uniform random number between 6 and 25 kWh without providing any specic reason for this assumption. Since PHEV batteries are different and most vehicles would have batteries with different charging proles, and also, it is very unlikely that customers in a single area own similar type of PHEV from the same company, it is very important to consider the distribution of different types of vehicles and batteries in eets of PHEVs in the analyses. Furthermore, [5], [8][10], [16][18] assumed empty battery for vehicles at the time of arriving home. However, SOC at the time of arriving home should be determined based on daily miles driven, vehicle type, AER, and different operation modes of PHEV. It should be noted that required energy for run the PHEV in charge depleting mode is provided simultaneously by internal combustion engine and electric motor. So, different modes of PHEV operation should be considered to determine SOC of the battery and consequently required energy to have a full charge battery. In this paper, a comprehensive model for PHEV is developed taking into account various aspects of PHEV and different PHEV characteristics, i.e., battery capacity, SOC, required energy for charging PHEV, and starting time of charging. Also, the number of PHEV in a distribution network is precisely determined based on their penetration levels and number of vehicles in the distribution network. The approach presented in this paper is unique because it includes: 1) PHEV owners behavior during different weekdays and weekends of summer and winter, 2) precise determination of PHEV characteristics such as PHEV type, AER, battery capacity, and PHEV distribution prediction in the network based on their characteristics, 3) precise accounting of PHEV energy consumption in daily trips based on the battery capacity and different modes of PHEV operation, 4) accurate prediction of PHEV penetration levels in the distribution systems for the upcoming years, 5) investigation of the impacts of different PHEV penetration levels on the performance of distribution systems in short and long-term simulations, and 6) consideration of load growth along with the increase in PHEV penetration level, in the time periods of the study. To this end, different transportation and PHEV reports like New York ISO report [4], 2009 national household travel survey (2009 NHTS) [19], PNNL report [20], EPRI report [21], U.S. Department

TABLE I ECPM AND BATTERY CAPACITY FOR PHEV33 OF DIFFERENT VEHICLE TYPES

of Energy report [22], and [23] are used to thoroughly extract vehicle owners behavior, PHEV characteristics, and other required assumptions such as load growth. One of the main contributions of this paper is proposing a comprehensive model to study the PHEV impacts on distribution system load demand for upcoming years. This helps the distribution network planners to have a precise prediction of the future load demand in order to implement an optimal scheduling for short and long term planning. This model shall be applied to 34 node IEEE distribution system to evaluate and discuss PHEV impacts on peak load demand, voltage deviation, and total power losses in different cases. Also, a sensitivity analysis is performed to investigate the impacts of different PHEV modes of operation on the network load prole. The rest of the paper is outlined as follows. A summary of PHEV characteristics and related assumptions for studying their impacts are presented in Section II. In Section III, the PHEV characteristics are extracted to approach precise assumptions for PHEV specications in Section IV. Simulation results are presented and discussed in Section V. The paper is concluded in Section VI. II. PHEV CHARACTERISTICSAND ASSUMPTIONS New generation of vehicles such as PHEVs, which have an electric motor together with an internal combustion engine, are recently becoming more popular and would broadly be seen on the roads in the future. As noted earlier, the PHEV characteristics should be specically studied before investigating their impacts on the distribution system. It is assumed that PHEVs are charged to their capacity once per day after their last trip arrival time. In this section, the PHEV characteristics and related assumptions required to study the impacts of PHEV on distribution system are described and analyzed through the following subsections. A. PHEV Battery Capacity The capacity of battery is the key factor to determine the number of miles driven by a PHEV in the electric mode. The Pacic Northwest National Laboratory (PNNL) has estimated the electrical energy consumption per mile (ECPM) and battery capacity for a PHEV33 of different vehicle types (PHEVx indicates a PHEV with ) [20], as shown in Table I. As noted earlier, AER is the possible distance driven by a PHEV with a full charge battery. So the product of ECPM and AER results in the battery capacity. (1) where is the usable capacity of PHEV battery.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
SHAFIEE et al.: INVESTIGATING THE IMPACTS OF PLUG-IN HYBRID ELECTRIC VEHICLES ON POWER DISTRIBUTION SYSTEMS 3

TABLE II SIZE OF BATTERY FOR VARIOUS PHEVS (KWH)

Table II shows the battery capacity for PHEVs with AER of 30, 40, and 60 miles according to (1). Referring to this table, PHEVs have a wide range of battery size, from 7.8 kWh to 27.6 kWh. Since the energy required for charging a PHEV depends on the battery capacity, it is necessary to determine the types and AERs of PHEVs to analyze their impacts. B. State of Charge (SOC) SOC is a measure of the amount of energy stored in a battery. It is similar to the fuel gauge in conventional internal combustion cars. In this paper, SOC refers to the percentage of energy remained in the battery when PHEV arrives home, after daily trips. Although many papers, [17], [18], and [24], [25], assumed that a PHEV travels in the electric mode until the battery becomes empty and then switches to the charge sustaining mode, the electric motor and internal combustion engine are optimally working together. So, PHEV can operate in blended mode in which the internal combustion engine helps the electric motor provide the required energy to run the vehicle. Therefore, in charge depleting mode, either the entire or a fraction of the required energy is supplied by the battery. To cover all the possible PHEV operations, a factor for each PHEV is dened. This factor represents the percentage of distance that PHEV drives in the electric mode. Accordingly, the SOC of a PHEV would be (2) is the distance driven in where is the total driven distance, the electric mode. The battery will be empty if reaches AER. C. Required Energy to Charge PHEVs Battery The amount of energy required to charge the PHEV battery after the last trip arrival time depends on the SOC. The energy is drawn from the utility grid to fully charge the battery and is expressed mathematically by (3) is the required energy to fully charge the battery. where represents the chemical energy consumed in the battery. The actual electrical energy is calculated by (4). (4) where is the efciency of charging PHEV battery and is the actual energy which should be transferred from the grid to

charge the battery. In [5], [11], and [15] the efciency is assumed to be 88%, 90%, and 90%, respectively. For further clarication, consider a PHEV with mid-size SUV type and AER equal to 40 miles. So, referring to Table II, battery capacity would be 15.2 kWh. Assume that this vehicle drives 32 miles in a day with . Thus, SOC is calculated by (2) as and would be 9.73 kWh using (3). Assume that the efciency of the charger is 88%. Thus, would accordingly be . D. Charging Level Due to various power system structures and different charging stations, so far a number of charging levels to charge PHEV battery have been considered. For instance, in [5], charging level is assumed to be 4 kW for all PHEV based on Belgian standard outlet (230 V, 4.6 kW). Reference [7] has considered 120 V/15 A and 240 V/50 A, respectively, as normal and rapid charging levels. The ORNL report [26] assumed two charging scenarios; the rst one is 120 VAC circuit supplied through a 15 A circuit breaker charged at a peak rate of 1.4 kW, and the other one is 240 VAC with a 30 A circuit breaker charged at a peak rate of 6 kW. In [10], a more reasonable assumption was taken into account for charging level in which different charging rates are specied by a factor of the battery capacity (C), i.e., 0.2C, 1C, and 2C. Accordingly, 0.2C is named slow charging rate at which it takes 5 hours to reach the full charge battery when the battery is initially empty. E. Load Growth Most utilities represent their customer demand on a class by class basis, using smooth, and 24-hour peak load curves [27]. These curves represent the demand characteristics of customers in each class, i.e., residential, commercial, and industrial. With the presence of PHEVs, the total load of a residential distribution network includes the daily household loads and the charging of PHEVs. To investigate the impacts of PHEVs, especially on the load curve, it is important to account the level of load growth during the time period of study. The system load increases as the new customers are added to the grid or the existing customers add new appliances or replace their existing equipments with devices that require more power. In conclusion, energy usage within an electric utility system grows for two reasons; new customer additions and new uses of electricity [22]. These two factors should be predicted to study PHEV impacts on the load curve in different years. F. PHEV Distribution in the Network To plan the future distribution network with the widespread presence of PHEVs, it is essential to have a precise prediction of PHEV distribution in the network. At rst, the number of vehicles, whether PHEV or not, in the network should be estimated. Then, based on the specic PHEV penetration level, the number of PHEVs is determined. Distribution system operator would have no control over the location of future PHEV charging points, and no direct control over the duration and frequency of PHEV charging. Thus, PHEV is randomly distributed

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4 IEEE TRANSACTIONS ON SMART GRID

along the houses based on the number of PHEVs in a given year. As mentioned before, periods of charging are related to the PHEV owners interest. Obviously, PHEV characteristics should be determined to study the impacts of PHEVs on the distribution systems.

B. VEHV2PUB.csv This le contains information about 309164 vehicles. Each vehicle has 92 attributes. Four of them, i.e., VEHTYPE, VEHID, HOUSEID, and BESTMILE, are utilized here. The rst three attributes are the same as those in the former le. BESTMILE indicates the best estimation of annual miles that each vehicle drives [28]. So, daily mileage is calculated by dividing BESTMILE by 365. In order to develop BESTMILE, different items including: weekday and weekend terms, self-reported vehicle mile travel (ANNMILE/365), age and education level of the respondent, vehicle age class and its type, size, and life cycle of the house, size of the area where the house located in, and the gender of the primary driver are involved [28]. The authors of [24] used ANNMILE instead of BESTMILE to calculate annual miles driven which likely leads to inaccurate results. However, as mentioned before, to calculate BESTMILE, many other attributes are considered in addition to ANNMILE and thus, BESTMILE sounds more accurate than ANNMILE. In conclusion, DAYV2PUB indicates trip characteristics and VEHV2PUB determines vehicles characteristics. In order to achieve PHEV characteristics, it is necessary to know when each trip ends and also how many miles are driven by the vehicle in a day. Thus, these two les should be combined to yield the useful data. HOUSEID, VEHID, and VEHTYPE are common in both les and are used to combine the les. Thus, a database is approached which indicates about 163 000 vehicle trips and each vehicle trip has these attributes: VEHTYPE, BESTMILE, ENDTIME, TRAVDAY and TDAYDATE. It is worthwhile to note that it is necessary to accurately determine the starting time of charging vehicles in macro and micro scale due to horizon time of study. In macro scale, the year is divided into summer and winter [5], [6], [26] and [29], while in micro scale vehicle owners driving pattern in weekdays and weekends is investigated. Here, macro and micro scale studies are approached together. To this end, TRAVDAY and TRAVDATE in recently created database are used to create four les for weekdays and weekends of the summer and winter.

III. STUDYING 2009 NHTS Transportation reports are the best sources of information about vehicle and trip characteristics. The data which are not related to the owners interest are accessible in many references and reports. Extracting characteristics which depend on the PHEV owners behavior are generally more complicated. In this study, 2009 NHTS database is used to extract these characteristics. This database can be easily obtained by a eld research based on designed questionnaire for residential customers in each country. NHTS provides comprehensive data on travel and transportation patterns in the United States. Detail transportation data from 1995 to 2009 can be found in NHTS website [19]. In [24], the authors used 2001 NHTS to extract the required data. It should be noted that, although NHTS has been studied before, a comprehensive and accurate analysis has not been yet prepared to extract the reliable data for PHEV characteristics. In this section NHTS is exactly deliberated and the associated data from the report are extracted and explained in details. The 2009 NHTS contains data associated with 150147 houses. These data are collected from all areas of the country and include household, vehicle owners, vehicle, and daily trip level data. The survey has been conducted from March 2008 through May 2009. Data collected for daily trips include trip purpose, mode of transportation used, duration of the trip, time of day the trip took place, travel day, and the data related to a private vehicle trip such as the number of people in the vehicle, driver characteristics, and vehicle attributes. From the databases of 2009 NHTS, two Microsoft Excel les 1) DAYV2PUB.csv and 2) VEHV2PUB.csv are used to investigate vehicle owners behavior and extract valuable vehicle characteristics. In the following, these two les are separately explained and then their relationship is claried. A. DAYV2PUB.csv This le contains information on almost 1041000 trips. Every trip has 150 attributes. In our study, nine attributes, household ID number (HOUSEID), person ID number (PERSON ID), travel day trip number (TDTRPNUM), vehicle ID number (VEHID), type of vehicle (VEHTYPE), travel day trip end time (ENDTIME), travel day of week (TRAVDAY), travel date (TDAYDATE), and the number of house vehicles (HHVEHCNT), are used. HOUSEID, PERSON ID and TDTRPNUM determine trip characteristics taken by a particular member of a specic house. VEHID and VEHTYPE indicate the specic vehicle of a house used in a particular trip and also determine the type of that vehicle. ENDTIME indicates the time in which the trip ends, and TRAVDAY and TDAYDATE determine the date of trip.

IV. DETERMINING PHEV CHARACTERISTICS AND ASSUMPTIONS FOR INVESTIGATING PHEV IMPACTS PHEV characteristics, depended on vehicle owners behavior, are extracted from 2009 NHTS by analyzing four created les. These data include daily miles driven, starting time of charging, number of vehicles per house, and vehicle type of houses. References [6], [11], and [30] assumed that the PHEV owners plug in their vehicles when they arrive home after their last trip in a day. Therefore, the last trip arrival time of the vehicles can be the starting time of charging and is considered here. Other associated reports and references are used to determine AER, PHEV penetration level, daily load prole, and residential load growth. All of these data are required to thoroughly model and analyze PHEV impacts on the distribution system.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
SHAFIEE et al.: INVESTIGATING THE IMPACTS OF PLUG-IN HYBRID ELECTRIC VEHICLES ON POWER DISTRIBUTION SYSTEMS 5

Fig. 1. Percentage of vehicles versus daily miles driven for the weekdays of summer. Fig. 3. Percentage of vehicles versus their home arrival time for the weekends of summer. TABLE III PERCENTAGE OF HOUSES WITH DIFFERENT NUMBER OF VEHICLES

Fig. 2. Percentage of vehicles versus their home arrival time for the weekdays of summer.

A. Daily Miles Driven As mentioned before, SOC of PHEVs and the energy required to charge the battery subsequently depend on the miles driven by vehicles. Four gures can be sketched from the BESTMILE attribute of each vehicle drives [28], in the four created les. For instance, Fig. 1 represents individual and cumulative distribution of vehicles on the basis of their mileage in weekdays of summer. Individual bar curves in Fig. 1 indicate that the common mileage of vehicles is in the range of 2025 miles. It can also be seen from this gure that about 55% of the vehicles drive less than 30 miles per day. B. Last Trip Arrival Time of Vehicles Vehicles can be classied based on their last trip arrival time at weekdays and weekends of summer and winter, using ENDTIME attribute [28], in the four created les. Figs. 2 and 3, respectively, show the percentage of vehicle home arrival time at weekdays and weekend of summer. It is clear that there is a noticeable difference between peoples driving pattern at weekdays and weekends. At weekdays, majority of people come home between 16:00 and 21:00, after their working hours. On the other hand, at weekends, people often stay home or go shopping without regular scheduling. C. Number of Vehicles Per House As the PHEV penetration level is one of the most important factors affecting the impact of PHEV on distribution system, estimation of the number of PHEV distributed in houses is important. Investigation of HOUSEID and HHVEHCNT attributes from DAYV2PUB.csv le [28] results in the percentage of houses with different number of vehicles. It is worthwhile to

TABLE IV PERCENTAGE OF EACH TYPE OF VEHICLE IN NHTS 2009

compare the number of vehicles reported in 2009 NHTS and 2001 NHTS report, as shown in Table III. This table certies that the percentage of houses which have a specic number of vehicles does not change signicantly from 2001 to 2009. Referring to Table III, the average number of vehicles per house in 2001 and 2009 is 2.12. This result is close to what was claimed in [7] for Washington State, i.e., in average 2.3 vehicles per house. D. Vehicles Type Analysis As emphasized before, type of PHEV has a direct impact on the size of battery as well as energy consumed by the PHEV. The market share of each type of vehicle is extracted from type of vehicle attribute from the VEHV2PUB.csv as shown in Table IV. It can be realized from Table IV that the most common vehicle is type 1, i.e., compact sedan. E. AER 2009 NHTS does not have data on AER of PHEVs, since the purpose of this survey is not to study the electric energy issue and there was not any PHEV in the market at the time of publishing the survey. The penetration of PHEVs with different AERs is assumed to be the same as Table V [10]. AER of PHEVs is determined randomly based on Table V. Based on the type and AER of each PHEV, the size of battery is extracted from Table II.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6 IEEE TRANSACTIONS ON SMART GRID

TABLE V PERCENTAGE OF PHEVS WITH DIFFERENT AERS

Fig. 6. Number of houses till 2050 in U.S..

Fig. 4. Medium penetration level of PHEV in the distribution system.

Fig. 7. IEEE 34-node test feeder [34].

Fig. 5. An average household load curve in summer and winter.

F. PHEV Penetration Level The number of PHEVs depends on the PHEV penetration level. According to [21], PHEV penetration level is estimated in three levels: low, medium, and high. The medium penetration level between 2010 and 2050 is depicted in Fig. 4 Accordingly, in medium PHEV penetration level, 35% of total vehicles will be PHEVs by 2020. PHEV penetration rate increases up to 54% and 62% by 2035 and 2050, respectively. It should be noted that the residential load is signicantly grown along with the increment of PHEV penetration levels during these years. G. Daily Load Prole Authors in [31] took available data from [32] and [33] to generate an average household load prole for summer and winter of 2010. The load prole covers 24 hours on a 15-min time basis as shown in Fig. 5 for an arbitrary day during summer and winter. These load proles, assigned to each house, are used in the PHEV impacts studies. H. Load Growth Load growth is also considered in studies conducted in this paper. According to [22], the mean incremental growth rate of annual residential electricity consumption will be 1.3%, due to new uses of electricity. The number of houses is assumed to increase similar to Fig. 6 [23]. Note that the daily load prole determined in the previous subsection belongs to 2010; so the load growth for future years is calculated based on that of 2010. V. NUMERICAL RESULTS This section represents the impacts of PHEVs on different terms of distribution system including system power consumption, total system losses, and voltage deviation at load points. In

the previous section, the PHEV characteristics and related assumptions were precisely extracted from the accessible reports. In this section, the impacts of charging PHEVs for the weekdays of winter and summer are explored. Since the load proles have lower peak and average at weekends than weekdays, PHEV impacts are just studied at weekdays. The proposed approach is general and is not case dependent. The following assumptions are just sample cases and are either reasonably justied or extracted from available reports and publications. A. Description of System Under Study The IEEE 34-node test feeder [34] is studied here. This radial network is shown in Fig. 7. The network medium voltage is 24.9 KV and low voltage is 230 V. So, this network is represented as a residential radial network. I The network contains 33 load points. Eight points, 810, 818, 820, 822, 826, 838, 856 and 864, are single phase and the remaining 25 are three phase. Two houses are assigned to each phase of load points [16]. Therefore, the total number of houses is equal to 166. According to Table III, the number of vehicles in the mentioned radial network is equal to 352. These vehicles are assigned randomly to the houses based on the percentage noted in Table III. From Table IV, the type of these vehicles is determined randomly. Daily miles driven and last trip arrival time of vehicles for the weekdays of summer and winter are extracted from presented data of Sections IV-A and IV-B and to determine when a vehicle is plugged into the outlet and how many miles are driven by each vehicle. These data are subsequently determined for PHEVs. Although the distribution of factor , due to the lack of data, cannot be accurately predicted, it is reasonable to assume that more than half of the daily trips are taken on the electric mode [35]. This leads to have , a random number between 0.5 and 1. So, daily miles driven, type of PHEV, AER, and factor are incorporated to calculate the energy consumption of the PHEV. Household load proles are assumed to be the same as the ones presented in the previous section. Due to impacts of reactive power on lines congestion, power losses, and voltage magnitudes, the reactive power consumption of household load and PHEV battery should be considered. To this end, according

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
SHAFIEE et al.: INVESTIGATING THE IMPACTS OF PLUG-IN HYBRID ELECTRIC VEHICLES ON POWER DISTRIBUTION SYSTEMS 7

to [36] and [37], the average power factor of a typical household and PHEV battery are respectively assumed to be 0.9 and 0.95. The charging level of PHEVs should also be determined. As mentioned before, it is more realistic to consider charging rate as a coefcient of the battery capacity rather than using a denite charging rate. So, the charging rate of a battery with large capacity is reasonably higher than a lower capacity battery. The charging level of 0.2C (C represents the usable capacity of the battery) is considered here, which means that an empty battery will be fully charged in about 5 hours [10]. Referring to Table II, the lower and upper bands of charging rate are and , respectively. This range of charging level is commonly available in most single-phase 230 V outlets in residential households without having to reinforce the wiring. The fast charging is not considered, because it requires a higher voltage connection and is not already available in houses. Also, dc/ac and ac/dc conversion efciency for the battery is assumed to be 88% [21]. Load ow analysis is performed to calculate the total load, feeder power loss, and voltage deviation in different load points of the network. In this section, load ow analysis is executed by open-source distribution system simulator (OpenDSS). OpenDSS is an electric power distribution system simulator for advanced analysis of distribution systems. OpenDSS is a multi-phase simulator tool in which each phase of load points can be specied separately. Also, power ow is simulated sequentially over successive time steps (e.g., 15 min) considering all circuit dynamic facilities (e.g., regulators). These OpenDSS capabilities provide interactions of the variations of PHEV load patterns and seasonal conventional load [38]. B. Case Studies Denition The following case studies are conducted to investigate the impacts of PHEV: Case I: single year study; this case focused on determining the effects of different PHEV penetration levels on the performance of distribution system, i.e., voltage deviation, system losses, and peak to average ratio (PAR), in a single year, 2020. Case II short-term study; The impacts of PHEV on distribution system performance within the period of 2020 and 2026 are investigated. Case III: long-term study; this case is devoted to the investigation of the impacts of PHEV between 2020 and 2050, just, on the load curve. As mentioned before, some of the PHEV characteristics are randomly distributed. So, for a more accurate analysis, for each scenario study, about 3000 samples are taken over these parameters to achieve a precise average for PHEV consumption in 15 min time interval. In the following subsections, the mentioned cases are deeply studied. C. Case I This investigation is effectuated in year 2020, in which the low, medium, and high penetration levels are respectively 11.3%, 35% and 45% [21]. The load of each house grows 13.8% compare to that of year 2010.

Fig. 8. Impacts of PHEV charging on total load curve for different PHEV penetration levels in summer of 2020.

Fig. 9. Impacts of PHEV charging on total losses for different PHEV penetration levels in summer of 2020.

Fig. 10. Impacts of PHEV charging on total load curve for different PHEV penetration levels in winter of 2020.

Fig. 11. Impacts of PHEV charging on total losses for different PHEV penetration levels in winter of 2020.

The number of houses is assumed to be 166, as mentioned before. The case with no PHEV is taken as a reference case. Figs. 811 and Table VI show the results of PHEV charging impacts on the distribution system in summer and winter. Table VI illustrates various specications of the grid with different PHEV penetration levels. The rst two rows of Table VI present the impacts of PHEVs on the peak load. Table VI demonstrates that growing PHEV penetration level leads to a signicant peak load increment, since majority of people arrive home after their last daily trip between 16 and 21 and plug-in their vehicles at this time which coincides with the households peak load. These facts are also be seen in Figs. 8 and 10 which show the feeder load curve in summer and winter of 2020, respectively. The results shown in Figs. 8 and 10 imply that load demand with PHEVs does not change noticeably from 3:00 to 11:00 compared to the reference load curve. This point shows that no

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8 IEEE TRANSACTIONS ON SMART GRID

TABLE VI GRID SPECIFICATIONS WITH LOW, MEDIUM, AND HIGH PENETRATION LEVELS OF PHEV IN 2020

Fig. 12. Impacts of PHEV charging on total load curve for medium PHEV penetration levels in summer of different years.

PHEV charges when the load level is low. So, although the average load increases due to PHEV charging, the peak load increases more than that and it causes PAR increment as presented in Table VI. It is desirable to have PAR close to one. Higher PAR might result in substantial cost increment for utilities in long-term since it requires new investment in generation and transmission capacities to serve higher peak load. Also, utilities take disadvantages of peak increment in short-term due to the operation cost rising. Unlike peak load, PAR is lower in winter than summer. The reason for this is that, load curve in winter is smoother than summer, as shown in Figs. 8 and 10. Figs. 9 and 11 and the third row of Table VI show total system losses during a day in summer and winter of 2020 with different PHEV penetration levels. The results verify the loss increment due to PHEV penetration level growth. It is worthwhile to draw a comparison between PHEVs power consumption, which can be concluded from Figs. 8 and 10, and loss variation curves sketched in Figs. 9 and 11. It can be deduced that charging PHEVs takes place simultaneously with the peak of loss curve. This occurrence results in the increment of loss contribution in the peak load. With regards to the fourth row of Table VI, increasing the number of PHEVs also causes an increase in voltage deviation; 45% PHEV penetration leads to more than 25% increment in the maximum voltage deviation, for both summer and winter, compared to the reference case. However, this deviation can likely be limited to its cap, using voltage regulators in the network. D. Case II In this case, three denite years are selected to probe the impacts of medium PHEV penetration level between 2020 and 2026: Year 2020: PHEV penetration level is 35% and the load prole of each house grows 13.8% compared to that of 2010 shown in Fig. 5. Year 2023: PHEV penetration level is 47% and the load prole of each house grows 18.3% compared to that of 2010. Year 2026: PHEV penetration level is 50% and the load prole of each house grows 23% compared to that of 2010. It is assumed that the topology of the test feeder is not changed during these years. However, the number of houses increases according to Fig. 6 and these houses are randomly

Fig. 13. Impacts of PHEV charging on total load curve for medium PHEV penetration levels in winter. TABLE VII SPECIFICATIONS FOR SHORT-TERM SIMULATION IN SUMMER AND WINTER

distributed among the existing load points. Simulation results for the three mentioned years are presented in Table VII, Figs. 12 and 13. The cases without PHEV are the reference cases for each year. The gures and table bring miscellaneous impacts of PHEV charging on the 34-node test feeder in a typical weekday of summer and winter. Table VII presents the peak load, PAR, total losses, and maximum voltage deviation of load points for summer and winter. Figs. 12 and 13 depict the feeder load curve including losses, with and without PHEV for short-term study in summer and winter, respectively. The following results can be concluded: The second and third rows of Table VII show, respectively, the peak load and PAR of the network for different years. The peak load and PAR increment, compared to the reference cases, are higher in 2023 than 2020 and do not change signicantly from 2023 to 2026. The reason for this is that, the growth rate of PHEV penetration level is higher within 20202023 than that of period 20232026 and majority of PHEVs charge at the time of household peak load. These can be seen from Figs. 12 and 13. The fourth row of Table VII describes the ratio of total power losses to total load in the reference case of the same

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
SHAFIEE et al.: INVESTIGATING THE IMPACTS OF PLUG-IN HYBRID ELECTRIC VEHICLES ON POWER DISTRIBUTION SYSTEMS 9

TABLE VIII LOAD VERSUS DIFFERENT VALUES OF FACTOR

Fig. 14. Impact of PHEV charging on peak load for long-term investigation during 2020 and 2050 in summer and winter.

Fig. 15. Impact of PHEV charging on PAR for long-term investigation between 2020 and 2050 in summer and winter.

operation modes on the network load prole. Doing so, different values of , i.e., 0.5, 0.6, 0.7, 0.8, 0.9, and 1, are considered for all the PHEVs in a single year, 2020, with different PHEV penetration levels. Table VIII illustrates the impacts on peak load of factor . Higher implies that PHEVs drive more on electric mode, so the battery of PHEVs would likely require more power from the grid to be fully charged. Consequently, as presented in Table VIII, peak load increases on the basis of increment. Also, a higher penetration level of PHEV leads to a more increment in peak load versus growth. VI. CONCLUSION Proliferation of PHEVs in distribution systems requires a deep study on their impacts. This study should be effectuated on the specications of distribution systems such as load curve, feeder losses, and voltage of load points. In this paper, a comprehensive investigation of PHEV impacts was made on distribution systems. To this end, the PHEV characteristics and other required assumptions were rstly extracted from available travelling surveys and reports, and then analyzed to achieve a valuable model for the investigation of PHEV impacts. These data were applied to a test distribution system in winter and summer in short and long-term simulation, considering load growth. In this study, the average annual load growth was utilized to incorporate the impacts of load growth along with increase in PHEV penetration level. The prole of increased load curve would likely be affected by the types of new appliances and usage pattern. This issue, however, is not in the scope of current paper. The results of PHEV impacts investigation veried that, the voltage deviation is not a sophisticated issue in the distribution systems, while peak load and loss increment are both the big concern to the widespread use of PHEVs in distribution systems due to coincidence of daily peak load and charging time of PHEVs. The represented results indicate that the problem is more complicated for winter days due to higher average load. The feeder was also studied under different penetration levels of PHEVs in year 2020. In addition to PAR and loss increment due to PHEV penetration growth, the study certied a noticeable likeness between the PHEV charging and total loss curve. It sounds essential to control the time of charging PHEV to prevent the distribution congestion and feeder loss increment. This issue as an open research area is currently under our study. ACKNOWLEDGMENT The authors would like to thank the Iranian National Science Foundation with gratitude for their nancial support.

year. With the PHEV penetration level increment, power loss rises signicantly. The last row of Table VII shows the maximum voltage deviation among the nodes during 24 hours. The results certify that PHEV charging causes an increment in the voltage deviation of load points. Also, the increment of PHEV number leads to a more voltage deviation. The table also shows that the voltage deviation is higher in winter than in summer due to the higher load demand of winter. E. Case III This case provides important information about peak load demand in a long-term investigation between 2020 and 2050, notwithstanding that the feeder structure is changed during these years. The PHEV penetration levels during these years are simply extracted from Fig. 4. Also, based on the previous mentioned data, the number of houses increases 27.5% and load of each house grows 47.3% from 2020 to 2050. The results are depicted in Figs. 14 and 15 for summer and winter weekdays. As expected, the peak load increases due to the PHEV presence and PHEV penetration level increment. The presented results indicate that the PHEV charging coincides with the peak of household energy consumption. The results shown in Fig. 15, in both summer and winter curves, represent that PAR increases before year 2026, since in this period, the rate of PHEV penetration level is high, and PHEV charging is coinciding with peak load curve. This occurrence causes the peak load increment to be more than average load increment. Obviously, PAR lessens in the next years due to the higher increment of average load than peak load. F. Sensitivity Analysis The distribution of would signicantly affect the consequent result of the load curve. As mentioned earlier, the distribution of factor cannot be accurately predicted. Therefore, a sensitivity analysis on the impacts of different values of is constructed here to investigate the impacts of all the possible PHEV

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
10 IEEE TRANSACTIONS ON SMART GRID

REFERENCES
[1] U.S. Department of Transportation [Online]. Available: http://climate. dot.gov/about/transportations-role/overview.html [2] S. Vazquez, S.M. Lukic, E. Galvan, L.G. Franquelo, and J.M. Carrasco, Energy storage systems for transport and grid applications, IEEE Trans. Ind. Electron., vol. 57, pp. 38813895, 2010. [3] M. Rastegar, M. F. Fotuhi, and F. Aminifar, Load commitment in a smart home, Elsevier J. Applied Energy, vol. 96, pp. 4554, 2012. [4] N. Y. ISO, Alternate Route: Electrifying the transportation sector, Tech. Rep., 2009. [5] K. Clement-Nyns, E. Haesen, and J. Driesen, The impact of charging plug-in hybrid electric vehicles on a residential distribution grid, IEEE Trans. Power Syst., vol. 25, no. 1, pp. 371380, Feb. 2010. [6] S. Shao, M. Pipattanasomporn, and S. Rahman, Challenges of PHEV penetration to the residential distribution network, in Proc. IEEE PES Gen. Meet., 2009. [7] K. Schneider, C. Gerkensmeyer, M. Kintner-Meyer, and R. Fletcher, Impact assessment of plug-in hybrid vehicles on Pacic Northwest distribution systems, in Proc. IEEE Power Energy Soc. Gen. Meet. Conv. Del. Elect. Energy 21st Century, Pittsburgh, PA, USA, Jul. 2008. [8] S. Deilami, A. S. Masoum, P. S. Moses, and M. A. S. Masoum, Realtime coordination of plug-in electric vehicle charging in smart grids to minimize power losses and improve voltage prole, IEEE Trans. Smart Grid, vol. 2, no. 3, pp. 456467, Sep. 2011. [9] A. S. Masoum, S. Deilami, P. S. Moses, M. A. S. Masoum, and A. Abu-Siada, Smart load management of plug-in electric vehicles in distribution and residential networks with charging stations for peak shaving and loss minimization considering voltage regulation, IET Proc. Gener., Transm., Distrib., vol. 5, no. 8, pp. 877888, 2011. [10] L. P. Fernndez, T. G. S. Romn, R. Cossent, C. M. Domingo, and P. Fras, Assessment of the impact of lug-in electric vehicles on distribution networks, IEEE Trans. Power Syst., vol. 26, no. 1, pp. 206213, Feb. 2011. [11] J. Taylor, A. Maitra, M. Alexander, D. Brooks, and M. Duvall, Evaluation of the impact of plug-in electric vehicle loading on distribution system operations, in Proc. IEEE Power Energy Soc. Gen. Meet., 2009, pp. 16. [12] K. Clement-Nyns, E. Haesen, and J. Driesen, The impact of vehicle-to-grid on the distribution grid, Elsevier J. Elect. Power Syst. Res., vol. 81, no. 1, pp. 185192, Jan. 2011. [13] M. J. Rutherford and V. Yousefzadeh, The impact of electric vehicle battery charging on distribution transformers, in Proc. 26th Annu. IEEE Appl. Power Electron. Conf. Expo. (APEC), Mar. 611, 2011, pp. 396400. [14] S. M. M. Agah and A. Abbasi, The impact of charging plug-in hybrid electric vehicles on residential distribution transformers, in Proc. 2nd Iranian Conf. Smart Grids (ICSG), , May 2425, 2012, pp. 15. [15] W. Su and M. Y. Chow, Performance evaluation of an eda-based large-scale plug-in hybrid electric vehicle charging algorithm, IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 308315, Mar. 2012. [16] E. Sortomme, M. M. Hindi, S. D. J. MacPherson, and S. S. Venkata, Coordinated charging of plug-in hybrid electric vehicles to minimize distribution system losses, IEEE Trans. Smart Grid, vol. 2, no. 1, pp. 198205, Mar. 2011. [17] M. Kintner-Meyer, K. Schneider, and R. Pratt, Impacts assessment of plug-in hybrid vehicles on electric utilities and regional U.S. power grids. Part 1: technical analysis, J. EUEC, vol. 1, 2007 [Online]. Available: http://www.euec.com/content/EUEC-2007.aspx [18] A. Hajimiragha, C. A. Caizares, M. W. Fowler, and A. Elkamel, Optimal transition to plug-in hybrid electric vehicles in ontario, canada, considering the electricity-grid limitations, IEEE Trans. Ind. Electron., vol. 57, no. 2, pp. 690701, Feb. 2010. [19] National Household Travel Survey [Online]. Available: http://nhts. ornl.gov [20] M. Kintner-Meyer, K. Schneider, and R. Pratt, Impacts assessment of plug-in hybrid vehicles on electric utilities and regional U.S. power grids part 1: Technical analysis, PNNL Rep., Nov. 2007 [Online]. Available: http://www.pnl.gov/energy/eed/etd/pdfs/phev_feasibility_analysis_combined.pdf [21] Environmental assessment of plug-in hybrid electric vehicles, Nationwide Greenhouse Gas Emissions, 1015325, Fin. Rep., EPRI and NRDC, vol. 1, pp. 156, 2007. [22] Energy Information Administration, U.S. Department of Energy, Annual energy outlook 2007 with projections to 2030, Feb. 2007 [Online]. Available: ftp://tonto.eia.doe.gov/forecasting/03832007.pdf

[23] Y. Zeng, K. C. Land, Z. Wang, and D. Gu, U.S. Family household momentum and dynamics: An extension and application of the profamy method, Population Res. Policy Rev., vol. 25, no. 1, pp. 141, 2006. [24] Z. Darabi and M. Ferdowsi, Aggregated impact of plug-in hybrid electric vehicles on electricity demand prole, IEEE Trans. Sustainable Energy, vol. 2, no. 4, pp. 501508, Oct. 2011. [25] H. Turker, S. Bacha, D. Chatroux, and A. Hably, Low-voltage transformer loss-of-Life assessments for a high penetration of plug-in hybrid electric vehicles (PHEVs), IEEE Trans. Power Del., vol. 27, no. 3, pp. 13231331, Jul. 2012. [26] S. W. Hadley, Impact of plug-in hybrid vehicles on the electric grid, Oak Ridge National Laboratory, TN, USA, Tech. Rep. ORNL/TM2006/554, 2006. [27] H. L. Willis, Power Distribution Planning Reference Book Revised and Expanded, 2nd ed. New York: Marcel Dekker, Inc., 2004. [28] 2009 National Household Travel Survey Users Guide (NHTS), 2011. [29] G. Putrus, P. Suwanapingkarl, D. Johnston, E. Bentley, and M. Narayana, Impact of electric vehicles on power distribution networks, in Proc. IEEE Veh. Power Propulsion Conf., Sep. 2009, pp. 827831. [30] C. Camus, C. M. Silva, T. L. Farias, and J. Esteves, Impact of plug-in hybrid electric vehicles in the portuguese electric utility system, in Proc. IEEE Power Eng., Energy, Elect. Drives Conf., 2009, pp. 285290. [31] S. Shao, T. Zhang, M. Pipattanasomporn, and S. Rahman, Impact of TOU rates on distribution load shapes in a smart grid with PHEV penetration, in Proc. IEEE PES Transm. Distrib. Conf. Expo., Apr. 1922, 2010. [32] RELOAD Database Documentation and Evaluation and Use in NEMS [Online]. Available: http://www.onlocationinc.com/LoadShapesReload2001.pdf. [33] The National Energy Modeling System: An Overview 2003 [Online]. Available: http://www.eia.doe.gov/oiaf/aeo/overview/. [34] Radial Test FeedersIEEE Distribution System Analysis Subcommittee [Online]. Available: http://www.ewh.ieee.org/soc/pes/dsacom/ testfeeders/index.html [35] A. Karnama, Analysis of integration of plug-in hybrid electric vehicles in the distribution grid, Masters thesis, Stockholm, Sweden, 2009. [36] Consumer report [Online]. Available: http://news.consumerreports.org/home/2009/07/power-factor-kilowatt-hours-volt-amperes-edison-electric-institute-compact-uorescent lightbulbs.html [37] L. Kelly, Probabilistic modeling of plug-in hybrid electric vehicle impacts on distribution networks in British Columbia, Masters thesis, University of Waterloo, Waterloo, ON, Canada, 2009. [38] OpenDSS [Online]. Available: http://sourceforge.net/projects/electricdss/.

Soroush Shaee received the B.S. and M.Sc. degrees in electrical engineering from Sharif University of Technology, Tehran, Iran, in 2010 and 2012, respectively. His research interests include power system reliability, power market, smart grid, and PHEV operations.

Mahmud Fotuhi-Firuzabad (SM99) received the B.Sc. and M.Sc. degrees in electrical engineering from Sharif University of Technology, Iran, and University of Tehran, Iran, in 1986 and 1989, respectively, and the M.Sc. and Ph.D. degrees in electrical engineering from the University of Saskatchewan, Canada, in 1993 and 1997 respectively. He is presently a professor and Head of the Electrical Engineering Department at Sharif University of Technology, Tehran, Iran. Dr. Fotuhi-Firuzabad is a member of the center of excellence in power system control and management. He serves as an Editor for the IEEE TRANSACTIONS ON SMART GRID and as the Guest Editor-in-chief of a Special Issue (Microgrids) of the IEEE TRANSACTIONS ON SMART GRID.

Mohammad Rastegar (S12) received the B.S. and M.Sc. degrees from Electrical Engineering at Sharif University of Technology, Tehran, Iran, in 2009 and 2011, respectively. He is currently pursuing the Ph.D. degree in the Electrical Engineering Department at Sharif University of Technology, Tehran. His research interests include power system reliability, demand response, smart home, and smart grid.