Cost analysis of plug-in hybrid electric vehicles using GPS-based longitudinal travel data

Energy Policy 68 (2014) 206–217

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Cost analysis of plug-in hybrid electric vehicles using GPS-based longitudinal travel data Xing Wu a,n, Jing Dong b, Zhenhong Lin c a

Lamar University, Department of Civil Engineering, 4400 MLK Boulevard, Beaumont, TX 77710, USA Iowa State University, Department of Civil, Construction and Environmental Engineering, 350 Town Engineering Building, Ames, IA 50011, USA c Oak Ridge National Laboratory, National Transportation Research Center, 2360 Cherahala Boulevard, Knoxville, TN 37932, USA b

H I G H L I G H T S








A spatial and longitudinal travel dataset was used to study PHEVs0 operating costs. Whether PHEVs have lower energy costs than CGVs/HEVs depends on charger coverage. Under small charging coverage PHEV40 is more costly than HEV if one0 s DVMT is large. If the gas price is $3, PHEV10 is the least costly even if the battery cost is $200/kW. Impact of fast charging is trivial on energy cost, but significant on charging time.

art ic l e i nf o

a b s t r a c t

Article history: Received 28 September 2013 Received in revised form 18 December 2013 Accepted 23 December 2013 Available online 31 January 2014

Using spatial, longitudinal travel data of 415 vehicles over 3–18 months in the Seattle metropolitan area, this paper estimates the operating costs of plug-in hybrid electric vehicles (PHEVs) of various electric ranges (10, 20, 30, and 40 miles) for 3, 5, and 10 years of payback period, considering different charging infrastructure deployment levels and gasoline prices. Some key findings were made. (1) PHEVs could help save around 60% or 40% in energy costs, compared with conventional gasoline vehicles (CGVs) or hybrid electric vehicles (HEVs), respectively. However, for motorists whose daily vehicle miles traveled (DVMT) is significant, HEVs may be even a better choice than PHEV40s, particularly in areas that lack a public charging infrastructure. (2) The incremental battery cost of large-battery PHEVs is difficult to justify based on the incremental savings of PHEVs0 operating costs unless a subsidy is offered for largebattery PHEVs. (3) When the price of gasoline increases from $4/gallon to $5/gallon, the number of drivers who benefit from a larger battery increases significantly. (4) Although quick chargers can reduce charging time, they contribute little to energy cost savings for PHEVs, as opposed to Level-II chargers. & 2014 Elsevier Ltd. All rights reserved.

Keywords: Plug-in hybrid electric vehicles Operating cost Battery cost

1. Introduction Electrification of transportation is widely regarded as an effective solution to energy security, climate change, and air quality (Ohnishi, 2008; National Research Council (NRC), 2010, 2013). The EV Everywhere Grand Challenge, announced by President Obama in March 2012, aims “to produce plug-in electric vehicles (PEVs) as affordable and convenient for the American family as gasoline-powered vehicles by 2022” (USDOE, 2013). However, fast growth of the PEV market faces two barriers. One is the high cost of battery packs. For example, according to a US

n

Corresponding Author. Tel.: þ1 409 880 8757. E-mail addresses: [email protected] (X. Wu), [email protected] (J. Dong), [email protected] (Z. Lin). 0301-4215/$ - see front matter & 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.enpol.2013.12.054

Department of Energy (DOE) report (2013), the battery cost was $500/kWh in 2012. Another barrier is the lack of public charging facilities (Lin, 2012). Though plug-in hybrid electric vehicles (PHEVs) also have issues such as battery safety, durability, bulkiness, etc., they are less dependent on charger availability, compared to battery electric vehicles (BEVs). PHEVs can operate on gasoline when the battery is depleted. An adequate charging infrastructure, however, can increase a PHEV0 s share of driving on electricity, thus increasing energy savings and promoting consumer acceptance. This paper aims to study the impacts of battery cost and charging infrastructure coverage on market acceptance of PHEVs. PHEVs combine an internal combustion engine (ICE) with a battery which can be charged with grid electricity. PHEVs can operate in the charge-depleting (CD) mode, in which little or no fuel is consumed and little or no tailpipe pollutants are emitted.

X. Wu et al. / Energy Policy 68 (2014) 206–217

After the CD range is exhausted, PHEVs can continue to operate in the charge-sustaining (CS) mode, using the ICE as the major power source, in virtually the same fashion as that of a hybrid electric vehicle (HEV). Having the ability to partially substitute electricity for gasoline, PHEVs can reduce lifecycle greenhouse gas (GHG) emissions compared with conventional vehicles, unless the grid electricity comes from coal (Hawkins and Singh, 2012). A less controversial merit of PHEVs is enhancing the energy security of the nation (Vyas et al., 2009; Lin and Greene, 2011). These benefits come from operating PHEVs in the CD mode. Therefore, it is important to make a full use of the CD mode in PHEVs0 operations. The maximum distance that a fully charged PHEV can operate in the CD mode, known as the CD range, is determined by the effective battery capacity.1 To take full advantage of a PHEV, motorists would hope to operate the vehicle mostly in the CD mode and return home with an empty battery. A long CD range is usually associated with a large and more expensive battery pack. Depending on their travel needs, different motorists might prefer batteries of different sizes. Lin (2012) estimated the optimal electric range for each individual in a national driver sample by tradeoffs of battery cost and energy cost, forming a national distribution of optimal ranges due to variation of driving patterns. Clearly, the impacts of the battery capacity and public charging facility coverage are highly correlated. With an extensive coverage of charging facilities that allow frequent charges, small batteries may meet motorists0 needs; on the other hand, if the government subsidy to PEVs increases, customers may prefer buying PEVs with large batteries, and thus reduce the need for investment in public charging facilities. Therefore, it is necessary to incorporate such correlations in the study of the long-term benefits and costs of PHEVs. For example, Peterson and Michalek, 2012 employed the 20090 s National Household Travel Survey (NHTS) data to investigate the cost of adopting PHEVs with different CD ranges, considering increasing battery capacity and infrastructure coverage. Zhang et al. (2011) used the 20090 s NHTS data taken in the South California to study energy consumption of PHEVs with different CD ranges under three charger coverage scenarios. These NHTS data were converted to a typical one-day travel pattern data. The results were compared with that of conventional gasoline vehicles (CGVs) and hybrid electric vehicles (HEVs), showing that a HEV could reduce 45% fuel consumption (in gallons) compared to a CGV and PHEV40 can help reduce additional 70% fuel consumption (in gallons), compared to a HEV. Furthermore, using the same dataset, Zhang et al. (2013) studied the operating costs of PHEVs and BEVs, assuming optimal charging strategies based on a time-of-use (TOU) electricity rate (which varies by season of a year) within a day. However, the NHTS data are aggregate data based on a cross-sectional survey, which cannot reflect the longitudinal variation in travel patterns of motorists. Furthermore, the NHTS data were collected through phone interview. The accuracy of the travel temporal and spatial information is low. Based on one school-day travel data collected in Austin, Texas, in 2005 or 2006, Dong and Lin (2012) studied the fuel savings and total energy cost of PHEVs under several hypothetical coverage levels of public chargers. These data were recorded by global-positioning-system (GPS) devices installed in vehicles. Therefore, they promise a high accuracy of the temporal and spatial information. However, studies in travel demand modeling and analysis have suggested great variations in motorists0 tripmaking behavior, including daily variations in the trip frequency, trip length, trip chaining, departure time choice and its connections with demographic variables (Pas and Sundar, 1995; Elango et al., 2007; Lin et al., 2012). Specifically, daily vehicle miles traveled

1 The battery0 s capacity is also affected by operational discharge strategies. The total battery capacity is larger than the effective capacity, mainly for protecting the battery life.

207

(DVMT) varies from one day to another for a particular motorist and also varies among motorists. Both the day-to-day variation in the DVMT and motorist heterogeneity could significantly impact the energy consumption of PHEVs (Lin and Greene, 2011). In this paper, we focus on the impacts of two factors – battery capacity and charger coverage – on the energy costs from the perspective of motorists (i.e., we do not consider the cost of building public charger facilities) based longitudinal travel data of multiple motorists. By assuming different scenarios of charger coverage, we want to answer two questions: (1) How much energy cost savings over the long term could PHEVs bring compared with CGVs or HEVs? (2) Is a large-capacity battery worth buying for motorists, considering the trade-off between incremental battery costs and operating cost savings?

2. Data and methods 2.1. Longitudinal travel data Recently, the Puget Sound Regional Council (2008) conducted a household travel choice study to determine how motorists change their travel behavior in response to tolling that varies by location and time of day. The study area was the Seattle metropolitan area, as shown in Fig. 1. A total of 451 vehicles from 331 households (randomly selected) participated in the study, and their detailed travel behavior over up to 18 months (from October 2004 to April 2006) was recorded through GPS devices installed in their vehicles. Khan and Kockelman (2012) studied potential market acceptance of PHEVs and BEVs based on these households0 DVMT revealed from these data.2 However, they assumed that all PHEVs were to be charged only at home. On the other hand, their dataset does not have spatial information associated with each trip, so that it would be difficult to consider the public charging opportunities for these trips. To promote the PHEV market, the public charging opportunities should also be considered. The Seattle dataset used in this paper includes detailed temporal and spatial information of each trip: start time, end time, start location, and end location. After data cleaning, a total of 758,612 trips from 449 vehicles were available. Of those, 415 motorists had active travel data for more than 90 days, and they made 749,828 trips in total. Note that even though the study spanned 18 months, some participants discontinued their participation at some point during this period for a variety of reasons. All 415 motorists0 travel data were used, and it was assumed that each vehicle represents one motorist. Therefore, a motorist0 s travel behavior was recorded as the vehicle0 s locations during the study period. It is found that most participants0 households were located in suburbs, as shown in Fig. 1. The vehicles participating in the study were all CGVs, so by using the data, we ignore the possible change of driving behavior between PHEVs and CGVs. This is appropriate because the focus is on the upfront battery cost and energy cost. Consideration of behavior change will distract the focus, although it could be an interesting extension of the study. 2.2. Battery schemes 2.2.1. Battery capacities and energy consumption rate Four types of batteries are considered: PHEV10, PHEV20, PHEV30, and PHEV40, where 10, 20, 30, and 40 refer to the CD 2 The data used in Khan and Kockelman (2012) were processed and provided by the National Renewable Energy Laboratory0 s Security Transportation Data Project. The dataset is smaller than the one we used here. For example, it has 269,357 trips from 264 households while our dataset has 758,612 trips from 331 households.

208

X. Wu et al. / Energy Policy 68 (2014) 206–217

Fig. 1. Map of the Seattle metropolitan area (the area inside the black rectangular is the Seattle downtown, and the red dots are the locations of the 331 households that participated in the study). (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.) Table 1 CD range, battery size and energy consumption rates of four types of PHEVs.a

CD range (mile) Battery capacity (kWh) Electricity use rate in CD mode (Wh/mile) Gasoline use rate in CS mode (gallon/mile) Gasoline use rate in CD mode (gallon/mile)

PHEV10

PHEV20

PHEV40

PHEV 30 (from interpolation)

R2

11 4.4 287.8788 0.0197 0.0018

21 7.6 337.3016 0.0230 0

38 16 341.6667 0.0288 0

30 (assumed) 11.9 337.9665 0.0258 0

N/A 0.996 0.8104 0.9988 –

a Note that the parameter values of PHEV30 come from interpolation based on three modes of PHEVs in the market, and the value of R2 indicates the accuracy of the interpolation.

range in miles. The Toyota Prius PHEV [CD range of 11 miles (Toyota, 2013)], Ford Fusion Energi PHEV [CD range of 21 miles (Ford, 2013)], and Chevrolet Volt [CD range of 38 miles (GM, 2013)] are used to represent typical PHEV10, PHEV20, and PHEV40, respectively. The differences in these vehicles0 market prices largely depend on the battery size. The fuel economy data are available online (USDOE and USEPA, 2013). A PHEV300 s fuel economy data is interpolated based on these data, as shown in Table 1. It is found that both the battery capacities and the gasoline use rates in the CS mode are highly correlated with the CD range (i.e., R2 is almost 1). Two linear models are Rcd ¼ 0:3914xcap þ 0:2

ð1Þ

Rcd ¼ 0:0003r cs;g þ 0:0168

ð2Þ

where Rcd is the CD range; xcap and r cs;g refer to the battery capacity (kWh) and gasoline use rate in the CS mode (see values in Table 1), respectively, while the relationship between the CD range

and the electricity use rate in CD mode can be modeled by a log curve: Rcd ¼ 38:8 lnðr cd;e Þ þ 206:05

ð3Þ

where r cd;e refers to the electricity use rate in the CD mode. The gasoline use rates in the CD mode in the Fusion and Volt are all 0, while the Prius uses a little gasoline (0.0018 gallon/mile) in the CD mode due to a different powertrain architecture. Hence, we hereby assume that PHEV300 s gasoline use rate in the CD mode is also 0. 2.2.2. DVMT distribution and tailored battery size A popular assumption about PHEVs is that they could be charged at home overnight (Peterson and Michalek, 2012). Based on this assumption, the optimal battery capacity is closely related with a motorist0 s DVMT (Lin, 2012). The DVMT varies among motorists, and even the DVMT for one motorist can vary from day to day (Lin et al., 2012). The Seattle dataset provides detailed trip data for each participant (represented by a vehicle). Therefore, the

X. Wu et al. / Energy Policy 68 (2014) 206–217

209

0.3

Proportion

0.25 0.2

PHEV10

PHEV20

PHEV30

PHEV40

0.15 0.1 0.05 0 0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

DVMT variability in terms of each motorist's typical DVMT Fig. 2. The distributions (normalized histograms) of the daily vehicle miles traveled (DVMT) and its variability of 415 vehicles which have more than 90 active days in the travel choice study. (a) One motorist0 s DVMT distribution. (b) Distribution of the typical DVMT of 415 motorists. (c) Distributions of DVMT variability based on the mode of each motorist0 s DVMT for four DVMT groups.

distribution of each participant0 s DVMT can be estimated. Fig. 2(a) reports the DVMT distribution of one participant in the study. We are most interested in the mode of each vehicle0 s DVMT, as it reflects the most frequent DVMT. The mode of a motorist0 s DVMT is called the motorist0 s typical DVMT. Fig. 2(b) reports the distribution of the typical DVMT of all participating vehicles. Based on the dataset, the average DVMT is 31.8 miles, close to 29.0 miles based on the 20090 s NHTS. Though using a similar dataset from the Seattle travel choice study, Khan and Kockelman (2012) reported that the average DVMT is 25.4 miles, probably because their dataset is smaller. It is unrealistic for battery manufacturers to make the battery capacity exactly equal to any given PHEV motorist0 s typical DVMT (i.e., the mode of a motorist0 s DVMT). A PHEV owner may select a battery capacity based on his/her typical DVMT. For example, if a motorist0 s typical DVMT is 25 miles, then a PHEV30 might be a good choice, and if the typical DVMT is greater than 30 miles, a PHEV40 could be a good choice. According to this battery selection scheme, among 415 motorists, 92 would select the basic PHEV10s; 131 would drive PHEV20s; 89 would adopt PHEV30s, and the remaining 103 would select PHEV40s. Apparently, the variability of one0 s DVMT impacts the effectiveness of such battery selection scheme. Note that the scheme is based on each participant0 s typical DVMT (i.e., the mode of one0 s DVMT), rather than the average DVMT. Therefore, for each participant, we specifically investigate the variability of DVMT in terms of his/her typical DVMT. From the dataset, it is found that a few DVMT of some motorists are very large (over 100 miles). Such large DVMT is unusual, and it can significantly increase the variability and cause big bias. Therefore, we exclude those unusual DVMT in calculating variability of one0 s DVMT, if the proportion of the days having such

unusual large DVMT is less than 5%. The motorists are divided into four groups according to the above described battery selection scheme, and the distributions of variability of four battery groups0 DVMT are reported in Fig. 2(c). It is seen that the DVMT variability is between 100 and 500 for most motorists. Therefore, the DVMT variability is not small, reflecting large variability of travel behavior. Therefore, using the average survey data aggregated into one day to estimate the energy cost of EVs would bring large bias. From Fig. 2(c), it is also seen that the distributions of DVMT variability among four groups are close, especially for three groups: PHEV20, PHEV30 and PHEV40. Therefore, it seems that the DVMT variability does not demonstrate an apparent change with the increase of DVMT. Khan and Kockelman (2012) show that even if all households use PHEV40, only 50% household0 s DVMT could be electrified. Based on our battery selection scheme, around 55% households would see their 50% DVMT could be electrified. Note that only 25% households adopt PHEV40 in our scheme. In the DOE report on the national goal of EV promotion for 10 years (from 2012 to 2022) (USDOE, 2013), the current battery cost was $500/kWh (in 2012). The goal is $125/kWh by 2022. Note that these are the production costs. For potential PHEV customers, the price markup factor should be considered, which reflects the cost of distribution, marketing, and profit. Here this factor is set to be 1.5, which is suggested by Plotkin and Singh (2009) and also used in Peterson and Michalek (2012) to estimate the retail price of PHEVs. For example, if the battery cost is reduced to $125/kWh, the market price would be $187.5/kWh. Based on this national goal, four battery price schemes are considered: $600, $450, $300 and $200. According to Table 1, the battery capacities of Prius (4.4 kWh) Fusion (7.6 kWh), and Volt (16 kWh) are used to characterize PHEV10, PHEV20, and PHEV40, respectively. The battery capacity of PHEV30

210

X. Wu et al. / Energy Policy 68 (2014) 206–217

Table 2 Incremental battery costs compared with PHEV10.

CD range (kWh/mile) Additional capacity required (kWh) Incremental battery cost based on PHEV10

$600/kWh $450/kWh $300/kWh $200/kWh

PHEV10

PHEV20

PHEV30

PHEV40

0.2879 –

0.3373 3.20

0.3380 7.54

0.3417 11.60

– – – –

$1920 $1440 $960 $640

$4500 $3375 $2250 $1500

$6960 $5220 $3480 $2320

Table 3 Scenarios of charger coverage. Scenario I

Scenario II

Home chargers only Home chargers & public chargers (449 home locations) in downtown Seattle (449 home locations þ69 downtown blocks) 2% coverage of public location

is assumed to be 11.9 kWh based on interpolation of the above data. The incremental battery costs, using PHEV10 as the benchmark, are listed in Table 2. 2.3. Charger coverage scenarios The cost-effectiveness of battery capacity also depends on the charger coverage. If the public charger coverage is extensive, a small battery pack may be sufficient; otherwise, if a PHEV can be charged only at home, a large battery pack may be preferred. In the dataset, the locations of a trip are represented by latitude and longitude coordinates. We have extracted the start and end locations of all 758,612 trips. When a vehicle is started, it may take a while for the GPS device to work properly. Therefore, a trip0 s start location recorded by the GPS device may not be accurate. Therefore, the 631,171 distinct end locations are used in the subsequent analysis. In addition, some nearby latitude and longitude coordinates might refer to the same location. For example, a motorist may not always park his/her car at the exactly same location in a parking lot. To better understand the purpose of a trip, a set of grid is defined to cover the entire Seattle area: in the downtown area, the grid cell is set to be 0.5  0.5 mile2; in suburbs, it is 1  1 mile2; and in outer suburbs, it is 5  5 mile2. Therefore, a grid cell (called blocks in the following) may contain multiple end locations (represented by latitude and longitude coordinates) of trips made by one motorist, and these trips may have the same purpose, such as for work or shopping. Four charger coverage scenarios are considered here, and for each scenario, the number of blocks that have public chargers is presented in Table 3. The first scenario assumes home charging only. Home charging is found more valuable than workplace and public charging (Lin and Greene, 2011), so it is not surprising that many existing studies on the PHEV energy analysis, such as Zhang et al. (2011), Khan and Kockelman (2012) and Peterson and Michalek, (2012), also consider the home charging as a basic or only charging scenario (though they may not use the words “baseline scenario”). This scenario implies no external investment on public charging facilities. A household location is regarded as a block. A trip is thought to end at home if the distance from this location to the location of the vehicle0 s household is less than 100 m. It implies that if its location is less than 100 m from home, the vehicle could be charged at home, and this home charger is used exclusively only for the vehicle belonging to this household.

Scenario III

Scenario IV

Home chargers and public chargers are available in popular destinations (449 home locations þ 657 blocks) 16% coverage of public location

Home chargers and public chargers are everywhere (449 home locations þ 4129 blocks) 100% coverage of public locations

From the database, there are a total of 515,085 distinct non-home locations where trips end, found in 4129 blocks. Another three scenarios are about public chargers in these non-home locations. Scenario II assumes that public chargers are available only in downtown Seattle. For simplicity, we assume that whenever a vehicle stops in the downtown area, it could be charged if the motorist so desires (as explained in Section 2.4). In this scenario, 69 blocks are identified having public chargers, covering only 2% of public area. Many trips attractions, such as shopping malls, office buildings, and restaurants, are located outside the downtown area. Therefore, in Scenario III, we assume that public chargers are available in some popular destinations. According to each motorist0 s trip-ends during the study period, some blocks are frequently visited, where workplaces or shopping centers are located. For one motorist and one block, the number of days this block was visited by this motorist can be calculated. For example, for one vehicle, one block located in downtown Seattle was visited 233 days out of the 386 travel days. Thus, it is possible that the motorist0 s workplace is located in this block. On the other hand, shopping trips might not happen as frequently as work trips. A person may go shopping once a week. For the same vehicle mentioned above, another block was visited 39 days out of the 386 days. This block may contain some shopping centers or restaurants that this motorist frequently visits. Another two blocks were also frequently visited by this motorist. These four blocks include 488 distinct non-home locations where this vehicle stopped. Note that the dataset does not tell the location of workplace of participants. Even though a block is visited by a participant frequently, it is not necessary to be a workplace. To simplify, we call such places as “popular destinations”. Note that locations are represented by latitudes and longitudes, while blocks are represented by grid cells. In this study, a block is regarded as a popular destination for a motorist if the ratio of days when this block is visited by this motorist to the total days when he/she traveled is larger than 10%. Using this criterion, we identified 657 blocks as popular destinations (see Fig. 3), covering 16% of the total number of blocks and containing 233,856 distinct locations where the motorists in the Seattle dataset had ever stopped during the study period. From Fig. 4, it is seen that the number of popular destinations increases steadily with the number of vehicles. It implies that the sampled participants are relatively independent: their popular destinations do not overlap significantly.

X. Wu et al. / Energy Policy 68 (2014) 206–217

211

Fig. 3. Popular destinations in the Seattle metropolitan area of 449 vehicles.

percentage of vehicles 700 600 500 400 300 200 100 0

37%

59%

80%

98%

100% 20%

# Blocks Percentage of blocks

15% 10% 5% 0%

80

167

263

357

441

Percentage of blocks

# blocks

18%

449

# vehicles Fig. 4. Number of popular destinations vs the number of vehicles from the Seattle travel dataset.

Finally, we consider an extreme case (Scenario IV), in which public chargers are everywhere (100% coverage). Therefore, a motorist could charge his/her PHEV at any public location if needed. This scenario represents very aggressive infrastructure deployment. Dong and Lin (2012) as well as Peterson and Michalek (2012) also considered such an ambitious scenario, which is regarded as an upper-bound of the charger coverage, while the home charge only scenario can be treated as the coverage lower bound. Since electricity is more economically efficient than gasoline, the energy cost of a PHEV in this scenario can also be regarded as its lower bound. We are interested to investigate how much more energy cost could be saved when the coverage increases across scenarios. 2.4. Computational model 2.4.1. Energy consumption model The estimation of PHEV energy consumption follows the method presented in Dong and Lin (2012). The travel distance of

each trip, denoted as di (miles), and the dwell time between two consecutive trips, as tj (hours), are given. The consumption of electricity and gasoline depends on the energy consumption rates and the distances traveled in the CD and CS modes, as follows: Ee ¼ ∑ di;cd r cd;e

ð4Þ

i

Eg ¼ ∑ðdi;cs r cs;g þ di;cd r cd;g Þ;

ð5Þ

i

where Ee and Eg are the energy consumptions of electricity and gasoline, respectively; di;cd and di;cs are the travel distances (miles) of trip i in the CD mode and CS mode, respectively; as defined above, r cd;e and r cd;g are the electricity and gasoline consumption rates (kWh/mile and gallons/mile) in the CD mode, respectively; and r cs;g is the gasoline consumption rate (gallons/mile) in the CS mode. The total energy cost is computed as follows: C en ¼ Ee pe þ Eg pg ;

ð6Þ

where C en is the energy cost of electricity and gasoline combined, and pe and pg are the price of electricity ($/kWh) and price of gasoline ($/gallon), respectively. Whether to recharge at stop j is determined by charger availability and user decision. If a battery is recharged at stop j, I j ¼ 1; otherwise, I j ¼ 0. If a charger is available, then X j ¼ 1; otherwise, X j ¼ 0. If a traveler decides to charge at stop j, Y j ¼ 1; otherwise, j, Y j ¼ 0. The determination of Y j will be described in Section 2.4.2. Therefore, we have Ij ¼ X j Y j :

ð7Þ

212

X. Wu et al. / Energy Policy 68 (2014) 206–217

When a PHEV is charged at stop j, the potential energy increase in the battery is determined as follows:

benefit and cost at stop j are determined, respectively, by (Dong and Lin, 2012):

Rp;j ¼ minðRcd  Rsoc;j ; Pt j =r cd;e Þ ;

Bj ¼ Rp;j ðr cs;g pg þ r cs;g crf þ gvÞ

ð8Þ

where Rp;j is the increased CD range (miles) if a PHEV is recharged at stop j; Rcd is the CD range (miles) which is also used above; Rsoc;j is the remaining CD range when arriving at stop j and is dependent on the previous trips P is the charging power (kW); and t j is the dwell time (hours). In Eq. (8), Rcd , r cd;e ; and P are all predetermined parameters, and r cd;e and Rcd are dependent on the type of the assigned PHEV. Rp;j can be regarded as the chargeable range at stop j. Note that the entire dwell time is assumed to be available for charging in this paper, i.e., no waiting time. The remaining CD range when arriving at stop j; denoted as Rsoc;j ; can be calculated based on the battery level at the previous stop Rsoc;j  1 , possible recharging, i.e., I j  1 Rp;j  1 , and travel distance dj : Rsoc;j ¼ maxð0; Rsoc;j  1 þ I j  1 Rp;j  1  dj Þ :

and C j ¼ Rp;j r cd;e pe þ crc ;

Finally, the travel distances of trip i in the CD and CS mode can be computed as follows: ð10Þ

di;cs ¼ di  di;cd :

ð11Þ

3. Parameter settings A PHEV with appropriate battery capacity is assigned to each participant based on the criterion described in Section 2.2.2. The total operating cost of each PHEV during the study period can be computed according to Eqs. (4)–(13). The parameters in these equations are set as follows. In the fuel economy estimation of USDOE and USEPA (2013), the prices of electricity and gasoline are set to be $0.12/kWh and $3.6/gallon, respectively. We use $0.12/kWh as the price of electricity, while the price of gasoline is roughly set at $3, $4 or $5 per gallon because it is more volatile. The charging power is set at 6 kW (i.e., a normal Level II charger), which requires around 4 h to fully charge a PHEV40 battery. Quick chargers with 50 kW charging power using 100–125 A direct current (DC) have been introduced to the market.

9

$3/gallon

9

8

$4/gallon

8

7

$5/gallon

Energy cost (cents/mile)

Energy cost (cents/mile)

2.4.2. Charging decision model The charging decision at stop j, denoted as I j , is determined by two factors, according to Eq. (7). The first one is the availability of chargers, that is, X j , and the second factor is the charging wish, that is, Y j ., which depends on the ratio of charging benefit (denoted as Bj ) to charging cost (denoted as C j ). A driver will charge his/her PHEV at stop j if the charger is present at this location, that is, X j ¼ 1, and the benefit and cost ratio of charging is larger than or equal to 1. That is to say, Y j ¼ 1 if Bj =C j Z1, and Y j ¼ 0 if Bj =C j o 1 (Dong and Lin, 2012). At stop j, the charging benefit and cost depend on the chargeable range Rp;j (see Eq. (8)). Gasoline can then be saved if the driver decides to charge at this stop (a charger is present). The charging

6 5 4 3 2 1 0

$3/gallon $4/gallon $5/gallon

7 6 5 4 3 2 1 0

8

PHEV10 PHEV20

PHEV30

PHEV40

$3/gallon $4/gallon $5/gallon

7 6 5 4 3 2 1 0

Overall

PHEV10

8

Energy cost (cents/mile)

Overall

Energy cost (cents/mile)

ð13Þ

where crf and crc refer to the refueling and charging hassle costs, respectively. When refueling a vehicle, a driver cannot leave the pump for the safety concern. The refueling hassle cost refers to the willingness of a driver to pay for the avoidance of visiting the gas stations; when charging a PHEV, a driver also has to conduct a series of actions, which a driver would want to avoid if not necessary. However, when charging a PHEV, a driver can leave the vehicle at the charger station and do other things (e.g., sleep during charging). Therefore, the charging hassle cost should be less than the refueling hassle cost. How to set the values of these parameters is discussed in Section 3. Here the value of environmental benefits due to the gasoline savings is not considered, because of the difficulty in its estimation.

ð9Þ

di;cd ¼ minðRsoc;i  1 þ I i  1 Rp;i  1 ; di Þ:

ð12Þ

PHEV20

PHEV30

PHEV40

$3/gallon

7

$4/gallon

6

$5/gallon

5 4 3 2 1 0

Overall

PHEV10 PHEV20 PHEV30 PHEV40

Overall

PHEV10

PHEV20

PHEV30

PHEV40

Fig. 5. Average energy cost per mile (cents) for different models of PHEVs, charger coverage scenarios, and the prices of gasoline. (a) Scenario I: Home Chargers Only. (b) Scenario II: Homeþ Downtown. (c) Scenario III: Home þ Popular Destination. (d) Scenario IV: Everywhere.

X. Wu et al. / Energy Policy 68 (2014) 206–217

Though expensive (up to $100,000 each) (Plug in America, 2013), these quick chargers require only about 20 min to fully charge a PHEV40 battery. In the quick charger scenarios, 50 kW is assumed as opposed to 6 kW for charging power. Electricity price is assumed as the same at $0.12/kWh, regardless of charging power. The discount rate is set to be 6% and three battery payback periods of 3, 5 and 10 years are used. As to the hassle cost estimation, Lin (2012) suggests $3 per refueling at gasoline stations; Dong and Lin (2012) used $0.27 as the perceived burden of charging, by assuming that the value of time is $16/h and plugin may take around 1 min.

4. Results 4.1. Energy savings of PHEVs The average energy cost per mile, by dividing the total energy cost in Eqs. (4) and (5) by distance, is used for comparison among different models of PHEVs, charger coverage scenarios, and gasoline prices. The results are reported in Fig. 5. Mid-range PHEVs (20 or 30 miles of CD range) may achieve the lowest energy cost, if the infrastructure coverage is limited as shown in Fig. 5(a) and (b), with Scenario I (Home Chargers Only) and Scenario II (Home and Downtown). A PHEV10 leads to higher energy costs as a result of insufficient charging opportunities, a shorter CD range and therefore higher shares of distance consuming gasoline. PHEV40s are also associated with higher energy cost, but for different reasons. They do lead to higher shares of electrified distance, but not the extent to offset the higher gasoline consumption rate in the CS mode (primarily using gasoline) than a PHEV10, PHEV20, or PHEV30, caused by the heavier battery. That is, if there is no matching coverage of infrastructure, carrying a heavy battery in the CS mode could, as in fact demonstrated in the results, offset the benefit of more electrified distance. Overall, more infrastructure coverage (from Fig. 5(a)–(d)) brings down the energy cost, regardless of the type of PHEVs, but PHEV40 benefits more from upgrading the infrastructure to cover popular destinations or all public places, relative to home and downtown coverage (from Fig. 5(b) to (c) or to (d)). Now we can try to answer the first question proposed at the beginning of this paper: how much energy cost savings could PHEVs bring compared with CGVs or HEVs? The fuel efficiencies of CGVs and HEVs are assumed to be 27.5 and 42 miles per gallon (MPG) (highway and urban travel combined), respectively. Since the total travel distance of each vehicle is known, the energy costs are calculated by assuming that all 415 motorists drove CGVs or HEVs. Table 4 presents a summary of energy savings when the price of gasoline is $4/gallon. On average, PHEVs could provide a 58–68% savings in energy cost compared with CGVs and a 36–51% savings compared with HEVs, depending on infrastructure coverage. With a large coverage of public chargers, the savings could be more. Using a similar dataset, Khan and Kockelman (2012) studied the cost comparisons in Table 4 PHEV energy cost savings (%) compared with conventional gasoline vehicles (CGVs) or hybrid electric vehicles (HEVs) of all 415 motorists during the study period (the price of gasoline is assumed to be $4/gallon). Charger coverage

Compared with CGVs I

II

III

Compared with HEVs IV

I

II

III

IV

switching from a CGV to a PHEV with similar vehicle characteristics under different gasoline prices (only home charging) for three DVMT groups. Because different parameters are used, the results are not compared directly. However, we can see that under the home charging scenario, the increase of energy cost savings with the increase of gasoline price is consistent with that in Khan and Kockelman (2012). Using the 20090 s NHTS data in Southern California, Zhang et al. (2011) found that PHEV10s and PHEV40s can reduce 45% and 70% fuel consumption in gallons compared to a HEV, respectively. Our findings are generally consistent with those of Zhang et al. (2011). However, we also found that in some cases, PHEVs are even less efficient than HEVs when the charger coverage is low (Scenarios I and II). As shown in Table 4, compared with HEVs, the energy savings of PHEVs could be  20%. The reason is that when the charger coverage is low, a PHEV40 may primarily operate in the CS mode while carrying a heavy battery. The fuel economy of PHEV40s in the CS mode is 35 MPG, less than that of HEVs (i.e., 42 MPG). For motorists who have to drive long distances frequently, PHEVs may not be a good choice without large public charger coverage. More infrastructure coverage seems to enhance the infrastructure equity among all PHEV motorists. As shown in Table 4, among motorists for the Scenario I (Home Chargers Only), the energy cost saving from CGVs can be as large as 73% (the most advantaged motorist), but can also be as small as only 21% (for the most disadvantage motorist). The variation is due to difference in spatial patterns of trips. The expansion of charging infrastructure, such as that from Scenario I to IV, brings almost no additional benefit for the most advantaged motorist, but substantially increases the savings from 21% to 53%, more than double, for the most disadvantage motorist. More infrastructure coverage seems to bring more benefits to disadvantaged PHEV motorists and narrow the gap of energy cost savings from GCVs. The same can be said for the savings from HEVs, according to Table 4. The effect of the gasoline price on PHEV energy savings, compared with CGVs and HEVs, is summarized in Fig. 6. When the price of gasoline increases, PHEVs save more in energy costs, compared with HEVs or CGVs under any charger coverage scenario.

4.2. Impact of battery size If a motorist adopts a large-battery PHEV due to his/her high typical DVMT, could the additional operating cost savings compensate for the incremental battery cost, compared to a PHEV10? The cost savings include two aspects: one is the energy cost savings and another is the reduced hassle cost associated with less frequent charging and refueling, i.e., saved time for charging or refueling. Note that the energy consumption rates listed in Table 2 take into consideration the reduced fuel economy associated with larger and heavier batteries. The refueling hassle cost depends on the quantity of gasoline consumed, as described in Section 2.4.2. 80%

CGVs, Scenario I

70%

CGVs, Scenario II

60%

CGVs, Scenario III CGVs, Scenario IV

50%

HEVs, Scenario I

40%

HEVs, Scenario II

30%

HEVs, Scenario III HEVs, Scenario IV

20% $3/gallon

Average Maximum Minimum

57.95 73.07 20.94

59.27 73.07 21.21

63.58 73.42 47.34

67.65 73.44 53.25

35.78 58.87  20.74

37.80 58.87  20.34

44.38 59.40 19.58

50.60 59.44 28.60

213

$4/gallon

$5/gallon

Fig. 6. Average energy cost savings (in percentage) of PHEVs compared with CGVs and HEVs under four charger coverage scenarios and three price schemes of gasoline.

X. Wu et al. / Energy Policy 68 (2014) 206–217

3 Note that it would bring some overestimation as there may be no travel on some days or data missing for some days. However, it is applied to all motorists from any battery group and what we compare is the annual energy cost gap among these groups. Therefore, such overestimation may not bring large bias to the results of the comparison

20 Home Chargers (Scenario I)

Number of PHEV users

The daily operating cost of each vehicle is calculated based on its total number of travel days and its annual operating cost is calculated by multiplying its daily energy cost with 365.3 The present value of each vehicle0 s total operating cost for a certain payback period can then be calculated (this could be regarded as the battery life). The base gasoline price is set at $4/gallon. Three payback periods, 3 years, 5 years and 10 years, and four battery unit price schemes (see Table 2), are considered. By comparing the present values of operating costs for each period, we find that PHEV10s would be the best option if the price of gasoline is no more than $4/gallon and the battery unit price is not less than $300/kWh, because the incremental battery cost paid by motorists for the large battery cannot be compensated for by the potential additional savings in energy and refueling and recharging hassle costs. Even if the unit battery cost is reduced to $200/kWh, close to the national goal in 2022 (USDOE, 2013), PHEV10s are still the most economical in most cases. The most important reason is that a PHEV400 s fuel economy is worse than that of a PHEV10, due to its additional vehicle weight. Such observations are well consistent with those in Peterson and Michalek (2012), who found that PHEVs demonstrates diminishing returns in fuel cost savings with the increase of battery size. The current fuel economy data were used to estimate the energy cost for the perceived lifetimes of 3, 5, or even 10 years, separately. Currently, PHEV40s are selling a little better than PHEV10s, probably due to higher subsidy for PHEV40s: a PHEV40 like Volt gets a $7500 tax credit, while a PHEV10 like Prius gets only $2500 (Internal Revenue Service, 2009). With the current tax credit policy, the incremental battery cost of a PHEV40 could be well offset. Actually, Peterson and Michalek (2012) found that the current federal PHEV subsidy tends to favor largebattery PHEVs. The impact of the battery size is further investigated considering different gasoline prices, assuming a unit battery cost of $200/ kWh and a payback period of 10 years. Note that the 10-year payback period is chosen as a conservative criterion. If a battery upgrade is difficult to justify within a 10-year payback, it will be more so with a shorter payback period. The results are reported in Fig. 7. It is found that even for a period of 10 years, no one can be benefit from a larger battery under any charger coverage scenario if the price of gasoline is as low as $3.00/gallon. However, if it rises to $5.00/gallon, the number increases significantly in all scenarios. For example, if the price is $5/gallon, 59 out of 131 PHEV20 users or 45 out of 103 PHEV40 users will see that the additional cost saving can offset the incremental battery cost under Scenario I (home chargers only), while if it drops to $4/gallon, that number is only 19 for PHEV20 users and 7 for PHEV40 users. Therefore, the implication is that higher gasoline prices make large-battery PHEVs much more attractive. Surprisingly, for those whose typical DVMT is between 10 miles and 30 miles (then they would select PHEV20s or PHEV30s), PHEV10s would become even more attractive with increasing public charger coverage, as shown in Fig. 7 (a) and (b). For example, when the price is $5/gallon, 59 and 55 out of 131 PHEV20 users would find that having the large battery is better than PHEV10 under Scenarios I and II, respectively; however, this number drops dramatically to only 21 under Scenario III. The reason is that for most PHEV20 or PHEV30 users, their most popular destinations (workplaces and shopping malls or restaurants) are less than 10 miles away from home. If chargers are available at these popular destinations, PHEV10s could be good

Home Downtown (Scenario II)

15

Popular Destination (Scenario III) Everywhere (Scenario IV)

10 5 0 PHEV20

Number of PHEV users

214

PHEV30

70

Home Chargers (Scenario I)

60

Home Downtown (Scenario II)

PHEV40

Popular Destination (Scenario III)

50

Everywhere (Scenario IV)

40 30 20 10 0 PHEV20

PHEV30

PHEV40

Fig. 7. Number of large-battery PHEV users who see the incremental operating cost savings over a 10-year payback period due to a large battery can offset the incremental battery cost, under four charger coverage scenarios and at three gasoline prices (the unit battery cost is $200/kWh). Note: when the gasoline price is $3.0/gallon, no PHEV user benefits from having a battery. (a) Gasoline price is $4.0/gallon. (b) Gasoline price is $5.0/gallon.

enough to keep PHEVs operating in the CD mode as much as possible, even though their typical DVMT is between 10 miles and 30 miles. When public chargers are available everywhere, such numbers increase again to 33 (but still less than those under Scenarios I and II) because a large battery can support long-distance travel in the CD mode and the destinations of those long-distance trips are usually not as popular. On the other hand, PHEV40 users (whose typical DVMT is larger than 30 miles) would be less sensitive to an increase in public charger coverage, compared with PHEV20 and PHEV30 users. When the gasoline price is $5/gallon, the number of PHEV40 users who would benefit from the large battery increases from 45 (Scenario I) to 48 (Scenario II) and then drops to 45 (Scenario III) and 42 (Scenario IV), because their popular destinations are far from home, and they make many long-distance trips. A large battery is able to support a long CD range, which is somewhat insensitive to the public charge coverage. The operating costs of various types of PHEVs were used in the above analysis. Considering only the energy cost saving, the number of users who benefit from a large battery drops in most scenarios, as shown in Fig. 8 (where the gasoline price is set to $5/ gallon and the same features are found when the gasoline price is $4/gallon). It implies that a large-battery PHEV could help save gasoline (less refueling hassle cost) and help reduce the frequency of recharging (less recharging hassle cost). However, there is an exception: when chargers are everywhere, the frequency of recharging for small-battery PHEVs will increase instead, because PHEV users always want to avoid using gasoline if its price is as high as $5/gallon. As a result, the increased recharging hassle cost of PHEV20s and PHEV30s (due to more recharging behavior) would be greater than the savings on energy cost and refueling hassle cost, therefore making PHEV20s and PHEV30s even less attractive, compared with PHEV10s. A PHEV40 is less sensitive to charge coverage than a PHEV20 or PHEV30, because it is able to

X. Wu et al. / Energy Policy 68 (2014) 206–217

215

50

70 Total Cost Savings

Total Cost Savings

Energy Cost Savings 50 40 30 20

# PHEV30 Users

# PHEV20 Users

60

10 0

40

Energy Cost Savings

30 20 10 0

Scenario I

Scenario II

Scenario III Scenario IV

Total Cost Savings

Scenario I

Scenario II

Scenario III

Scenario IV

Energy Cost Savings

# PHEV40 Users

50 40 30 20 10 0 Scenario I

Scenario II

Scenario III

Scenario IV

support a longer CD distance and requires less frequency of recharging. Therefore, a PHEV40 could help save on refueling and recharging hassle costs. 4.3. Impact of quick charging 4.3.1. Energy savings Our results show no significant energy savings for PHEVs from switching Level-II chargers to 50 kW quick chargers. Following the same calculation procedure conducted in Section 4.2, we calculate each vehicle0 s daily energy cost assuming that all public chargers are quick DC chargers (50 kW) under Scenario II (downtown), III (popular destinations), and IV (everywhere). By comparing the results, we estimate how much more energy cost can be saved for each PHEV for a 10-year payback period by upgrading to quick chargers (by comparing the net present value of energy costs). Again, the 10-year payback period is chosen as a conservative criterion. As shown in Fig. 9, even when the quick chargers are everywhere (admittedly too ambitious), 60% of 415 vehicles see a savings of only 3% (less than $160). Such savings are almost negligible given a period of 10 years! 4.3.2. State of charge The departure state of charge (SOC), i.e. the SOC after recharging and before the next trip, affects the energy cost of upcoming trips, but perhaps more important psychologically, it could affect the perceived probability of having an electrified trip (totally or largely). Therefore, the departure SOC reflects the utility of this charging activity. We chose to use the number of trips that start with at least 80% of SOC to measure the utility of charging. For example, Vehicle #43 (driving a PHEV40) has a total of 2300 trips recorded during 393 travel days. It is found that the number of trips that start with 80% or more SOC is increased by 26% if all public chargers at popular destinations were switched from LevelII chargers to quick chargers. On the other hand, for Vehicle #500 (using a PHEV10), the number of such trips is increased by only 1.3% under the same conditions. It is found that quick charging upgrades increase the departure SOC. Fig. 10 reports the average increase in the number of trips

Percentage of vehicles

Fig. 8. Number of large-battery PHEV users who benefit from having a large battery, compared with having a PHEV10 under four charger coverage scenarios (the price of gasoline is $5/gallon and the unit battery cost is $200/kWh). (a) PHEV20 (131 motorists). (b) PHEV30 (89 motorists). (c) PHEV40 (103 motorists).

100% 80% 60%

Downtown (Scenario II) Popular_destination (Scenario III) Everywhere (Scenario V)

40% 20% 0%

Percentage of energy cost savings Fig. 9. Percentage of vehicles at different levels of energy savings for a 10-year payback period, if all public chargers are quick chargers, compared with the results when all are Level-II chargers.

Downtown

Pop_dest

Everywhere

16% 14% 12% 10% 8% 6% 4% 2% 0% PHEV10

PHEV20

PHEV30

PHEV40

Fig. 10. Average increase of the number of the trips (in percentage) that have departure SOC of 80% or more under three scenarios of public charger coverage for four models of PHEVs.

with Z80% departure SOC, by PHEV type and by infrastructure coverage scenario, due to the upgrades. When quick chargers are available at all popular destinations, such increase is around 10% on average, while if they are everywhere, such increase is 14% on average. The results reported in Fig. 10 are consistent with the results on the energy cost savings found above. It is difficult to judge the significance of the impact on energy savings and on the perceived probability of electrifying the next trips, because the cost of a quick charging infrastructure is not fully considered in this study and the psychology of perceiving a departure SOC is not well understood.

216

X. Wu et al. / Energy Policy 68 (2014) 206–217

Table 5 Average actual charging time (minutes) of 415 vehicles at two types of public chargers (quick chargers as QCs and Level-II chargers) under three scenarios of public charger coverage. PHEV model

Percentage of vehicles

PHEV10 PHEV20 PHEV30 PHEV40

Public chargers at downtown

Public chargers at popular destinations

Public chargers everywhere

QC (50 kW)

Level II (6 kW)

QC (50 kW)

Level II (6 kW)

QC (50 kW)

Level II (6 kW)

0.41 0.87 1.24 2.06

24.45 54.47 78.88 101.31

1.66 3.50 5.90 8.57

29.51 61.88 91.63 124.57

3.77 6.45 9.22 12.63

40.39 73.31 102.30 136.46

100% Downtown (Scenario II)

80%

pop_dest (Scenario III)

60%

Everywhere (Scenario IV)

40% 20% 0% 0.8

0.85

0.9

0.95

1

Charge time savings due to public quick chargers (percentage) Fig. 11. Percentage of vehicles that have different levels of total charging time savings (in percentage), caused by quick chargers, and based on the charging time at Level-II public chargers.

4.3.3. Charging time So far, we have assumed that the dwell time between two trips could be used entirely for charging. However, in the real world, a charger may not always be ready immediately for use at a charger station, especially considering that charging time is much longer than refueling time. If the actual charging time can be reduced significantly, it would allow more PHEVs to be charged at the same locations. Therefore, we have to consider the actual charging time when comparing the effects of quick chargers and normal chargers. It is no surprise for the results to show that long-range PHEVs need longer charging time. Quick charging upgrades are found to significantly reduce charging time. Table 5 reports the average actual charging time of 415 vehicles at public chargers under three scenarios. For Level-II chargers, the average charging time ranges from half an hour to 2 h, while it is less than 10 min in most cases at quick chargers. Fig. 11 shows the distribution of charging time savings (in percentage) by quick charger upgrades. For all scenarios, a charging time saving of more than 90% for more than 80% vehicles (out of 415 vehicles) was observed. Therefore, by helping to save charging time so dramatically, quick chargers allow more PHEVs to be charged at one charging location, and allow more miles to be electrified and more petroleum use to be cut.

5. Discussion and conclusions This paper employed the longitudinal travel data from 749,828 trips made by 415 vehicles during 3–18 months from the Seattle metropolitan area, to address (1) how much energy cost could be saved for PHEVs compared with CGVs or HEVs, and (2) whether a large-battery PHEV can be justified based on the incremental energy cost savings. The dataset provides detailed spatial and temporal information of trips, which makes it possible for us to incorporate specific location-based public charging coverage in the analysis of PHEV operating costs. The data come from a travel choice study which aims to investigate how motorists change their

travel behavior in response to tolling that varies by location and time of day. All participants in the study used CGVs. In this paper, we directly employed this dataset and assumed that their travel behavior would not change if they were to switch to a different vehicle technology, such as PHEVs or HEVs. The average DVMT based on our dataset is close to that of the 20090 s NHTS data. However, it is not clear if conclusions from this study can be applied to other regions. Another caveat is that we do not consider the cost of building public charging infrastructure. In future studies, we expect to consider infrastructure cost and examine cost-effectiveness of different infrastructure scenarios. This study generates several methodological contributions. First, we used a longitudinal travel dataset to capture variability of travel behavior over time and among motorists. Second, we considered endogenous charging decisions. Third, PHEV ranges were assigned to motorists based on individual typical driving patterns, in order to avoid assigning ranges that are not suitable for individual motorists. By using realistic data for vehicle and infrastructure technologies and carefully setting up the infrastructure deployment scenarios, we have made several important observations. First, whether PHEVs have lower energy costs than those of CGVs or HEVs depends on coverage of public chargers. It is found from the travel data that PHEVs could help save around 60% or 40% in energy costs, compared with CGVs or HEVs, respectively. Midrange PHEVs (20 or 30 miles of CD range) may achieve the lowest energy cost, if the infrastructure coverage is limited (to home or downtown). When more public chargers are available, the savings could be even more. On the other hand, without large coverage of public chargers, HEVs may even be a better choice than largebattery PHEVs. The general results are well consistent with other studies, such as Khan and Kockelman (2012) and Peterson and Michalek (2012). The incremental battery cost of long-range PHEVs is difficult to justify based on the PHEV0 s incremental operating costs savings unless a government subsidy is offered for large-battery PHEVs. It is found that if the gasoline price is less than $4/gallon, even at battery cost of $200/kWh and for a period of 10 years, PHEV10s are the best option in most cases (without subsidy). However, the battery technology development for the next 10 years is not considered. The price of gasoline has a significant impact on the costeffectiveness of PHEVs, especially those with a large battery. When the price increases, PHEVs could provide a greater energy cost savings than CGVs and HEVs. More importantly, it was found that when it increases from $4/gallon to $5/gallon, the number of PHEV users who benefit from having a larger battery increases significantly under each scenario of charger coverage. Therefore, the subsidy policy and gasoline price are two key factors that impact the acceptance of largebattery PHEVs, which use less gasoline and save more energy. Policymakers should pay attention to these two factors. If chargers are immediately available when a PHEV motorist arrives and parks, the effect of having quick DC chargers (50 kW) would be almost negligible in terms of long-term energy savings, compared with the normal Level-II chargers (6 kW). The assumption of immediate availability is less realistic if more PHEVs and BEVs are on the road. The most important advantage of quick chargers lies in the fact that they can reduce the charging time dramatically. Therefore, they cut the recharging time for the previous PHEV, and make the charger more likely to be available for the next one.

Acknowledgments This research is sponsored by Lamar University 2013 Research Enhancement Grant 420212 and the U.S. Department of Energy,

X. Wu et al. / Energy Policy 68 (2014) 206–217

Office of Energy Efficiency and Renewable Energy, Vehicle Technologies Office, under Contract DE-AC05-00OR22725 with UTBattelle, LLC. The authors assume sole responsibilities for the content expressed. This manuscript has been authored by the Oak Ridge National Laboratory, managed by UT-Battelle LLC under Contract no. DEAC05-00OR22725 with the US Department of Energy. The US Government retains and the publisher, by accepting the article for publication, acknowledges that the US Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US Government purposes. References Dong, J., Lin, Z., 2012. Within-day recharge of plug-in hybrid electric vehicles: energy impact of public charging infrastructure. Transp. Res. Part D 17, 405–412. Elango, V.V., Guensler, R.L., Ogle, J.H., 2007. Day-to-day travel variability in the commute Atlanta, Georgia study. Transp. Res. Rec. 2014, 39–49. Ford, 2013. Inside the 2013 Ford Fusion Energi Plug-In Hybrid. Available at 〈http:// media.ford.com/images/10031/2013_FusionEnergi_hero.pdf〉. GM, 2013. Chevrolet Volt Specification. Available at 〈http://gm-volt.com/ full-specifications〉. Hawkins, T.R., Singh, B., 2012. Comparative environmental life cycle assessment of conventional and electric vehicles. J. Ind. Ecol. 17, 53–64. Internal Revenue Service, 2009. Notice 2009-89: New Qualified Plug-in Electric Drive Motor Vehicle Credit. Internal Revenue Service, November 30, 2009. Khan, M., Kockelman, K., 2012. Predicting the market potential of plug-in electric vehicles using multiday GPS data. Energy Policy 46, 225–233. Lin, Z., Greene, D., 2011. Assessing energy impact of plug-in hybrid electric vehicles: significance of daily distance variation over time and among drivers. Transp. Res. Rec. 2252, 99–106. Lin, Z., 2012. Optimizing and diversifying the electric range of plug-in hybrid electric vehicles for U.S. drivers. Int. J. Alternative Powertrains 1 (1), 108–194, http://dx.doi.org/10.4271/2012-01-0817.

217

Lin, Z., Dong, J., Liu, C., Greene, D., 2012. Estimation of energy use by plug-in hybrid electric vehicles: validating Gamma distribution for representing random daily driving distance. Transp. Res. Rec. 2287, 37–43. National Research Council (NRC), 2010. Transitions to Alternative Transportation Technologies — Plug-in Hybrid Electric Vehicles. Technical Report. The National Academies Press, Washington, DC NRC, 2013. Transitions to Alternative Vehicles and Fuels. Technical Report, National Research Council. Ohnishi, H., 2008. Greenhouse gas reduction strategies in the transport sector: Preliminary report. Technical Report: OECD/ITF Joint Transport Research Center Working Group on GHG Reduction Strategies in the Transport Sector. OECD/ITF, Paris. 〈http://www.internationaltransportforum.org/pbu/pdf/08GHG.pdf〉, last visit March 15, 2013. Peterson, S., Michalek, J., 2012. Cost-effectiveness of plug-in hybrid electric vehicle battery capacity and charging infrastructure investment for reducing US gasoline consumption. Energy Policy 52, 429–438. Plotkin, S., Singh, M., 2009. Multi-Path Transportation Futures Study: Vehicle Characterization and Scenario Analysis. US Department of Energy, Washington, DC Plug in America, 2013. Understanding Electric Vehicle Charging. 〈http://www. pluginamerica.org/drivers-seat/understanding-electric-vehicle-charging〉 (accessed 20.03.13). Puget Sound Regional Council, 2008. Traffic choices study. Technical Report, 〈http:// www.psrc.org/assets/37/summaryreport.pdf〉 (accessed 15.03.13). Pas, E., Sundar, S., 1995. Intrapersonal variability in daily urban travel behavior: some additional evidence. Transportation (Netherlands) 22, 135–150. Toyota, 2013. Toyota Prius Plug-in Hybrid Specification. Available at 〈http://www. toyota.com/content/ebrochure/2013/priusplugin_ebrochure.pdf〉. USDOE, 2013. EV Everywhere Grand Challenge Blueprint. The United States Department of Energy, available at 〈http://www1.eere.energy.gov/vehiclesand fuels/electric_vehicles/pdfs/ eveverywhere_blueprint.pdf〉. USDOE and USEPA, 2013. Fuel Economy. Available at 〈http://www.fueleconomy. gov〉. Vyas, A., Santini, D., Johnson, L., 2009. Potential of plug-in hybrid electric vehicles to reduce petroleum use: issues involved in developing reliable estimates. Transp. Res. Rec. 2139, 55–63. Zhang, L., Brown, T., Samuelsen, S., 2011. Fuel reduction and electricity consumption impact of different charging scenarios for plug-in hybrid electric vehicles. J. Power Sources 196, 6559–6566. Zhang, L., Brown, T., Samuelsen, S., 2013. Evaluation of charging infrastructure requirements and operating costs for plug-in electric vehicles. J. Power Sources 240, 515–524.