Battery electric propulsion: An option for heavy-duty vehicles? Results from a Swiss case-study

Transportation Research Part C 88 (2018) 107–123

Contents lists available at ScienceDirect

Transportation Research Part C journal homepage: www.elsevier.com/locate/trc

Battery electric propulsion: An option for heavy-duty vehicles? Results from a Swiss case-study

T

⁎

Emir Çabukoglu, Gil Georges , Lukas Küng, Giacomo Pareschi, Konstantinos Boulouchos Aerothermochemistry and Combustion Systems Laboratory, ETH Zurich, Sonneggstrasse 3, 8092 Zurich, Switzerland Swiss Competence Center for Energy Research on Efficient Technologies and Systems for Mobility, Zurich, Switzerland

AR TI CLE I NF O

AB S T R A CT

Abbreviations: BEV Battery Electric Vehicle CNG Compressed Natural Gas ENTSO-E European Network of Transmission System Operators for Electricity GTE Goods Transport Survey ICE Internal Combustion Engine LNG Liquified Natural Gas LSVA Distance-specific Heavy-duty Vehicle Tax MPW Maximum Permissible Weight PHEV Plug-in Hybrid Electric Vehicle pkm Passenger-kilometer SOC State of Charge tkm Tonne-kilometer vkm Vehicle-kilometer Mt Megaton WHVC World Harmonized Vehicle Cycle

Road freight is the most energy-intensive freight mode (per tkm) and runs almost exclusively on fossil fuels. Electrification could change that, but can batteries really power actual heavy-duty operations? This study introduces a data-driven, bottom-up approach to explore the technical limits of electrification using real data from the entire Swiss truck fleet. Full electrification increased the total Swiss electricity demand by about 5% (3 TW h per year) over its current level and avoid about 1 megaton of CO2 per year (accounting for emissions of generation). Realizing this potential required (1) an allowance to exceed current maximum permissible weight regulations, (2) a high-capacity grid access for charging at the home-base (at least 50 kW ) and (3) a supporting intra-day energy infrastructure (we explored battery swapping). Boosting the gravimetric energy density of the battery cells was generally beneficial, but only effective if the aforementioned conditions were met. Thus, right now, battery electric trucks are no drop-in replacements for their Diesel counterparts. To allow their wide-spread usage, the road-freight sector would have to transform well beyond the vehicle. The required changes are substantial, but not unthinkable. Therefore, we think electric trucks deserve further exploration, in particular regarding their costs, life-cycle impact, technological variants and comparison to competing technologies.

Keywords: Road freight heavy-duty truck Battery electric propulsion Technology potential

⁎

Corresponding author at: Aerothermochemistry and Combustion Systems Laboratory, ETH Zurich, Sonneggstrasse 3, 8092 Zurich, Switzerland. E-mail address: [email protected] (G. Georges).

https://doi.org/10.1016/j.trc.2018.01.013 Received 5 September 2017; Received in revised form 15 January 2018; Accepted 15 January 2018 Available online 04 February 2018 0968-090X/ © 2018 Elsevier Ltd. All rights reserved.

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al. Annual transport performance Switzerland

1. Introduction 1.1. Why truck electrification is hard but important Heavy-duty road freight1 are the most energy-intensive freight mode (in units energy per tkm) (EEA, 2008). Since Diesel fuels virtually the entire fleet, it is an important CO2 emitter: trucks produce roughly 11% of the transportation CO2 emissions of Switzerland (BAFU, 2017), and more than 16% worldwide (Kahn Ribeiro et al., 2007). Passenger cars produce the lion share of emissions, but electrification could change that in the coming decades. Meanwhile, recent studies expect the demand for road freight to grow roughly twice as fast: + 33% (in tkm) until 2040, compared to a “mere” + 18% for motorized individual mobility (in pkm) (ARE, 2017); internationally, the International Transport Forum even expects a + 160% growth (in tkm) until 2050 (ITF, 2017). Not surprisingly, the International Energy Agency calls for rapid electrification of road freight (Teter, 2016). But the average Swiss heavy-duty vehicle is more than ten times heavier than the average car – in terms of curb weight, i.e. before adding any payload (Rohner, 2004). To achieve the same autonomy range, electric trucks must thus carry thousands of kilos of batteries, as opposed to hundreds in cars. To make matters worse, trucks generally drive much greater distances on a daily basis. From a purely technical perspective, this raises a fundamental question: are trucks electrifiable to a meaningful extent? 1.2. What has been done thus far (state of the art) We found no empirical work specifically on the electrifiability of trucks, presumably due to the lack of commercial products; see Section 1.2.2. However, there is a rich body of literature for passenger cars; see Section 1.2.4. A number of studies approached the topic from a market angle; see Section 1.2.3. We found no previous analytical work on the technological potential; but methodologies similar to our own have been used in environmental impact assessment, namely concerning CO2 and pollutant emissions; see Section 1.2.1. 1.2.1. Environmental impact assessment of current heavy-duty vehicles There are two types of assessment methodologies: bottom-up and the top-down. The bottom-up approach assesses the impact of each vehicle individually and then it sums up; the top-down approach uses aggregated data, such as fuel sales and average emission factors. Cai et al. (2012) compared the two methods. They concluded that the bottom-up approach incurs a high uncertainty from factors such as fuel type, driving behavior and road conditions. However, it allows more sophisticated analyses, as it considers what kind of vehicle causes them. McKinnon and Piecyk (2009) evaluated three different bottom-up methods against a top-down reference. The difference were the data sources: namely survey and counting station data for the distance, as well as survey, empirical and test-bench data on the fuelconsumption. The authors concluded that the survey-data was prone to under-reporting of trip distances; counting station data produced the most reliable estimates. Despite their weaknesses, bottom-up methods remain the preferred impact assessment tool for freight transport (Cai et al., 2012). We found examples from Finland, China, Canada, Turkey and Spain: Liimatainen and Pöllänen (2010) determined trends of energy efficiency in Finnish road freight using an energy demand model and statistical data. Yang et al. (2015) used questionnaires, GPS devices and an energy demand model to track driving behavior and compute CO2 emissions. Lukomskyj (2003) computed emissions in Alberta using traffic counts. Ozen (2013) and Burón et al. (2004) calculated emissions using COPERT, a modeling tool developed by the European Environment Agency (EEA). 1.2.2. Practical experience: commercially available electric truck products There are too few heavy electric road vehicles to draw conclusions for the entire freight sector: In public transport, trolley buses have been commonplace in cities for a long time; more recently, plug-in electric and all-electric solutions have been tested in many countries, including Switzerland (such as ABB’s electric bus in Geneva (ABB Communications, 2013)). In road freight, Siemens is currently testing its “eHighway” technology: direct-electric hybrid trucks, powered via overhead lines (Siemens, 2016). Pure and independent electric freight vehicles remain exotic: at the time of writing we identified the Swiss E-Force One (E-FORCE ONE, 2017), EMOSS (EMOSS Full Electric Truck, 2017) and Terberg electric trucks (Terberg Yard/Port Tractor, 2017). MercedesBenz (Mercedes-Benz Urban eTruck, 2017) and recently Tesla (Tesla Semi, 2017) announced products. 1.2.3. Uptake of electric propulsion in heavy-duty vehicles In its “commercial vehicles study” on the “future of German road freight and bus and coach travel up to the year 2040” (Adolf 1 We follow the Swiss legal definition: any automobile road-vehicle (a) dedicated to the transport of goods for commercial purposes and (b) whose maximum permissible weight exceeds 3.5 metric tonnes (Muncrief and Sharpe, 2015).

108

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

et al., 2016), Shell takes a (technically detailed) market angle – meaning the primary hurdle to change is market adoption, not technical limits: two adoption scenarios were evaluated using a fleet substitution dynamics model, tracking the light, medium and heavy duty (the European N1, N2 and N3 classes) fleets separately. The authors concluded that electrified propulsion would become important primarily to the light-duty sector; the medium duty sector would further be dominated by Diesel (88.4% of registrations); and about half of the heavy-duty fleet would be fueled by LNG. The project “KomDRIVE” focused more on technology, but it considered only vehicles up to 12 t maximum permissible weight (MPW) (Arndt et al., 2016). Based on surveys conducted among companies participating in the project, the authors found that most of the trucks under 12 t MPW would be electrifiable. 1.2.4. Bottom-up assessment of electrification potential of vehicle fleets When it comes to the technological potential and in particular the limits imposed by energy autonomy, research seem focused on passenger cars. There is a whole host studies that tracked current car usage (using GPS or surveys) and then checked whether BEVs (and in some cases, PHEVs) could sustain the same mobility. A first, large group, considered only the distance driven, not the energy demand: Pearre et al. (2011) tracked 484 passenger cars, and Jakobsson et al. (2016) used GPS data from over 7000 cars from Sweden and Germany, to see which ones can be substituted by BEVs. Khan and Kockelman (2012) followed 264 households in Seattle with GPS devices. They analyzed how many of them could satisfy their daily mobility needs with BEVs or PHEVs. Björnsson and Karlsson (2015) explored how far PHEVs would really drive in all-electric mode, using GPS data from 432 conventional Swedish cars, tracked over at least 30 days. In a subsequent publication (Björnsson and Karlsson, 2017), they also investigated if BEVs or PHEVs are more appropriate for the electrification of the two-car household based on the GPS data from 64 two-car households in Sweden. Christensen et al. (2010) generated a virtual fleet using a discrete choice model; they then investigated to what extent fast-charging is necessary in addition to home charging. Weiss et al. (2014) modeled the yearly mobility profile of the German car fleet using mobility surveys. They looked specifically for vehicles traveling less than 100 km per day. In our case, this approach is not feasible, because there are no vehicles from which we could derive the effective autonomy range. A second group of studies relied on energy demand models to estimate if vehicles are electrifiable: Greaves et al. (2014) based their analysis on GPS data from 166 vehicles; Neubauer and Wood (2014) used 317 year-long trip histories from the Puget Sound Regional Council’s Traffic Choices Study. Gnann et al. (2015) focused on passenger cars operated in commercial settings in Germany. Next to the technical electrification potential they also compared the total cost of ownership. This would work in our case, but the results of these studies apply only to a subset of the fleet. A last interesting group did essentially the same as the second group, but substituted the empirical tracking data with synthetic data from traffic flow simulations to cover the entire fleet. Namely the projects ARTEMIS (Waraich et al., 2014) and THELMA (Gassmann et al., 2014) relied on MATSim (Horni et al., 2016). Crossing the technological and market angles, they used fleet substitution dynamics models (similar to the Shell-study) to generate virtual fleets, containing a certain amount of BEVs. The latter were then analyzed using energy demand models and various impact assessment techniques. 1.3. Scope and goals of this study We explore the maximum penetration depth that battery-electric powertrain technology can achieve in the heavy-duty fleet under ideal conditions. We explicitly disregard the market and the substitution dynamics of the fleet – in contrast e.g. to the Shell study (see Section 1.2.3). Just as studies seen in Section 1.2.4, our analysis was based on an energy demand model. However, we covered the entire fleet using an empirical data source: Switzerland is one of only three countries in the world to electronically monitor every heavy-duty vehicle on its entire road network for tax purposes. That, to our knowledge, makes this is the only study covering an entire national heavy duty fleet. We will limit ourselves to battery-electric propulsion, since it is the most constraining and thus well illustrates the strengths of our approach. We will explore other technologies, in particular fuel cell systems and hybrids, in a subsequent publication. This paper is structured as follows: Section 2 explains how we combine the performance monitoring data with survey data to derive operation profiles for each vehicle in the fleet and how we assess whether such a profile allows for electrification. Section 3 discusses the data that went into our model. Section 4 illustrates our findings, and discusses the sensitivity of our findings on our assumptions. 2. Methodology 2.1. Overview Our method distinguishes itself from those seen in Section 1.2 in two key points: first, we resolve the varying amount of payload every vehicle carries on a daily basis for a whole year. Second, instead of summing up the energy demand, our aggregation procedure counts how many vehicles could still accomplish all 365 of their daily schedules with an electric drivetrains. This happens in three steps: 1. Section 2.2 explains how we derived the unique daily usage profiles of every individual vehicle in the fleet. 2. These usage profiles were then translated to mechanical energy using a vehicle dynamics model and the World Harmonized 109

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Vehicle Cycle (WHVC) (Silberholz et al., 2014). This step also involved characterizing the vehicle as a physical system. This is explained in Section 2.3. 3. Based on technical characteristics of the original vehicle, such as its nominal propulsion power, heuristics defined a substitute battery-electric powertrain. Section 2.4 details how we check if that powertrain could provide the required energy on every day of the annual vehicle usage profile. The outcome is the electrification potential of the fleet. Most of our findings are based on how that potential changed, if we varied certain boundary conditions, such as the gravimetric energy density of the battery cells or the charging behavior – see Section 2.5. Since we found battery swapping crucial, we also explored the resulting infrastructural requirements – see Section 2.6. 2.2. Reflecting vehicle usage – the demand side Vehicle usage is described by patterns: they consist of (1) the distance driven and (2) the average payload carried on each day2 of the model year. We used a data-driven approach, basing our profiles on the past performance of the actual fleet. Section 3 describes the structure and content of the datasets used; in this section, we discuss the process. 2.2.1. Classification of data-sources Generally, we distinguish four kinds of suitable data sources: Vehicle flow data. Data from a counting station network or traffic flow simulation can be used to estimate the fleet performance (in vkm). Such information is available in many countries (including Switzerland). However, it only provides aggregated results. Representative surveys. As part of governmental oversight, all European Union member states (and Switzerland) routinely survey their heavy duty fleets (Regulation (EU) No 70/2012). The results can serve on their own to study electrification potential. However, they do generally not follow the same vehicle for much more than a few consecutive days (7 in the case of Switzerland). Vehicle tracking systems are on-board data-loggers recording the position or instantaneous velocity of a vehicle at a frequency, high enough to resolve basic vehicle physics (typically ⩾ 1 Hz). This enables sophisticated analyses, but is expensive to deploy at a large scale; also payload tracking may still require surveys. Performance monitoring systems are data-logging systems similar to vehicle tracking systems, but with much lower temporal resolution. The low-tech variant are handwritten logbooks; their digital counterparts are e.g. fuel cards and remote fleet monitoring systems. Their key advantage is that they are already in place; but they tie the analysis to one operator and their infrastructure. 2.2.2. Generation of usage profiles We used a combination of performance monitoring and survey data: our account of the distance is based on data from a performance monitoring system; the Swiss Federal Government uses it to collect the heavy vehicle road tax “LSVA” (see Section 3.1). The data provides a very detailed and precise account of the distance each individual vehicle drove in the past. However, the system does not monitor the payloads carried. To estimate payloads, we used survey data from the Swiss Federal Office for Statistics. The survey, “GTE”, describes the goods flows generated by a representative subset of the fleet. We consider GTE representative of the daily average payload carried by certain types of vehicles; thus we assigned each LSVA record an average payload using conditional sampling. By necessity, we characterize vehicles according to the attributes shared by both the GTE and LSVA datasets, namely the maximum permissible weight, curb weight, body type and whether a trailer was towed. However, the daily average payload and the vehicle characteristics are not statistically independent; for example, heavier vehicles can carry more payload. Therefore GTE was clustered using a decision tree (Sklearn – Decision Tree Classifier, 2017), with payload as the target variable: the procedure groups vehicles of similar behavior, even if their characteristics do not match exactly. This minimizes the number of vehicle groups, ensures statistical representativeness and speeds up the sampling process. 2.3. From usage profiles to useful energy demand In this section, we move from usage data to energy. The basis is the longitudinal force equation Eq. (1):

Fprop (v (t )) =

1 dv (t ) ·ρ ·cD·Af ·v 2 (t ) + m ·g ·cr + m · 2 air dt

(1)

where:

• g is the standard acceleration due to gravity the ambient air density • cρ isisthe vehicle’s aerodynamic drag coefficient • air

D

2

In fact there can be multiple entries for one day: every time a vehicle attaches or detaches a trailer, the day is split in what we call “stages”.

110

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

• A is the vehicle’s frontal area • c is the tire rolling resistance coefficient • m is the total mass of the vehicle, consisting of the towing vehicle (curb weight), the trailer (curb weight or 0 if there is no trailer f

r

attached) and the payload – see Eq. (2)

m = mtowingvehicle + mtrailer + mpayload

(2)

g and ρair are ambient properties which marginally vary with the location of the vehicle. Otherwise, all parameters are vehiclespecific, depending primarily on its body-type, its maximum permissible weight (MPW) and the kind of towed trailer (if any). We grouped the Swiss fleet along those axes; for each group we defined one typical representative, whose parameter values we use for all vehicles in that group. Assuming purely dissipative braking, the distance-specific, useful mechanical energy required along a known speed signal v (t ) (here: the WHVC) is the positive propulsion work plus any non-propulsive work Eq. (3):

etotal =

∫WHVC Ptotal (t ) dt 1 ( (Fprop (v (t )))·Fprop (v (t )) + Paux ) dt = ·∫ d WHVC ∫WHVC v (t ) dt

(3)

where  is the Heaviside step function3; Paux is the total auxiliary mechanical power demand of the various non-propulsive subsystems, from air conditioning to power steering. Next to the type of vehicle, etotal depends on the variable payload; etotal must therefore be recomputed for every stage.

2.4. From useful to end energy demand (Diesel and electric) The powertrain must provide the necessary useful energy etotal from one (or several) mobile energy storage device(s), such as a tank or battery. Once depleted, the storage device must be replenished at a stationary energy source – a fuel or charging station. We call the exchanged energy “end energy”. We describe the powertrain using constant overall efficiencies: they relate the useful energy to the end energy from which it was extracted; Table 1 lists the assumed numerical values. As a first step, we compute the demand for Diesel of the unmodified vehicles. Then, as explained in Section 2.4.1, we redesign the each vehicle and check if it can still satisfy its driving missions.

2.4.1. Powertrain design rules In our modeling perspective, we remove the conventional powertrain – assuming everything else stays the same – and install a replacement battery-electric powertrain. The latter follows a fixed set of rules, which are exogenous to our model: 1. We assume that the electric motor has the same nominal power output as the original Diesel engine. This ensures comparable acceleration performance, gradability and top speed. 2. The gear-box becomes superfluous; its weight and that of the engine are discounted according to Table 1. 3. The AC electric machine is connected to the DC battery using an inverter. We assume their combined efficiency and power-specific weights as listed in Table 1. 4. The fuel-tanks become superfluous; the space they occupied now houses batteries. To not artificially reduce the electrification potentials, we use the largest available volumes in each market segment (see Table 2). In some vehicles, this would lead to an excessive weight increase. Therefore, the latter is capped at 5% of the original vehicle’s MPW. We assume that vehicles may still carry the same payload, based on paragraph 6 of the EU directive 2015/719: [The use of alternative powertrains] for heavy duty vehicles or buses may generate extra weight, but reduces pollution. That extra weight should not be counted as part of the effective load of the vehicle, since this would penalize the road transport sector in economic terms. However, the extra weight should not result in the load capacity of the vehicle being increased either. While the text specifies no limit to the “extra weight”, heavier vehicle do more damages to the road surfaces (Poulikakos et al., 2013) – not to mention the arising cost and safety concerns.4 We computed the weight of the battery pack using the gravimetric and volumetric energy densities in Table 1. These are the expected cell-level characteristics of the Panasonic 2170, the cell developed for the Tesla Model 3, derived from the Panasonic 18650 (Panasonic, 2012). We assumed the cylindrical cells are arranged in uniform triangular packing. A safety clearing radius around the cells limits the packing density to 70%; this matches the Tesla S battery pack, weighing 544 kg at an energy content of 85 kW h (Roper, 2016). The Heaviside step function  (x ) = 1 if x > 0 ; otherwise it is 0. The 5% threshold value is an assumption. We chose it simply because it puts the maximum weight increase (for the heaviest vehicle) at a reasonable-sounding 2t . We are not claiming that all future vehicle designs will follow this logic; we seek to establish a baseline to tell whether battery volume or mass is more constraining. 3 4

111

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Table 1 Technical properties of the components (or assemblies) used in powertrain design and energy demand computations. Quantity

Component

Weight-to-power ratio [kg/kW]

ICE + gear-box Electric motor Power electronics

Value

Source

2.17 0.80 0.10

Plotkin and Singh (2009) Plotkin and Singh (2009) Plotkin and Singh (2009)

Gravimetric density [W h/kg]

Battery (cell)

240

Plotkin and Singh (2009)

Volumetric density [W h/l]

Battery (cell)

727

Plotkin and Singh (2009)

Efficiency [–]

ICE + gear-box Battery charger Motor + power electronics

0.40 0.85 0.86

Neubauer and Wood (2014) Lin (2017)

Emission factors [g CO2 /kWh]

Diesel Swiss consumer mix CCGT ENTSO-E 2010 (European mix) ENTSO-E 2015 (European mix)

272.00 138.50 325.00 343.80 413.60

Röthlisberger (2014) Messmer and Frischknecht (2016) Willnow (2013) E-Control (2011) E-Control (2016)

Table 2 Largest available fuel tank capacity by vehicle category, as currently available in the Swiss market. We assume that this is the maximum space available for battery installation in electric trucks – although the volume may only be partially used (see Section 2.4.1). Note that the values reflect the product-portfolio of the market-leader in Switzerland at the time of writing (Mercedes-Benz LKW Konfigurator – Actros, 2017). For Euro VI trucks, the volume of the tank holding the reduction agent may be discounted as well, leading to somewhat larger tanks. As shown later in the sensitivity analysis, this has virtually no impact on electrifiability. Vehicle type

MPW [kg]

Fuel tank volume [l]

Rigid

3500–18,000 18,000–25,000 25,000–26,000 26,000–40,000

200 1260 1060 1260

Articulated

3500–18,000 18,000–23,300 23,300–25,000 25,000–26,000 26,000–40,000

200 1300 730 490 820

2.5. Determining the electrification potential The design rules yield the new curb-weight and nominal battery capacity for each vehicle in the fleet. If the vehicle became heavier, we recompute its energy demand (using Section 2.3). A vehicle is not electrifiable if any stage of its usage profile requires more electricity than the initial charge of the battery. 2.5.1. Charging behavior We assume that vehicles charge at night, for at most 12 h, at 50 kW – the typical power of a fast-charging station for cars (Falvo et al., 2014). For some vehicles that may be insufficient; which is why we varied charging power levels. Note that starting the day with a partial charge does not disqualify a vehicle from electrification; only the inability to complete its mission does. For many missions, more charging power is not the issue; the vehicle needs to recharge during the day. Given the size of truck batteries, fast-charging is not really an option. Therefore we considered battery swapping, modeled as an instantaneous, full recharge; the number of swaps per and vehicle was limited. 2.5.2. Evaluation The total CO2 emissions of the original Diesel fleet served as a validation point. We then explored the reductions resulting from (1) increasing the battery pack and cell densities (assuming technological improvements in the future), (2) increasing the charging power available at night, (3) allowing vehicles to swap their batteries during the day (assuming an infrastructure develops in the future) and (4) combining both battery swapping and technology improvements. We used the following aggregate indicators:

• the CO emissions avoided through electrification of the fleet, assuming CO -neutral electricity generation. It is a proxy of the avoided fossil fuel demand. In a few explicitly designated instances, we included upstream emissions from electricity production. • the share of vkm and tkm driven by battery electric trucks. 2

2

112

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

• the number of electrifiable vehicles, as a proxy for fleet operator’s investment in new vehicles. Multiplied by a presumed average • • •

50 kW charging power, this yields the worst-possible charging power demand – it occurs if all vehicles initiate their daily charging process approximately at the same time. the totally required electric charging energy at the end of each day, as a measure of the impact on the energy providers. Divided by 12 h , it yields the best-case overnight charging power. This emulates a perfect “smart-charging” scheme, distributing the charging load evenly among all vehicles. the aforementioned best-case and worst-case charging power as a proxy for grid–impact. The charging demand from batteries recharging in swapping stations comes on top. the total number of times that any vehicle swapped its battery on any given day, as a proxy for the number of additionally required batteries (beyond those already in the vehicles).

2.6. Battery swapping infrastructure requirements Contrary to what we assumed in Section 2.5.1, battery swapping is neither instantaneous nor ubiquitously available; vehicles lose time swapping and waiting for a free swapping station. We explored this temporal aspect, because it relates to the minimal infrastructural requirements. However, we disregarded the spatial component, by assuming that stations are spatially evenly distributed according to the actual demand. Our tool was a multi-agent, discrete event simulation: it goes through each vehicle’s day, simulating the interaction with swapping stations as they occur. This happens simultaneously for all vehicles, within the strict, deterministic rule set, illustrated in Fig. 1. Its core are two logical agent types: vehicles and stations. Each vehicle in the fleet is represented by one vehicle agent. Its mission is covering a pre-defined distance at a constant average speed. It keeps driving until it completes its mission or its battery runs dry. If the latter happens, the vehicle is presumed at a swapping station. The departure time, driving quota, battery capacity and specific energy demand are individually defined for each vehicle, according to its usage profile (see Section 2.3). The swapping stations are represented by a single agent: the manager. It is in charge of distributing incoming vehicles onto a finite set of swapping slots, each of which can service one vehicle at a time. Two queues form the heart of the manager: the holding bay is a first-in-first-out queue. All visiting vehicle agents register here on arrival. The second queue, the service area, holds all vehicle agents currently in service – it only accepts nslots of vehicles. The manager shuffles vehicles from the holding bay to the service area, until either the first is empty, or the second full. Vehicles remaining in the holding bay await their turn; vehicles in the service area stay for tswap seconds and then resume driving. −1 , reducing the operationally undesirable down-time for swapping to a We used this model to determine the minimal nslots and tswap reasonable limit. 3. Data used in the computations In the following, we review the data sources mentioned in Section 2.2.2. We used survey data to estimate the carried loads, and

Fig. 1. Flow-chart of the interactions between the swapping-station and vehicle agents of our discrete-event simulation, modeling congestion at the battery swapping infrastructure: when their battery runs dry, vehicles message the “swapping station” agent. It services maximally nslots vehicles simultaneously in its “service queue”. If the latter is full, new arrivals wait in the “queue” until a slot frees up. The time spent “at station” then exceeds tswap , the duration of the swapping process.

113

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

data from an automated performance monitoring system for the driven distance. In the following sections, these data sources are explained in detail. 3.1. Driven distance: “LSVA” performance monitoring data The Swiss federal government raises a road tax called “LSVA” from all5 heavy vehicles. The charge is proportional to the distance a vehicle drove within Switzerland; the rate depends on the maximum permissible weight and Euro class. To collect the tax, the Swiss Federal Customs Administration operates an automated performance monitoring facility. Its core are on-board logging devices; every vehicle must carry one. The device not only records performance, but also what kind of trailers were towed (because this may change the maximum permissible weight). While other countries operate similar systems,6 Switzerland is one of only three countries in the world to tax the usage of any of its roads, not just motorways. This makes “LSVA” a uniquely complete account of the activity of every vehicle in the fleet; Fig. 2a to c give a glimpse into the wealth of available information. 3.2. Carried payload: “GTE” survey data As part of a European commitment to the decarbonization of transport, all European Union member states and Switzerland routinely monitor their heavy duty fleets (Regulation (EU) No 70/2012). In Switzerland, this takes the form of a survey the Swiss Federal Office of Statistics conducts every year: they contact over 5000 truck-drivers, collecting technical information on their vehicles and the trips they took (including the spatial origin and destination, they type, form an weight of the goods transported, and of course the distance covered). The data can be used to compute various indicators, including distance-based metrics such as the transport performance; we only rely on the payload information.

4. Results and discussion In the following, we present a broad range of results produced by our model. Section 4.1 opens with the status quo, meaning conventional vehicles; the results serve as a reference, but also as a sanity check by comparison to other sources. In Section 4.8 we discuss the impact of modeling uncertainties. Then we dive into electrification: Section 4.2 defines the upper bracket, if all vehicles were electrified – regardless of whether that is feasible. The limitations of current-day technology are the subject of Section 4.3. Sections 4.4 and 4.5 explores the role of battery technology and energy infrastructures. Finally, Section 4.6 combines the two in settings leading to high degrees of electrification; we discuss the resulting fleet composition and infrastructural requirements. 4.1. Reference point: the current fleet We applied the procedure explained in Section 2 to compute the total distance driven, the transport performance and CO2 emissions of current fleet – see Table 3. The values match those of the Federal Statistical Office’s Goods Transport Survey (BFS, 2016). With regards to CO2 emissions, there are no official figures excluding foreign vehicles. So we computed the value in Table 3 ourselves: it is based on the “distance traveled” and HBEFA emission factors (Keller, 2010), weighted between the different road types (urban, rural and motorway) according to the WHVC. Note that, in 2015, the fleet consisted of more than 52,800 active vehicles, subject to LSVA. More than 99% of them were powered by conventional Diesel powertrains. That yields an average of almost 35,000 vkm and 233,266 tkm per vehicle per year.

4.2. Maximum CO2 mitigation potential To get an idea of what electrification of the heavy-duty fleet means, we asked for the electricity demand and CO2 reduction if the entire fleet was electrified; however, we did not care (yet) whether this is technically and practically feasible. To model full electrification, we let the vehicles swap their battery as many times as necessary, i.e. the number of daily swaps allowed is unlimited (see Section 2.5.1). We found a total electricity demand of 3.06 TW h per year (at the plug, i.e. disregarding transmission losses) – that’s more than 5% of the current electricity consumption of Switzerland (BFE, 2016). Fig. 3a illustrates how the electricity mix determines the CO2 avoidance: obviously, CO2 -neutral electricity leads to a full reduction (1.43 Mt ). The Swiss consumer mix of 2014 (Messmer and Frischknecht, 2016) yields a 70% reduction (1.01 Mt ). Beyond 470 g CO2 /kW h , emissions from electricity production exceed that of the original Diesel fleet. 5 There are exceptions, such as military and emergency services vehicles, buses in public transportation and certain types of self-propelled mobile machinery, such as truck-mounted cranes. 6 New Zealand, Switzerland, Austria, Germany, Czech Republic, Poland, Slovakia, France, Belgium and several states of the U.S.A. use public performance monitoring systems for heavy-duty vehicles.

114

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Fig. 2. Different visualizations of the daily driving distance of the Swiss heavy-duty fleet, according to LSVA in 2015.

Table 3 Comparing model results for the current, conventional fleet with reference values from the Swiss Federal Office of Statistics. All values are aggregates over the entire fleet and the year 2015. Distance traveled [vkm]

Transport performance [tkm]

CO2 emissions [Mt]

Our model

1836·106

12,092 · 106

1.433

Reference (GTE)

1853·106 −0.92%

12,317 · 106 −1.82%

−2.37%

Relative difference

1.468

4.3. Electrification using current-day technology Next we explored the potential of current-day technology – Section 2.4.1 defines its characteristics. Concerning charging behavior, we disallowed battery swapping and over-night charging was limited to 50 kW over 12 h . Under these conditions, only 12% of the vehicles (over 6500 units) were electrifiable – Table 4 provides more details. The CO2 avoidance was a disappointing 2.1%. The reason is simple: we used real data, reflecting the inhomogeneous usage profiles within the fleet; electrification picked the easy targets: lowperformance vehicles. Although their number is relatively large, they never contributed much CO2 to begin with. Thus their CO2 mitigation potential is low. Unsurprisingly, the resulting power demand was low too: Fig. 4a shows that the peak charging power demand in a completely unmanaged dumb-charging scheme (“worst case”) could go up to 200 MW ; perfect smart-charging (“best case”) brought that down to 25 MW – a fraction of a percent of the peak power demand in Switzerland (BFE, 2016). The total, annual charging energy demand was 64 GW h . 4.4. Considering improvements in gravimetric energy density of the battery cells Battery energy densities have continuously improved over the last decade and are expected to continue doing so (Van Noorden, 115

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Fig. 3. Avoided CO2 emissions against CO2 intensity of electricity and daily peak charging power demand from electric trucks in case of full electrification In both plots, the “(1) Best Case” corresponds to a perfect smart-charging scheme: it smoothly distributes the load over 12 h . The “(2) Worst Case” corresponds to a dumbcharging scheme, of all vehicles beginning to charge at the same time.

Table 4 Absolute numbers of the Swiss fleet by vehicle class, and relative share of electrifiable vehicles in each class using current day technologies (240 W h/kg cell density, no battery swapping, 50 kW over-night charging). Vehicle class

Vehicles In the fleet

Electrifiable

Rigid trucks Articulated trucks

42,434 10,385

14.6% 3.4%

3500 kg ⩽ MPW < 12,000 kg 12,000 kg ⩽ MPW < 18,000 kg 18,000 kg ⩽ MPW < 26,000 kg 26,000 kg ⩽ MPW < 32,000 kg MPW ⩾ 32,000 kg

5582 6060 20,483 9748 10,946

10.0% 6.0% 11.4% 18.9% 13.3%

All trucks

52,819

12.4%

Fig. 4. Daily peak charging power demand from electric trucks assuming current-day technology and no battery swapping. In both plots, the “(1) Best Case” corresponds to a perfect smart-charging scheme: it smoothly distributes the load over 12 h . The “(2) Worst Case” corresponds to a dumb-charging scheme, of all vehicles beginning to charge at the same time.

2014). This raises the question, how the electrification potential would improve in turn. We took it to the extreme, namely 2000 W h/kg – the density of a hypothesized future battery, believed to enable a 189 seat electric airliner (Hornung et al., 2013). Noticeably, even at 2000 W h/kg the model deemed only 70% of the vehicles electrifiable; the emissions were not even cut in half (Fig. 5a). As Fig. 5b clarifies, the reason was not a lack of energy, but power: at roughly twice today’s cell density, 50 kW no longer sufficed to charge the largest batteries in time; beyond 1000 W h/kg even 100 kW became limiting. Keep in mind that we are talking 116

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Fig. 5. Improving battery energy densities – no swaps allowed, charging power: 50 kW , packaging factor: 70% and CO2 -free electricity.

about sustained, overnight charging, not isolated fast-charging. The effects of the packaging factor and electricity mix on the avoided CO2 emissions are shown in Section A. Lastly, we explored the sensitivity to our design rules (see Section 2.4.1). All results we have seen thus far assumed the former fuel tank volume is filled up with batteries until it is full or the weight increase relative to the original vehicle (MPW) reaches 5% . Fig. 6 displays what happened when we imposed different limitations – curve 3 being our reference. Curve 4 strikes the eye: if we allowed no weight increase, the mitigation potential was practically zero for cell densities below 800 W h/kg . Lifting the weight restriction (curve 1) more than doubled the avoidance below 480 W h/kg , i.e. double the status quo. As densities increased, the relative difference between curves 1 and 3 reduced. Thus, exemptions in maximum permissible weight are a prerequisite for electric trucks – at least in the short to medium term. The juxtaposition of curves 2 and 3 revealed that volume was restrictive in its own right beyond 480 W h/kg , but far less so than mass. This suggests that chassis may not require fundamental redesign to house more batteries – at least not in the medium-term (see also Section 4.7). 4.5. Allowing battery swapping We also explored battery swapping as an alternative means of improving electrifiability, independently of innovation on the vehicle side. Fig. 7a shows the dramatic effect using current-day vehicle technology: a limit of three swaps per day produced an avoidance of over 50%; without swapping, that would require cell densities of 1000 W h/kg and 100 kW of charging power (see Fig. 5b). Fig. 7b clarifies that many vehicles did not need their three swaps: on the busiest day (27th of May), there were only 11,522 swaps – that is about 5 swaps for every 18 vehicles. Note that this is also the minimum number of extra batteries needed (in addition to those in the vehicles); their capacity would sum to 3.3 GW h . The distribution of the number of batteries swapped and the effect of the electricity mix on the avoided CO2 emissions is shown in Appendix B. Finally, Fig. 8 leads to the same conclusion on the sensitivity to the design rules, as we have seen to the cell density: if vehicles were not allowed to become heavier, the potential dropped almost to zero for 3 swaps and less. 4.6. Doing both to reach imposed targets (Considering batteries with higher energy densities and allowing battery swapping) We have seen that the gravimetric energy density and number of allowed swaps per day can each significantly improve

Fig. 6. Effect of allowance for additional vehicle weight for different battery energy densities on the extent of avoided CO2 emissions.

117

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Fig. 7. Allowing battery swaps – cell energy density: 240 W h/kg , charging power: 50 kW , packaging factor: 70% .

Fig. 8. Effect of allowance for additional vehicle weight for different number of daily swaps allowed.

Fig. 9. Improving battery energy densities and allowing battery swaps – charging power: 50 kW and packaging factor: 70% .

electrifiability. What if they were combined? We explored this using a full factorial evaluation of our model; Fig. 9a and b illustrates the results as contour plots. Note that the plots reflect 50 kW of charging power; we omitted plots for 100 kW because they (visually) only really differ beyond 600 W h/kg . The almost perfectly hyperbolic trade-off curves express the intuitive trade-off between the number of swaps and the battery-cell density: for instance, two swaps per day and 315 W h/kg were enough for 50% CO2 avoidance – the same was achieved by either 3 swaps (see Fig. 7a) or 900 W h/kg and 100 kW of charging power (see Fig. 5b). Note that we extended the definition of the iso-curves beyond integer values of swaps. Those partial swaps represent opportunitycharging during the day: we have seen that many vehicles that relied on swapping, finished their day with only marginally depleted 118

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

batteries. They may have been able to forgo the last swap, had they recharged their batteries during the day. Substituting more than fractions of swaps by opportunity charging is challenging though, given the large capacities (up to 550 kW h ) of the batteries. 4.7. Pathways to high degrees of electrification Fig. 9a lays out three points (A, B and C); Table 5 summarizes their characteristics; Table 6 reveals the fleet compositions they produced. We chose those points as possible milestones along two alternate trajectories of the Swiss fleet to full electrification. In this section, we explore the effort of reaching them. Milestone A, which we already encountered in Section 4.5, essentially applies the currently available technologies to their full potential: Carried by a well-developed battery swapping infrastructure, electric trucks cut emissions in half. At that point (in our model-perspective), innovation in battery technology or infrastructure could drive further improvement: milestone B assumes a 75% increase in battery gravimetric energy density; milestone C assumes double the number of allowed swaps. Both led to 90% CO2 avoidance and electric powertrains in all but 5% of the vehicles. The fleet configuration of milestones B and C was almost identical. A distinguished itself from B or C by the number of electrifiable articulated trucks: they are more often used for long-haul transport. Remarkably, trucks between 12 and 18 tonnes (MPW) had the lowest electrification potential in all scenarios. This was a direct consequence of the design rules (see Section 2.4.1): relative to their total weight, those vehicles have small fuel-tanks. The inverse logic granted very heavy vehicles the highest electrification potential. Presumably, vehicles designed from ground up for electric propulsion would even this out. Of course the vehicles are only half of it: Fig. 10a compares the maximum number of batteries swapped. As before, we consider it a proxy of the number of additionally batteries required. Remarkably, it remained almost constant, even though over 25% more trucks were electrified. B hence minimized the ratio of extra batteries to vehicles; C maximized it to over 1 battery for every 2 vehicles. This suggests a poor outlook for electrification without advances in battery technology. B and C drew about the same amount of charging power: idealized smart-charging (“Best Case”) caused a load of 700–800 MW, plug-and-charge (“Worst Case”) was beyond reasonable. Thus smart-charging is a must in both cases. 4.7.1. Congestion in battery swapping stations Another important distinction between the three cases was the swapping station infrastructure itself. We reflected its development in terms of the number of stations and their throughput (in vehicles per hour). We computed the time loss in the different cases using the discrete-event simulation explained in Section 2.6 – under the assumption that time loss is the only consideration. Fig. 11 presents the results as 30 min iso-swapping-time curves for each case: on the curves, no vehicle spent more than 30 min at the stations. Fig. 11 uncovers the intuitive trade-off between the number of slots and the throughput. We also added the curves representing different average velocity assumptions, namely 30 and 50 km/h : they bracket the spread of the true driving behavior around the 40 kph average velocity of the WHVC. The effect of jumping 10 kph is severe; however, such a large deviation is highly unlikely. To put things into perspective: Switzerland has 65 large motorway fuel stations at the time of writing (Erdöl-Vereinigung, 2017). If each installed two swapping stations, a throughput of at least 17 veh/h (± 5) would put the maximum waiting time under 30 min in cases A and B. Doubling the number of swaps (case C) more than doubled the required throughput (37 ± 7 veh/h ). That means that a swap had to be completed in under 100 s: this includes driving the vehicle in position, unfastening and removing the depleted battery, putting the replacement in place, securing it and driving the vehicle away – keeping in mind that many of those batteries are heavier than a fully loaded passenger car. The requirements of cases A and B were virtually identical, even though there were more electrified vehicles under B; we already discussed this in Section 4.7. Electrification could thus grow on a more or less stable infrastructure as batteries improve in gravimetric energy density. 4.8. Sensitivities to remaining technical parameters To ensure that uncertainty on the remaining technical parameters of Table 1 does not disrupt our conclusions, we reran all analyses using the extreme values displayed in Table 7. Since the effects were mild, we only report the encountered extrema of the CO2 avoidance. Generally, the effect was negligible for cases of very low or very high avoidance. This indicates that the parameters discussed in this section are not the driving forces behind electrification. Nevertheless, they had a noticeable effect on average case. That is why Table 7 only features case A, with its 50% CO2 avoidance. Parameters affecting the mechanical energy demand of the vehicles provoked the strongest deviations. Thus detail-improvements Table 5 Three pathways to high electrification. Case A relies on current-day technology and a moderate amount of swapping (as already discussed under Section 4.5). Case B presumes improved batteries to reach 90% avoidance. Case C does the same using current-day battery technology. Case A: Current-day B: Battery improvement C: Infrastructure

Cell density [W h/kg ]

Allowed swaps [–]

Charging power [kW]

CO2 avoidance [% ]

240 420 240

3 3 6

50 50 50

50 90 90

119

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Table 6 Distribution of fleet in three cases: Case A (50% of the CO2 emissions avoided): Three swaps allowed and a gravimetric energy density of 240 W h/kg , Case B (90% of the CO2 emissions avoided): Three swaps allowed and gravimetric energy density of 420 W h/kg , Case C (90% of the CO2 emissions avoided): Six swaps allowed and gravimetric energy density of 240 W h/kg . Type of vehicle

Number of vehicles

Share of electrified vehicles Case A

Case B

Case C

Rigid trucks Articulated trucks

42,434 10,385

82.6% 32.1%

96.6% 88.4%

96.7% 89.1%

3500 kg ⩽ MPW < 12,000 kg 12,000 kg ⩽ MPW < 18,000 kg 18,000 kg ⩽ MPW < 26,000 kg 26,000 kg ⩽ MPW < 32,000 kg 32,000 kg ⩽ MPW

5582 6060 20,483 9748 10,946

71.7% 50.5% 59.0% 86.4% 99.0%

93.3% 82.0% 94.5% 99.7% 99.9%

93.4% 82.5% 94.9% 99.7% 99.9%

All trucks

52,819

72.7%

95.0%

95.2%

Fig. 10. Effect on infrastructure in three cases: Case A (50% of the CO2 emissions avoided): Three swaps allowed and gravimetric energy density of 240 W h/kg , Case B (90% of the CO2 emissions avoided): Three swaps allowed and gravimetric energy density of 420 W h/kg , Case C (90% of the CO2 emissions avoided): Six swaps allowed and a gravimetric energy density of 240 W h/kg .

Fig. 11. Infrastructural needs in swapping stations for three cases shown in Fig. 9a and three different average velocity assumptions – waiting for max. 30 min per day.

e.g. of aerodynamics would help; on the other hand, with ± 5% points, they are not what enables electrification. The same holds for the efficiency of the electric powertrain, which we found to be the second most important parameter. Evidently, the weight-to-power ratios of the powertrain components are secondary to that of the battery. The efficiency of the battery charger, just as the efficiency of the ICE, scaled the end-energy demand; the effect on electrifiability was negligible. However, at battery densities exceeding 800 W h/kg , we observed ± 3 percent points avoided CO2 . The reason is that the plug-to-battery efficiency effectively reduces the power reaching the battery. For very large batteries, this can make the difference between a battery being rechargeable in 12 h or not.

120

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Table 7 Observed maximal sensitivities of the CO2 avoidance to uncertain technical parameters of Table 1 for case A. With 50% CO2 avoidance, this case was generally the most sensitive. Note that the effect is indicated in percent points, not percent. Quantity

Component

Value

Δ avoidance (in % points)

Weight-to-power ratio [kg/kW]

ICE + gear-box Electric motor Power electronics

2.17 ± 1.08 0.80 ± 0.40 0.10 ± 0.05

± 1% ± 1% ± 1%

Average efficiency [–]

ICE + gear-box Battery charger Motor + power electronics

40% 85% ± 5% 86% ± 5%

± 0% ± 0% ± 3%

Physical properties

Car body Fuel tank volume

best/worst + 5%

± 5% −0.5%

5. Conclusions and outlook We quantified the technological electrification potential of the Swiss heavy-duty fleet. Compared to passenger cars, we found truck electrification to be extraordinarily hard – however, to our surprise, not so hard that we would completely rule it out. We identified three conditions for a high electrification potential – they do not apply to passenger cars, at least from a technical perspective (Waraich et al., 2014): the first and foremost is an exemption from maximum permissible weight regulations. If electric trucks have to operate within the same legal confines as conventional trucks, the achievable CO2 avoidance do not justify the effort. Second, the home bases of electric trucks require a high-capacity grid connection: currently, a minimum of 50 kW per vehicle is to be expected. At high penetration levels, some form of coordinated “smart charging” becomes unavoidable. Finally, truck electrification requires a supporting intra-day energy infrastructure, such as swapping stations – no matter the gravimetric energy density of the batteries. With double the gravimetric energy density and roughly 130 swapping stations (two for every motorway fuel-station in Switzerland), full electrification moves into reach. The technical requirements of such a swapping station would be enormous though: it would have to house and charge at least one hundred batteries; during peak hours, swapping even the heaviest batteries (up to 3 t ) must complete within 3.5 ± 1 min. In conclusion, electric trucks are clearly no drop-in replacements for their Diesel counterparts – as opposed to passenger cars. Unless somebody starts building the necessary infrastructure, they will probably never be more than a niche product – it is not just a matter of technology evolution. W hether or not that transformation will come is a different question. Yet the need for action is real, as evidenced by the mounting political concerns about the sustainability of road-freight (see Section 1); Compared in particular to its big EU neighbors, Switzerland transports a rather low share of its freight by road (EC, 2017). It is also significantly smaller. In direct, international comparison, Switzerland’s infrastructure requirements could thus be rather benign. Nevertheless, battery-electric trucks shouldn’t be dismissed out of hand – at least not without proper assessment. To the best of our knowledge, this is the first quantitative assessment of the technological potential of truck electrification on a complete national fleet. As this showed, the operation profiles of heavy-duty vehicles are highly diverse, even though there are significantly fewer trucks than cars – the ratio is 1 truck to 83 cars in Switzerland (Rohner, 2004). Thus analogical reasoning, mean-value based analyses and observations on small samples can therefore be deceptive. There remain many opportunities for further research: Would the costs of such an endeavor be prohibitive? Does the cradle-tograve impact of such large batteries outweigh the operational CO2 avoidance – in particular given the potentially faster battery degradation in trucks? Could “electrified highways” eliminate the need for battery swapping? Are there other ways to exploit economies of scale (or rather density) of the energy infrastructure? Would fleet and logistics managers from various sub-sectors really opt for electric trucks? And of course, what happens if competing technologies, namely hybridization and fuel-cells, are considered? Acknowledgments The authors gratefully acknowledge the financial support of the Swiss Federal Office of Energy (BFE) [SI/501311-01] and the Swiss Competence Center for Energy Research in Efficient Technologies and Systems for Mobility (SCCER mobility), funded by the Swiss Commission for Technology and Innovation (CTI). We thank the Federal Statistical Office (BFS), the Federal Roads Office (ASTRA) and the Federal Customs Administration (EZV) for providing the different datasets. We thank Michael Hugentobler for operationalizing the GTE dataset. We thank Jonas Bütikofer for the intial cataloguing of the vehicle fleet. We thank Prof. Dr. Andrea Vezzini and Dr. Alejandro Santis for sharing their knowledge of battery electric propulsion technology. Appendix A. Considering improvements in gravimetric energy density of the battery cells – effects of the packing factor and the electricity mix on the avoided CO2 emissions Fig. 12a demonstrates how tighter cell packing improved CO2 avoidance, but high battery cell densities still outweighed the effect. Note that we chose the 90.7% as an absolute maximum: it presumes a pack composed only of cells, arranged in a maximally tight hexagonal lattice (Pickover, 1989). The demands on the cell walls would be extraordinary: they would be the sole structural element,

121

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Fig. 12. Improving battery energy densities – no swaps allowed, charging power: 50 kW , packaging factor: 70% and CO2 -free electricity if not stated otherwise.

contain thermal dilation and provide all the cooling. Fig. 12b confirms that a high electrifiability only translated to high CO2 avoidance if the upstream emissions were minimal. Appendix B. Allowing battery swapping – the distribution of batteries swapped in different weekdays and the effect of the electricity mix on the avoided CO2 emissions Fig. 13a reveals that the demand for swaps dropped to almost zero on weekends. Also it exhibited similar seasonal fluctuations as the driving demand (see Fig. 2c). As in the previous cases, a low CO2 -intensity mix was a requirement for high mitigation potentials (see Fig. 13b).

Fig. 13. Allowing battery swaps – cell energy density: 240 W h/kg , charging power: 50 kW , packaging factor: 70% . In figure (a), the boxes show the interval between the upper and lower quartiles while the horizontal line in the box shows the median. The vertical line begins at 10th and ends at 90th percentile.

References ABB Communications, 2013. Large-capacity, Flash-charging, Battery-powered Pilot bus takes to the Street. < http://www.abb.com/cawp/seitp202/ eae3e161cc498140c1257b7b00552bcd.aspx > . Adolf, J. et al., 2016. Diesel Oder Alternative Antriebe – Diesel Oder Alternative Antriebe – Womit Fahren Lkw Und Bus Morgen?. < https://www.shell.de/medien/ shell-publikationen/shell-nutzfahrzeug-studie.html > . ARE, 2017. Perspektiven des schweizerischen personen- und güterverkehrs bis 2040. Technical Report. Arndt, W.-H., Döge, N., Marker, S., 2016. Elektrifizierungspotential kommerzieller Kraftfahrzeug-Flotten im Wirtschaftsverkehr als dezentrale Energie-Ressource in städtischen Verteilnetzen: komDRIVE. Universitätsverlag der TU Berlin. BAFU, 2017. Entwicklung der Emissionen von Treibhausgasen seit 1990. < http://www.bafu.admin.ch/bafu/de/home/themen/klima/daten-indikatoren-karten/ daten/treibhausgasinventar.html > . BFE, 2016. Schweizerische Elektrizitätsstatistik 2015/ Statistique suisse de l’ électricité 2015, Technical Report. < https://www.bundespublikationen.admin.ch/ cshop_mimes_bbl/8C/8CDCD4590EE41EE79781873EBED9E3AB.pdf > . BFS, 2016, GTE, Technical Report. < https://www.bfs.admin.ch/bfs/de/home/statistiken/mobilitaet-verkehr/erhebungen/gte.assetdetail.7501.html > .

122

Transportation Research Part C 88 (2018) 107–123

E. Çabukoglu et al.

Björnsson, L.H., Karlsson, S., 2015. Plug-in hybrid electric vehicles: How individual movement patterns affect battery requirements, the potential to replace conventional fuels, and economic viability. Appl. Energy 143, 336–347. Björnsson, L.H., Karlsson, S., 2017. Electrification of the two-car household: PHEV or BEV? Transp. Res. Part C: Emerg. Technol. 85, 363–376. Burón, J.M., López, J.M., Aparicio, F., Martín, M.Á., García, A., 2004. Estimation of road transportation emissions in Spain from 1988 to 1999 using COPERT III program. Atmosph. Environ. 38 (5), 715–724. Cai, B., Yang, W., Cao, D., Liu, L., Zhou, Y., Zhang, Z., 2012. Estimates of China’s national and regional transport sector CO2 emissions in 2007. Energy Policy 41, 474–483. Christensen, L., Kveiborg, O., Mabit, S.L., 2010. The market for electric vehicles – what do potential users want? In: 2010 12th World Conference on Transport Research (WCTR), pp. 1–25. E-Control, 2011. Gesamtaufbringung nach ENTSO-E 2010. Technical Report. < https://www.e-control.at/documents/20903/-/-/8c1acae6-16b1-40e8-bf8696a3082e4e9f > . E-Control, 2016. Gesamtaufbringung nach ENTSO-E 2015. Technical Report. < https://www.e-control.at/documents/20903/26393/ENTSOE2015Gesamtjahr201604-04.pdf/e7e7eb5e-d947-45ce-9314-b8da5d13def8 > . EC, 2017. Europe in figures: Eurostat yearbook. Publications office of the European Union, Luxembourg. EEA, 2008. Energy Efficiency. < https://www.eea.europa.eu/publications/ENVISSUENo12/page027.html > . E-FORCE ONE, 2017. < http://eforce.ch/ > . EMOSS Full Electric Truck, 2017. < http://www.emoss.nl/en/electric-vehicles/full-electric-truck/ > . Erdöl-Vereinigung, 2017. Erdöl-Vereinigung Jahresbericht 2016. Technical report. < https://www.erdoel.ch/de/publikationen/jahresbericht/jahresbericht-2016 > . Falvo, M.C., Sbordone, D., Bayram, I.S., Devetsikiotis, M., 2014. EV charging stations and modes: International standards. In: 2014 International Symposium on Power Electronics, Electrical Drives, Automation and Motion, pp. 1134–1139. Gassmann, F., Lutz, P., Elber, U., 2014. Annual Activity Report 2014. CCEM – Project Report 2014, pp. 17–20. Gnann, T., Plötz, P., Funke, S., Wietschel, M., 2015. What is the market potential of plug-in electric vehicles as commercial passenger cars? A case study from Germany. Transp. Res. Part D: Transp. Environ. 37, 171–187. Greaves, S., Backman, H., Ellison, A.B., 2014. An empirical assessment of the feasibility of battery electric vehicles for day-to-day driving. Transp. Res. Part A: Policy Pract. 66 (1), 226–237. Horni, A., Nagel, K., Axhausen, K.W., 2016. The Multi-Agent Transport Simulation MATSim. Ubiquity, London. Hornung, M., Isikveren, A.T., Cole, M., Sizmann, A., 2013. Ce-Liner – Case Study for eMobility in Air Transportation. In: 2013 Aviation Technology, Integration, and Operations Conference. American Institute of Aeronautics and Astronautics, Reston, Virginia, pp. 1–11. < http://arc.aiaa.org/doi/10.2514/6.2013-4302 > . ITF, 2017. ITF Transport Outlook 2017. Technical Report. < http://www.oecd.org/about/publishing/itf-transport-outlook-2017-9789282108000-en.htm > . Jakobsson, N., Gnann, T., Plötz, P., Sprei, F., Karlsson, S., 2016. Are multi-car households better suited for battery electric vehicles? – Driving patterns and economics in Sweden and Germany. Transp. Res. Part C: Emerg. Technol. 65, 1–15. Kahn Ribeiro, S., Kobayashi, S., Beuthe, M., Gasca, J., Greene, D., Lee, D.S., Muromachi, Y., Newton, P.J., Plotkin, S., Sperling, D., Wit, R., Zhou, P.J., 2007. Transport and its infrastructure. Climate Change 2007: Mitigation of Climate Change, pp. 323–385. Keller, M., 2010. Pollutant Emissions from Road Transport, 1990 to 2035. BAFU (Bundesamt für Umwelt), p. 131. Khan, M., Kockelman, K.M., 2012. Predicting the market potential of plug-in electric vehicles using multiday GPS data. Energy Policy 46 (January), 225–233. Liimatainen, H., Pöllänen, M., 2010. Trends of energy efficiency in Finnish road freight transport 1995-2009 and forecast to 2016. Energy Policy 38 (12), 7676–7686. Lin, Z., 2017. The Energy Consumption and Cost Savings of Truck Electrification for Heavy-duty Vehicle Applications, vol. 495. Lukomskyj, B., 2003. Quantification of Greenhouse Gases Produced by the Road Transportation Sector in Alberta Using a Traffic Volume Methodology. Technical Report August. Alberta Infrastructure and Transportation. McKinnon, A., Piecyk, M., 2009. Measurement of CO2 emissions from road freight transport: a review of uk experience’. Energy Policy 37 (10), 3733–3742. Mercedes-Benz LKW Konfigurator – Actros, 2017. < https://toc.mercedes-benz.com/LKW_Konfigurator_TOC_de1/toc.dll? > . Mercedes-Benz Urban eTruck, 2017. < https://www.daimler.com/products/trucks/mercedes-benz/urban-etruck.html > . Messmer, A., Frischknecht, R., 2016. Umweltbilanz Strommix Schweiz 2014. treeze Ltd, Fair Life Cycle Thinking. Muncrief, R., Sharpe, B., 2015. Overview of the heavy-duty vehicle market and CO2 emissions in the European Union. International Council on Clean Transportation Working Paper (December), 1–14. Neubauer, J., Wood, E., 2014. The impact of range anxiety and home, workplace, and public charging infrastructure on simulated battery electric vehicle lifetime utility. J. Power Sources 257, 12–20. Ozen, M., 2013. Estimation of Road Freight Transportation Emissions in Turkey (Ph.D. Thesis). MIDDLE EAST TECHNICAL UNIVERSITY. Panasonic, 2012. Panasonic NCR18650A Data Sheet. Technical Report. Pearre, N.S., Kempton, W., Guensler, R.L., Elango, V.V., 2011. Electric vehicles: How much range is required for a day’s driving? Transp. Res. Part C: Emerg. Technol. 19 (6), 1171–1184. Pickover, C.A., 1989. Circles which kiss: a note on osculatory packing. Comput. Graph. 13 (1), 63–67. Plotkin, S., Singh, M., 2009. Multi-Path Transportation Futures Study: Vehicle Characterization and Scenario Analyses. Argonne National Lab. Poulikakos, L.D., Heutschi, K., Soltic, P., 2013. Heavy duty vehicles: impact on the environment and the path to green operation. Environ. Sci. Policy 33, 154–161. Rohner, H., 2004. MOFIS – das Motorfahrzeuginformationssystem der Eidgenössischen Fahrzeugkontrolle. < http://www.experience-online.ch/de/9-case-study/ 2023-mo?s > . Roper, D., 2016. Tesla Model S. < http://www.roperld.com/science/TeslaModelS.htm > . Röthlisberger, R., 2014. Faktenblatt CO2 -Emissionsfaktoren des Treibhausgasinventars der Schweiz. BAFU – Bundesamt für Umwelt – Abteilung Klima. Siemens, 2016. World’s First eHighway opens in Sweden. < https://www.siemens.com/press/en/pressrelease/?press=/en/pressrelease/2016/mobility/ pr2016060319moen.htm > . Silberholz, G., Six, C., Fredriksson, J., 2014. GRPE-HDH Research Project. < https://wiki.unece.org/download/attachments/18546752/HDH-17-07e.pdf?api= v2%0A > . Sklearn – Decision Tree Classifier, 2017. < http://scikit-learn.org/stable/modules/tree.html > . Terberg Yard/Port Tractor, 2017. < http://www.terbergbenschop.nl/en/products/tractors/yard-tractors/yt202-ev/ > . Tesla Semi, 2017. < https://www.tesla.com/semi/ > . Teter, J., 2016. Energy Technology Perpectives 2016 – Sustainable Transport in an era of Urbanization. Technical Report. IEA. < https://www.iea.org/media/ workshops/2016/egrdtransportsystemsofthefuture/2EGRD_Teter_25_10_2016.pdf > . Van Noorden, R., 2014. A better battery. Nature 507 (7490), 26–28. Waraich, R.A., Georges, G., Galus, M.D., Axhausen, K.W., 2014. Adding Electric Vehicle Modeling Capability to an Agent-Based Transport Simulation. Weiss, C., Chlond, B., Heilig, M., Vortisch, P., 2014. Capturing the usage of the German car fleet for a one year period to evaluate the suitability of battery electric vehicles – a model based approach. Transp. Res. Procedia 1 (1), 133–141. Willnow, K., 2013. Energy Efficiency Technologies ANNEX III Technical Report for Thermal Power Plants. WEC Knowledge Network (August). Yang, X.F., Liu, H., Man, H.Y., He, K.B., 2015. Characterization of road freight transportation and its impact on the national emission inventory in China. Atmosph. Chem. Phys. 15 (4), 2105–2118.

123