Can Charging Infrastructure Used Only by Electric Taxis Be Profitable? A Case Study From Karlsruhe, Germany

Electric vehicles are a promising instrument for reducing greenhouse gas emissions and other local pollutants, such as nitrogen oxides. Especially the taxi sector is a promising field of application. First, their high driving distances are necessary for electric vehicles to economize against conventional vehicles. Second, taxis are operated predominantly in city centers, where the pressure to reduce local emissions is greatest. However, while the electrification potential of taxi fleets has gained great attention in literature, the question of whether charging infrastructure can be profitable if only used by taxis has not been answered yet. To answer this question, we analyze 161 taxis in Karlsruhe, Germany. We find that a high share of electric taxis would come at lower cost than their diesel counterparts. Thus, a 25-45% share of taxi driving could be electrified, if only the taxi user perspective would be taken into account. However, while ten charging sites would be necessary to do so, only one charging site could be refinanced through exclusive taxi use. Consequently, the electrification potential would fall to about 3%, if only this one profitable charging site was built. Yet, in the future, electric driving will become cheaper. In 2025, already 55% electrification potential would be possible, given that both electric driving and charging infrastructure operation would be profitable. Altogether, while currently taxi charging infrastructure would require public funding or other business models, in the medium to long term, an exclusive use by taxis would be sufficient.

SECTION I. Introduction

Electric vehicles (EV) are a promising solution to decarbonize the transport sector as well as to reduce local pollution in cities [1]. While their potential to electrify passenger car travel is well understood for privately owned cars (e.g., [2], [3]), especially the commercial sector is an interesting first market for electric vehicles in Germany due to their regular driving and their high share in new vehicle sales [4].

Nevertheless, driving behavior within the commercial sector is inhomogeneous and the taxi sector might be peculiarly suited for electrification. Taxis are operated within a limited driving radius in areas with high population density and their high daily driving distances allow economizing the high investment of EV against conventional vehicles. However, the limited driving ranges and long charging times may be greater challenges for taxi use [5]. Although electric taxis (eTaxis) are used in a growing number of cities, in particular in China [6], the question of how both electric taxis and their charging infrastructure can be profitable, is not sufficiently understood. Accordingly, to the best of the authors' knowledge, a rollout of charging infrastructure, which would both enable broad taxi electrification and a profitable operation, has not yet been analyzed.

A. Existing Literature and Scope of This Study

Electric taxis are applied in various cities, with a broad rollout of large electric taxi fleets in Chinese Megacities [7]. Although electric vehicles are found to be suitable for taxi application in principle, very often a lack of an appropriate charging infrastructure is stated. Yet, charging infrastructure requires high investments and currently, electric taxis often have to rely on public charging infrastructure. Accordingly, taxi drivers are dissatisfied with charging infrastructure availability as often discussed in media, e.g., for Washington, D.C. [8], Shenzen [9], Amsterdam [10] or Frankfurt [11].

Given the starting diffusion of electric taxi fleets, there is a still limited number of scientific literature analyzing electric taxi data. For example, Tian et al. [12] compare the operational patterns of conventional taxis with 600 electric taxis in the Chinese cities Shenzen, Beijing and Hangzhou and find that electric taxi operation can be profitable given supportive policies. Hagman and Langbroek [6] analyze conditions for electric taxi driving in the Greater Stockholm region (Sweden) and underline the importance of specific support such as companies booking preferably zero emission taxis or a priority queuing at taxi stands to compensate for charging times. Accordingly, the authors find systematic differences in driving patterns of electric and conventional taxis. In their economic analysis, eTaxis show a higher profit margin than conventional vehicles, given the above mentioned support instruments. Based on electric taxi data from Daejeon, South Korea, Baek et al. [13] emphasize the efforts needed to rollout charging infrastructure as well as financial incentives to make electric taxi driving profitable and attractive. Kim et al. [14] analyze driving and charging data from ten electric taxis in Seoul, South Korea. The authors find most taxis to charge less than 2.5 times per day. Charging frequency shows a pronounced peak in the late evening and night and a smaller peak at noon. From their analysis, the authors conclude that taxis use preferably charging infrastructure in the city center and thus, these stations should be revamped to meet taxi charging demand. Rao et al. [15] analyze charging behavior of 129 electric taxis in Shenzen, China. Among other, their results show a dependency of the charging duration on charging start time. In addition, one taxi is seldom at the same charging station more than five times a day. Zou et al. [16] analyze electric taxi data from Beijing and find average electric taxi drivers to charge two times per day, with average daily driving distances of ∼120 km. Their driving data was recorded in 2013/2014, with vehicle all electric range well below 200 km.

A large body of literature simulating electric taxi driving based on conventional data supports the electrification potential of taxi fleets. One group of these studies focuses on the vehicle side and takes charging infrastructure availability as one important constraint. Yang et al. [5] analyze the potential to replace conventional taxis with battery electric vehicles (BEV) in Nanjing, China. The authors find that a broad charging infrastructure coverage, especially for fast charging, is needed for high electrification rates. However, also the use of taxi apps to facilitate customer acquisition may increase vehicle utility and thus might be interesting to increase BEV acceptance. Similarly, Fraile-Ardanuy et al. [17] analyze mobility data from San Francisco, USA. With battery capacities below 30 kWh, already 75% of the taxis could be electric and would lead to global cost reductions of 20%. For New York City, USA, Hu et al. [18] state a need for more public chargers that would allow electrifying more than half of the taxi fleet with BEV. Deyang et al. [19] use simulation to determine techno-economically optimal battery sizes for electric taxis based on taxi data from Shanghai, China. With their general approach, the authors find a suitable size of 60 kWh. In conclusion, in Yang et al. [20], the authors focus their analysis of taxi driving data (Changsha, China) to deduce policy recommendations. According to the authors, maintaining vehicle purchase subsidies is important, just as it is the support for public charging. Finally, the authors highlight the importance of a diversified BEV fleet due to different range requirements. Wang et al. 2015 [21] found electric taxis not being cost competitive in Shanghai without financial support, given the circumstances at that time. However, their results also showed a higher profitability of longer range BEV compared to lower range BEV. For small cities, Bischoff & Maciejewski [22] find electric taxis being able to provide taxi operation at a high service level. However, while the authors consider dispatching electric taxis by highest SOC rather than by first-in first-out to optimize battery charging, they do not provide an economic analysis.

In addition, there is a wide range of studies that use data of conventional taxis to determine suitable charging sites for a potential electric taxi fleet. Yang et al. [23] use dwell places of taxis in Changsha, China, for potential charger allocations. Analogously, Cai et al. [24] use taxi parking patterns from Beijing, China, to optimally site charging stations. The authors find the taxis' charging demand to overlap with the cities peak power demand. These studies often focus on the allocation of charging stations and their effect on electrification potential, but do not consider charging infrastructure profitability. In addition, often a given budget of a certain number of charging stations is used, e.g., as in [25] for Seoul, South Korea, in [26] for Vienna, Austria or in [27] for Beijing, China.

Altogether, existing studies on electric taxis often focus on either the vehicle or the charging infrastructure perspective. In addition, while the technical potential of eTaxis is frequently addressed, techno-economic analyses are rare. Since electric taxis have different charging infrastructure requirements than e.g., electric passenger cars or electric buses, findings from studies in these fields are not directly transferable. Accordingly, there is a research gap with regard to a combined techno-economic assessment of electric taxis and their charging infrastructure. The aim of this paper is thus to evaluate the electrification potential of electric vehicles (EV) in the taxi sector in Karlsruhe, Germany, by taking both a vehicle and a charging infrastructure perspective. We analyze, if today´s EV are technically and economically suitable for taxi operation in Karlsruhe under today´s and future conditions and if the charging infrastructure needed can be profitable if only used by these taxis. This allows deriving specific recommendations for action with regard to the rollout and political funding of charging infrastructure for electric taxis.

Here, we assume BEV taxis to be operated equally to non-BEV taxis, instead of assuming trip reallocation. This is mainly due to two reasons. First, we aim to analyze the electrification potential separately from rearranging trips (for example using app-based solutions) that also allow for a more efficient use of today's conventional taxis (e.g., because less cars are needed). Second, in Germany in general and in Karlsruhe in particular, the competition of companies with often not more than one taxi makes trip reallocation difficult.

SECTION II. The Taxi Sector in Germany and Karlsruhe

The taxi sector in Germany is part of the commercial branch “H - Transportation And Storage” (according to the Eurostat definition, see [28]). In this branch 40,590 vehicles were registered in Germany in 2017 [29] resulting in a vehicle stock of 208,963 vehicles (1.1.2018, [30]). For comparison: the total passenger car stock in Germany is 46 million. As taxis are not reported separately in these statistics, we use data from a special survey conducted by the Federal Ministry of Transport, Building and Urban Development [31]. There were 53,302 taxis in Germany in 2016, representing 36% of the vehicle stock in Sector H. Most of these taxis (87%) have less than eight seats.

The majority of the taxi companies in Germany owns only one or two taxis. In 2016, 73.6% of all 20,932 taxi companies owned only one car and another 11.6% owned two taxis. Accordingly, 14.3% of the companies own three or more taxis, resulting in 11.1 taxis per company on average.

The number of people transported by taxis has been very constant since 2010 and amounted to 431 million people in Germany in 2017. This resulted in a transport capacity of 2,917 million passenger kilometers. The taxi density varies largely among cities. In Karlsruhe, there is one taxi for 1,485 inhabitants, which corresponds well to the national ratio of 1,550 [32]. Currently, there is no battery electric taxi operating in Karlsruhe.

SECTION III. Taxi Driving Data

This case study results from a close cooperation with Taxi-Funkzentrale Karlsruhe, the largest taxi operator in Karlsruhe: 165 of 210 taxis are operated by Taxi-Funkzentrale. In total, the anonymized booking data of 161 taxis could be used (all diesel), representing 77% of the total 2017 taxi fleet in Karlsruhe. The data set contains start time, end time and distance of each trip as well as the information at which taxi stand the taxi is idling. We assume a taxi at a taxi stand if within 200 meters. The data was recorded during one month in summer 2017 (15. July 2017 to 14. August 2017). During this period, the taxis travelled 740,000 km in total. Summary statistics of the dataset can be found in Table I.

TABLE I Summary Statistics of the Taxi Driving Data

The driving data were pre-processed. We corrected the data set for GPS errors, which reduced the number of 8,536,000 data points in the original data set by approximately 13%. Since there is no timetable information in the original data set, we set a new trip if the time between two data points with zero velocity was more than five minutes. We corrected idle times when the location of these two data points was different. We then reduced the idle time by the time it takes to travel the route at an assumed constant speed of 50 km/h.

SECTION IV. Modelling Approach

In this work, we analyze the techno-economic feasibility of electric taxi driving, given that charging infrastructure, exclusively built for electric taxis, must be profitable. Thus, the research target is to quantify the potential share of electric taxis in the fleet and the corresponding number of profitable charging stations needed. To this aim, we first analyze taxi mobility patterns and determine the technical feasibility of electric driving for each driving profile, presuming different charging infrastructure rollout stages. Second, we determine the individual total cost of ownership (TCO) of electric driving compared to driving a conventional taxi. Third, based on the TCO results from the previous modeling step, we calculate a monthly basic fee to be paid by each electric taxi to make the operation of the charging stations profitable. Finally, we can assess, whether lower TCO of eTaxi driving might compensate for charging infrastructure cost. We present the different methodologies used for the individual modeling steps separately in the following sections.

A. Analysis of Technical Potential

1) Electric Driving Simulation

For electric driving, we simulate the battery state of charge (SOC) of the EV for every data point. We simulate different EV models using vehicle model specific parameters, as given in Table II. We assume a constant average energy demand for all driving situations. Accordingly, the SOC decreases proportionally to the distance driven of each trip D [km] (with departure time t and arrival time t + 1) and the EV model specific energy consumption $EC_{M}$ [kWh/km]:

$$\begin{equation*} SOC\left({t + 1} \right) = \ SOC\left(t \right) - \ D\ \cdot \ E{C_M}\ \cdot \ {K^{ - 1}}.\tag{1} \end{equation*}$$

TABLE II Technical Specifications of the EV Models Analyzed

$K$ describes the usable battery capacity (as it is given by EPA range and energy consumption) in kWh. Thus, a SOC safety buffer is implicitly taken into account, as designed by the vehicle manufacturer. Here, ${\rm{SOC}}$ is the battery nominated state of charge within a range of 0 and 100% and it is initially set to 100%. A BEV is assumed feasible if all of the observed driving days are within its driving range. For a PHEV, we assume all electric driving until zero SOC is reached and conventional driving after. That is, we model PHEV in charge depleting mode until reaching zero remaining electric range. The individual electric driving share for PHEV is then given as all electric km divided by total km driven.

2) Technical Vehicle Parameters

We simulate electric taxi driving for EV models available on the market (both BEV and PHEV). In our analysis, we use vehicle all electric range (AER) and energy consumption (EC) according to the U.S. Environmental Protection Agency (EPA) test cycle (model year 2018/2019). EPA test cycle ratings comprise a series of driving cycles, among them a highway and a city cycle as well as tests under hot and cold temperature test conditions. Thus, they represent a broad application range. Furthermore, they are found to be in the magnitude of the empirical energy consumption of electric taxis measured in field tests in Germany (e.g., the EPA energy rating for the Tesla Model S 85 deviates less than 5% on average compared to the field test results in [33]). In addition, the EPA range was found to be a good average for empirical driving ranges of PHEV [34]. The vehicle models and their parameters used for simulation are summarized in Table II. EPA measures energy consumption including charging losses. While for the BEV driving simulation we do not consider these losses, they are included in the economic analysis.

Conventional diesel taxis have a high fuel consumption, also due to long idle periods with heating needs. In our analysis, we use a diesel fuel consumption of 10 l/100 km which is an empirical value stated in [32]. We tested the sensitivity of our results with an average consumption of 8 and 14 l diesel per 100 km as discussed later. For conventional driving of PHEV we assume a proportional consumption to diesel of 11.6 l/100km gasoline. Although PHEV fuel consumption could be lower due to regenerative braking, this assumption hardly affects the quality of our results.

3) Charging Simulation

For recharging the electric taxis during idling periods, we distinguish two cases:

1. Charging at a taxi stand with high power charging: If a vehicle is standing at a taxi stand with charging infrastructure available, we assume charging if the vehicle is idle for more than five minutes. Charging infrastructure is available at selected taxi stands.

2. Additional home charging with low power. If a vehicle is parking for more than five hours, we assume home charging wherever the vehicle is parked except for a taxi stand. Charging power is 3.7 kW. Home charging is assumed to be available in addition to fast charging. The reasoning behind this approach is that if a taxi driver would be willing to adopt an EV, she should be inclined to park the vehicle where charging is possible. For the scenarios without home charging, only fast charging at taxi stands is available.

Charging power P [kW] might be restricted by the maximum charging power of the vehicle (see Table II) or the charging infrastructure $(P = \min({P_{\max,vehicle}},\ {P_{\max,CI}}))$. In a previous study [35], we found that increasing charging power above 50 kW has a very limited effect on the electrification potential. Accordingly, we limit our analysis on a maximum of 50 kW charging power for all chargers.

We assume, that the total idling time $\Delta t_{I}$, can be used for charging, less a handling time of one minute $\Delta t_{H}$. For fast charging at a taxi stand, we assume a maximum SOC of 80%, for home charging S$OC_{max}$ is 100%. Due to comfort aspects, we assume that taxi drivers are only willing to charge at a taxi stand if the SOC is below 75%. The battery state of charge after charging SOC(t + 1) is calculated as:

$$\begin{align*} {SOC\left({t + 1} \right)} &= {\rm{min}}\left(SO{C_{max}},\ SOC\left(t \right) \right.\\ &\left.\quad\; + \left({\Delta tI\ - \Delta tH} \right)\ \cdot \ P\ \cdot \ {K^{ - 1}} \right) \tag{2} \end{align*}$$

4) Charging Infrastructure Availability

We rank the charging sites by the taxi arrival rate, i.e., the taxi stand with the most frequent taxi visits is chosen as starting point for the charging infrastructure rollout. In the second stage, we assume an additional charging site at the second most frequented taxi stand and so on until all 20 taxi stands are equipped with charging infrastructure. Since parking and idling times of taxis are mainly at taxi stands - all 20 taxi stands are among the 25 most frequent stopping points [35] - the focus on these locations is justifiable.

B. Economic Analysis

1) Vehicle TCO

A total cost of ownership analysis (TCO) is a common approach to evaluate alternative drivetrains and especially the commercial sector uses TCO to assess vehicle investment decisions [4]. Annual TCO ($TCO_{a}$) are the sum of annualized capital expenditures ($a_{capex}$) and operating expenditures ($a_{opex}$). Capital expenditures of an investment I with calculatory lifetime T and interest rate i are given as:

$$\begin{equation*} {a_{capex\ }} = \ I\ \frac{{\left({1 + i} \right)T\ \cdot \ i}}{{\left({1 + i} \right)T\ - \ 1}}.\tag{3} \end{equation*}$$

Operating cost comprise cost for conventional and electric driving, cost for operation and maintenance ($\text{c}_{O \& M}$), yearly taxes ($c_{tax}$) and cost for emission tests (only ICEV and PHEV). Other cost, such as tire cost, are neglected, since they might be the same for all drivetrains. Energy cost for PHEV consist of fuel cost for conventional driving and electricity cost for electric driving. Cost for electric driving are given as the product of the electric driving share UF, the electric consumption $c_{el}$ [kWh/km] and the cost of electricity $p_{el}$ [€/kWh]. Accordingly, parameters for the conventional driving are the share of conventional driving (1-UF), fuel consumption $c_{fuel}$ [l/km] and fuel cost $p_{fuel}[ {{\text{C}}\!\!\!\!\!\!\!=} / \text{l}]$. For BEV, UF is 100% and for the diesel car UF is 0%. Given the annual vehicle kilometers driven (aVKT), yearly operating cost of a taxi are:

$$\begin{align*} {a_{opex\ }} &= aVKT\ \cdot \left({UF\ \cdot \ {c_{el}} \cdot \ {p_{el}} + \ \left({1\ - \ UF} \right) } \right.\\ & \quad \left. \cdot {\ {c_{fuel}} \cdot \ \ {p_{fuel}} + \ {p_{O \& M}}} \right)\\ & \quad\; + {p_{tax}} + {p_{ET}} \tag{4} \end{align*}$$

Here, we use the TCO analysis to assess the cost barrier to EV adoption and thus focus on the difference annual TCO of electric vehicles compared to conventional vehicles. We do not include taxi operation profits, since we assume electric operation only if all trips can be made by an electric taxi. Accordingly, drive train choice does not affect income. We discuss this assumption in the discussion section.

2) Economic Vehicle Parameters

In Germany, Mercedes-Benz (MB) dominates the taxi market with the upper middle class sedan MB E-Class. Due to its importance, MB offers a special “The Taxi”-edition that contains all necessary equipment such as the roof sign, taximeter and a hands-free kit. Thus for the electric models, there will be additional cost of 1,590 € (excl. VAT) for the installation of a taxi package. In addition, the MB E-Class “The Taxi” has a higher than basic equipment. Accordingly, the purchase price of the EV models does not necessarily correspond to their basic price, but additional equipment was added if appropriate (e.g., higher equipment variants, parking heater as additional equipment, etc.). The purchasing prices of the EV models are summarized in Table III.

TABLE III Purchase Price of the EV and ICEV Models Analyzed

In order to take into account the different vehicle segments of the electric vehicle models, their economic valuation is carried out in comparison to combustion vehicles from the same segment. These conventional reference vehicles therefore differ depending on the electric vehicle model (cf. Table III).

We do not take into account political incentives since we aim to assess the current technology independently from policy measures. In addition, vehicle manufacturers also offer exchange premiums for new diesel cars. For example, Daimler grants an exchange premium of 2000 to 6000 €, while the purchase premium for a battery electric vehicle is 4000 €.

Operating cost and related parameters are shown in Table IV. All values are without value added tax (VAT) since companies are eligible for a 100% VAT refund. No empirical data from EV driving of taxis in Karlsruhe is available. Thus, we rely on external cost parameters. For example, the diesel price is the average price for 2018 stated in the annual report of the German taxi and car rental association [32].

TABLE IV Operating Cost Vehicle (Base Year 2018, All Values w/o VAT)

Vehicle taxes are according to current law. Cost for operation and maintenance (O&M) are from literature on passenger cars. These are higher than material and service cost stated in [32] or [33] (0,036-0,041 €/km) due to additional personnel cost. Since no reliable data on resale values of electric taxis is available, for all vehicle types, we calculate the capital cost over a calculatory lifetime of 10 years without residual value at vehicle end of life. In our analysis, we assume an EV to charge for 0.21 €/kWh at all types of charging infrastructure. This price corresponds to the average electricity price for commercial customers and is well in line with charging tariffs at e.g., Tesla´s superchargers [35].

3) Modelling Charging Infrastructure Profitability

To analyze charging infrastructure profitability, we determine the monthly fee per electric taxi that would be necessary to make charging infrastructure operation profitable - if the generated income from selling electricity (to the taxis) is not enough to do so.

We first heuristically determine the number of charging points needed per taxi stand$\ c$ ($C{P_c}$) to cover eTaxi charging demand at any time. To this end, we determine the maximum energy charged at a taxi stand $c$ for all 15-minute periods ${t_{15}}$ over the whole observation period $T$. This number is then divided by the maximum charging power $P$ per charging point (with factor 4 to adjust for the 15-minute period):

$$\begin{equation*} \# C{P_c}\ = \left\lceil {\ \frac{{\mathop {{\rm{max}}}\limits_{t15\ \in T} \mathop \sum \nolimits_{v \in V,\ CI = c} {{\rm{kWh}}_{charged}}\left({v,c} \right)}}{{4*P}}} \right\rceil\tag{5} \end{equation*}$$

Since the charging infrastructure network must be sufficient only for the electric taxis and not for the entire taxi fleet, we focus the profitability analysis on the subset ($v \in V$) of all technically or techno-economic feasible taxis, respectively.

In a second step, we distribute the annual costs, which are not covered by the sale of electricity, evenly over all electric taxis in the form of a monthly basic fee. The monthly basic fee per electric taxi $({M{F_v}})$ is given as:

$$\begin{align*} & M{F_v} = \frac{{1\ \left[ {year} \right]}}{{12\ \left[ {months} \right]}}*\frac{1}{{\mathop \sum \nolimits_{v \in V} 1}}*\\ &\quad\; \times \left[{\vphantom{\mathop \sum \limits_{v \in V,\ CI = c}}}\! \mathop \sum \limits_c \# C{P_c}*\ \left(\!\! {{I_{CI}}\ \frac{{\left({1 + {i_{CI}}} \right){T_{CI}}\ \cdot \ {i_{CI}}}} {{\left({1 + {i_{CI}}} \right)\ {T_{CI}}\ - \ 1}} + Ope{x_{CI}}} \!\!\right) \right.\\ &\quad\; \left. - \left({{p_{el}} - {p_{ind}}} \right)*\frac{{365}}{T}\mathop \sum \limits_{v \in V,\ CI = c} {{\text{kWh}}_{charged}}\left({v,c} \right) \right] \tag{6} \end{align*}$$

We adjust the amount of energy charged by the factor $\frac{{365}}{T}$ to represent a yearly operation. For the parameters, please refer to Table V. We assume a charging station with two charging points with 50 kW each for BEV and 11 kW each for PHEV. The charging infrastructure operator can buy electricity for the industrial electricity price due to the high order volume [36]. Investments for charging infrastructure also comprise investments for the grid connection.

TABLE V Economic Parameters of CI (w/o VAT). Data Source: [36]

SECTION V. Results

A. Taxi Usage Patterns

Daily vehicle kilometers travelled (VKT, Fig. 1) of taxis are well above those of private vehicles - but mainly below 250 km. This might implicate a high electrification potential due to the constantly high daily VKT which are necessary to economize against conventional vehicles but which are well within the battery range of today´s BEV (c.f. Table II).

Fig. 1.

Distribution of daily vehicle kilometers travelled (VKT). Shown are all days of all vehicles.

In public, the potential of EV is often discussed based on average daily driving distances. Yet, a conventional vehicle will be replaced by an EV only when it can satisfy almost all driving needs (not assuming neither changes in driving behavior nor a car pooling potential - or in this case, a reallocation of suitable trips to electric taxis). To illustrate the difference between electrification needs for average driving and all driving, for every taxi, its average daily VKT is plotted against its maximum daily VKT in Fig. 2.

Fig. 2.

Distribution of mean vs. maximum daily driving distances for the taxicab data sample. Blue line indicates a linear fit (R² = 0.68) with 95% confidence band. Whiskers of boxplots show 1.5 interquartile range.

For the majority of the taxi drivers, maximum daily VKT are more than twice as high as mean daily driving distances. This means, that fast charging infrastructure is necessary for electric taxi driving. Due to the high charging demand and the local operation of taxis in Karlsruhe - 95% of all trips ended within a radius of 8 km around the city center [35] -, a charging network specifically designed for eTaxis has a high potential to be operated profitably.

Taxi stands are an obvious and a promising location for taxi charging infrastructure, since taxis often occupy taxi stands. Fig. 3 shows the average number of taxis idling at a taxi stand over the period of one week. However, while some of the taxi stands are rarely used (one taxi arriving every third hour), several taxi stands are used heavily with being occupied by multiple taxis over a period of six hours. These taxi stands are predestined as starting points for the built-up, as discussed later. Accordingly, the blue line in Fig. 3 shows the occupation pattern for the most frequented taxi stand, while the red line in contrasts shows the average over all 20 taxi stands.

Fig. 3.

Number of taxis at a taxi stand over time for the most frequented taxis stand (blue line) and averaged over all 20 taxi stands (red line). The figure shows an average week.

Further, the driven round-trip distance between two stops at a taxi stand is well within long range BEV for most of the taxis. Fig. 4 shows the cumulated density function of roundtrip distances starting and ending at a taxi stand (may be different taxi stands). One line represents the driving distance distribution of one taxi. Light blue lines indicate round-trip distances if a minimum idling time of 30 min between two consecutive round-trips is assumed, e.g., as necessary for charging. Dark blue lines show round-trip distances, independent of the idling time. Altogether, this underlines the high technical potential of long range BEV in combination with a dense charging infrastructure network at the taxi stands.

Fig. 4.

Cumulative density function of distances between taxi stands, one line per vehicle. Dark blue lines show distances of a path chain with no idling at taxi stands. Light blue lines show path chains with at least 30 min idle time between them.

Finally, taxis in Karlsruhe are only in operation for a limited time. Idling times sum up to 20 h per vehicle and day (median) and thus, taxis have a sufficient amount of parking time that is necessary for the electric taxis to be recharged. The high parking time might weaken the challenges of many taxis being at a taxi stand at the same time - as indicated by the daily recurring patterns (Fig. 3).

B. Technical Electrification Potential

The electrification potential of the taxi fleet depends on charging infrastructure availability, which is why we consider different stages of charging infrastructure rollout. The rollout of the taxi stands is prioritized by the number of taxi visits per day, as described in more detail in [37].

The electrification potential of the Karlsruhe taxi fleet with battery electric vehicles (BEV) increases with charging infrastructure availability at taxi stands, with a decreasing benefit for more than half of the taxi stands. With long range BEV, such as the Tesla Model S or the Hyundai Kona, the electrifiable share of kilometers increases from 30–50% for five charging sites to a share of ∼40–60% with ten charging sites. More charging infrastructure has only a limited effect (∼70% for the Tesla with 20 sites and ∼45% for the Kona, see also Fig. 5). Especially the combination of very high vehicle ranges and a medium to high availability of charging infrastructure is necessary for high electrification shares above 40% (Fig. 5). BEV with ranges below 300 km might only be interesting if operated very locally. This in turn would make reallocation of trips necessary (and taxi operation economically less interesting, see Discussion).

Fig. 5.

Electric driving share for different charging infrastructure rollout stages as function of EV range. Only charging at taxi stands is assumed. Left: BEV models, right: PHEV models.

Especially in the beginning, the additional availability of home charging infrastructure - or other types of infrastructure where taxis can charge during a larger parking period - could be important to enable electric driving. If home charging would be available, already five charging sites would allow for an electric driving share of 50–75% for the aforementioned vehicles - given a maximum potential of 60–80% with 20 charging sites.

Since PHEV potentially electrify parts of all trips, they might reach high electrification shares with lower all electric ranges (AER) (Fig. 5). With an AER of 85 km, the Chevrolet Volt could electrify almost the same driving share as the Tesla. However, all vehicles would need to be such a PHEV. In addition, the Chevrolet Volt is not available in Germany and the end of production was already announced. Additional home charging allows for 5 to 20 percentage points higher electric shares, depending on vehicle model and charging infrastructure availability. For PHEV, also the maximum charging power has a remarkable effect (cf. Fig. 5 for the Mitsubishi Outlander PHEV, range: 35 km). Despite its lower range, it can electrify more km than PHEV with higher AER due to its fast charging possibility (Table II). With home charging availability, the influence of fast charging power decreases.

C. TCO Analysis

The additional cost of driving a Tesla, compared to a MB E class 300d (diesel), are quite constant for the different charging infrastructure rollout stages, regarding both median cost and variance (Fig. 6). However, the underlying basic population varies from 25 BEV (16% of all taxis) for one taxi stand with charging infrastructure (and no home charging) and 111 BEV (69%) for ten charging sites up to 117 BEV (73%) for 20 charging sites.

Fig. 6.

Yearly total cost of driving a Tesla Model S (left) and a Jaguar iPace (right), compared to the conventional MB E class 300d as function of charging infrastructure availability. Only charging at taxi stands is assumed. Positive values denote higher cost for BEV driving. Boxplot shows the median as a solid line. Whisker show 1.5 times the interquartile distance.

The TCO of a Tesla was compared with that of a MB E class due to both cars being reference models for a BEV and a German taxi, respectively. That is, the comparison is the best available. However, the higher costs of the Tesla could be attributable to its higher vehicle class. Accordingly, we compared the same MB E class to the Jaguar iPace, a vehicle that is already used as a Taxi in Germany (Munich). As shown in Fig. 6, most of the taxis could be operated economically with a Jaguar iPace. However, due to the lower range, the share of technically feasible BEVs is lower than for the Tesla: 10 BEV (6% of all taxis) for one charging site up to 67 BEV (42%) for 20 charging sites. Since especially the lower range of the Jaguar might lead to lower cost compared to the Tesla, we conclude, that high electrification rates come at (currently) high cost. The results for the Hyundai Kona are comparable to the Jaguar iPace, but at lower cost (all vehicles with negative delta TCO, median of delta TCO: -4500 EUR/a).

Most PHEV come at lower cost than their diesel counterparts. The Hyundai Ioniq PHEV, the Chevrolet Volt and the BMW 530e come always at lower cost, even if charging is only possible at taxi stands. The majority of Prius and Outlander also come at lower cost, if only one taxi stand is equipped with charging infrastructure (Table VI). With additional home charging, all PHEV come at lower cost.

TABLE VI Delta TCO of PHEV With Charging Infrastructure Exclusively at Only One Taxi Stand

D. Cost Minimal Drivetrain

For every taxi, we determine the EV model with the minimal delta TCO compared to their diesel reference. Since the vehicles have different vehicle classes, the interpretation of the results might be difficult and we keep this section short. All taxis could be operated electrically, with either a PHEV or a BEV. The BEV with the lowest delta TCO is the Hyundai Kona, for PHEV it is the Chevrolet Volt (or Hyundai Ioniq, if the Volt is not available). The share of taxis with a PHEV being the cost minimal drivetrain is 90% if only one taxi stand is equipped with charging infrastructure. This PHEV share falls below 50% for the case that all taxi stands are equipped with charging infrastructure. Additional home charging availability strongly increases technical potential of BEV taxis (with BEV reaching 30-50% EV share).

E. Techno-Economic Electrification Potential

In the previous sections, we analyze the electrification potential separately from the economic potential. While all technically feasible Hyundai Kona are also techno-economic feasible, the techno-economic potential of the Jaguar iPace is only slightly below its technical electrification potential. In contrast, the techno-economic potential of the Tesla Model S is only half of its technical potential and thus falls below that of the Kona. However, the different vehicle segments of the different models have to be taken into account. Especially since taxi drivers are scarce in Germany, taxi drivers can make demands and a Hyundai Kona might not meet their comfort requirements. As the Tesla has a high technical potential and could be accepted by taxi drivers, further cost reductions are important for taxi electrification potential, as discussed in Section VI.

F. Charging Infrastructure Profitability

We now turn from the vehicle to the charging infrastructure perspective and determine the monthly fee that every electric taxi would have to pay for the use of the charging infrastructure. The reasoning behind this approach is the fact that the possibility to charge at a dedicated taxi charging infrastructure with no public access has its own value equal to all taxis, beyond their actual usage. Accordingly, we have refrained from adjusting the electricity price for fast charging in favor of the monthly basic fee. The monthly basic fee that would be necessary to refinance charging infrastructure at one taxis stand, amounts to at least 8 €/month per eTaxi (with home charging also being available). This value rises up to 130 €/month for refinancing 20 charging sites (c.f. Fig. 7). Please note that we spread infrastructure cost among all technically feasible BEV. Their number increases with the number of charging sites and thus, the infrastructure cost are spread among more vehicles. This scenario is the most profitable from the infrastructure perspective.

Fig. 7.

Monthly fee per eTaxi necessary for refinancing the charging infrastructure. All technical feasible taxis. Bar totals show monthly fee per vehicle for total charging infrastructure. Coloring differentiates cost per charging site. Left: home charging is available, right: only charging at taxi stands (no home charging).

Without home charging, although in total more energy is charged at the taxi fast charging stations, cost are divided by less vehicles (due to the lower technical potential). Thus, the monthly fees are higher and range from 20 to 170 €/mo (for all technically feasible vehicles).

If focusing only on techno-economic feasible vehicles, monthly fees are even higher. For the Hyundai Kona, the monthly fee would be 15 to 160 €/mo (with home charging), for the Jaguar iPace it would amount to 27–190 €/mo and for the Tesla Model S it would be the highest with 130–550 €/mo. Thus, under current economic conditions, if only Tesla would be considered as electric taxis, only one charging site would be profitable, leading to 3% electrified kilometers. For all eTaxis being a Jaguar iPace, six charging sites would allow for one fourth of all km being electrified. Due to its lower price, the full techno-economic potential of the Hyundai Kona could be exploited with 20 charging sites. However, its lower segment has to be taken into account, which might distort comparability to a certain extent. Accordingly, we focus on the Tesla Model S in the following since it probably is most comparable to current conventional taxis in Germany.

The advantage of PHEV is that almost all taxis are techno-economic feasible with either the Chevrolet Volt, the Hyundai Ioniq or the BMW 530e and thus, cost are divided by more vehicles. In addition, charging infrastructure for PHEV is less expensive due to the lower charging power. Accordingly, monthly fees are much lower for PHEV, ranging from 2–70 €/mo for the Volt and from 9–95 €/mo both for the Ioniq and the 530e if all charging would take place at these stations. With additional home charging, these values are almost the same with a deviation of not much more than 3 €/mo.

G. Future Prospects

Currently, charging infrastructure could hardly be profitable if only used by Tesla Model S taxis since the lower TCO of the electric vehicles (Fig. 6) could not compensate for the necessary fees for exclusive charging infrastructure usage (Fig. 7). However, the purchasing price of electric batteries will decrease in the future, mainly due to continuously falling battery prices [39]. In addition, due to more stringent CO₂-emission standards, conventional cars (ICEV) will probably become more expensive. Altogether, EV become financially more attractive in the near future so that charging infrastructure could become financially self-sufficient even though only used by taxis.

With battery prices falling, OEM might 1) either reduce the purchasing price of the vehicle or 2) increase battery capacity at constant purchasing price. The effects of increased battery capacities at constant purchasing price (case 2) can be interpreted from the results above. The Jaguar iPace, for example, is usually cheaper as its conventional counterpart (Fig. 9). Increasing battery capacity inversely proportional to battery price development would imply an increase in vehicle range by roughly 30% until 2030. This is comparable to the Tesla´s range, which we assume to be a medium term upper limit.

For the analyses of reduced purchasing prices (case 1), we focus on the Tesla Model S based on the parameters shown in Table VII. The decrease of the purchase price was determined on battery pack price development as stated in [39] (using a factor of 1.5 as a markup for the battery system and profit margin). The cost of charging infrastructure are kept constant due to low cost reduction potential. The fall in battery prices not only reduces the average TCO of a BEV (Fig. 8), but it also increases the number of taxis that could be operated technically and economically as BEV. The share of BEV (Tesla Model S) with techno-economic potential increases from ∼20–25% in 2020 to ∼45–60% in 2030, which means that in 2030 almost 85% of all technically feasible BEV are also cheaper than their conventional counterparts (under the assumptions made). In 2025, roughly 75% of all technically feasible BEV are also economically feasible

TABLE VII Economic Parameters for Future Development

Fig. 8.

Average Delta TCO of all techno-economic feasible Tesla Model S taxis. CP = number of taxi stands with charging infrastructure.

Fig. 9.

Impact of long distance round trips: share of roundtrips with a destination further than 50 km from the city-center and their share on the turnover per year for each taxi.

Altogether, an average Tesla Taxi in 2025 could afford to pay 180 € per months for charging infrastructure to be cost-neutral to an ICEV (see Fig. 8). As early as 2025, a charging network of ten charging sites could be financially self-sustaining, even if only half of all techno-economically feasible taxis were actually operated as BEV (and if no additional charging option was available). This would imply 30% of the taxi kilometers being electrified.

The maximum possible electrification potential that can be reached with Tesla taxis, given that both vehicle and charging infrastructure operation must be profitable, amounts to >50% electric km in 2025, as shown in Table VIII. In 2030, all taxi stands could be profitably equipped with charging infrastructure, leading to more than 60% electric km. Since in 2030 already ∼85% of all technical feasible taxis come also at lower cost, for higher electrification mainly technical solutions are necessary, such as additional charging opportunities for longer trips (see Discussion).

TABLE VIII Number of Taxi Charging Sites That Could be Refinanced Only by Taxi Use - And the Corresponding Electrified KM Share. Charging Only at Taxi Stands

H. Discussion

The design of our model has limitations because the results depend directly on the data and assumptions, especially on driving and charging behavior. To account for the corresponding uncertainties, the most important assumptions are discussed in the following.

We use driving data of 161 operating in Karlsruhe, Germany. The profiles show a comparatively low operation time as well as a local operation around the city center, but with recurrent long distance trips. Accordingly, our results are only generalizable to some extent, especially with regard to megacities with much more taxis in operation where driving patterns might be very different.

We assume same operation of electric and conventional taxis. First, this implies that taxis are charged only during idling periods. Especially additional charging possibilities might increase electrification potential of taxis, such as our analyses on home charging availability show. In addition, we do not assume rearrangement on trips. However, only one long distance trip might render technical operation impossible. But if charging infrastructure would be available at the destination of a long distance trip, e.g., a public charging point at a train station, the share of technical feasible electrified kilometers would increase by ten percentage points compared to the results shown in Fig. 5 (for a Tesla Model S with 20 charging sites). In addition, since long distance trips are recurrent (75% of the taxis have at least one trip per month), these are an important income for taxis. Monthly turnover of long distance trips (>50 km) can sum up to 1300 € (see Fig. 10) and thus, not being able to make these trips would be an economic barrier for taxi operation.

In our study, we analyze the BEV potential of the taxi fleet using EPA test cycle energy consumption for the whole year and vehicle life. This might overestimate electrification potential, since the range of the BEV might strongly decrease, e.g., at battery end of life or in cold temperatures. Accordingly, Goldschmidt et al. [41] identify a so-called winter-effect, which leads to lower assessment of electric driving performance among current electric taxi drivers. To assess the effect of lower AER on our results, we perform a sensitivity analysis, assuming only 80% of the BEV range (cf. Table II) to be available. We find a strong decrease in technical potential for all BEV models. The technical electrification potential of the Tesla decreases by ∼20 percentage points for all scenarios with at least five charging sites (cf. Fig. 5). The corresponding decrease for the Hyundai Kona are 12–15 and for the Jaguar iPace 5–10 percentage points. However, there is still a remarkable share of 20% (for the Jaguar iPace) to 40% (Tesla) of electrifiable kilometers, given that charging infrastructure is available at ten or more taxi stands. Accordingly, although the reduced all electric range has a high negative effect, a considerable share of taxis could still be operated fully electrically.

Since we had to rely on external data for our economic analysis, we have tested the sensitivity of our results to our main assumptions. For a consumption of 8 l/100 km (instead of 10) for the Diesel reference car, yearly TCO of a diesel car would decrease by ∼1300 € on average. For a Tesla, the techno-economic electrification potential would decrease from ∼60% to 40% (with 20 charging sites). Accordingly, future cost decrease of BEV taxi driving would become more important.

We only use a heuristic to determine the number of charging points per charging site needed, which could overestimate charging infrastructure needs. Nevertheless, even these high basic fees allow for an economic operation of BEV today, e.g., with the Hyundai Kona or the Jaguar iPace. In the near term future, also the Tesla could allow for a profitable operation of a dense taxi charging infrastructure.

In addition to collecting data from the real world of electric taxis, future research could also focus on seasonal differences in taxi use and energy demand, especially taking into account the additional power required for heating and cooling the cabin during idling phases. Furthermore, a reallocation of journeys will be analyzed as the next step to increase the BEV potential in a mixed fleet of PHEVs and BEVs to understand possible effects on the charging infrastructure demand (c.f. for example [42]).

SECTION VI. Conclusion

This paper analyzes the electrification potential of the taxi fleet in Karlsruhe, Germany. We focus on both, the vehicle and charging infrastructure perspective to better address interaction effects. Our analysis underlines the high electrification potential of taxi fleets - for which long range BEV are necessary. We find that charging infrastructure at half of all taxi stands is sufficient, but also necessary to reach high electrification rates. However, the cost of a charging infrastructure rollout only for taxis is very high. Accordingly, a profitable operation of charging infrastructure used exclusively by taxis is only possible to a very limited extent under current circumstances. For example, with a Tesla Model S taxi fleet, only one charging point would be economically viable in Karlsruhe today, allowing to electrify 3% of km. However, already in 2025, 15 charging sites could be refinanced due to lower vehicle cost, which would allow for over 50% of km electrified. Policies thus might focus on incentivizing especially charging infrastructure. In addition, policies might support home charging or adequate alternatives for taxis, since also PHEV have a high electrification potential, in particular with a home charging possibility.

ACKNOWLEDGMENT

This publication was written in the framework of the Profilregion Mobilitätssysteme Karlsruhe, which is funded by the Ministry of Economic Affairs, Labour and Housing in Baden-Württemberg and as a national High Performance Center by the Fraunhofer-Gesellschaft.