2 THE DEMAND FOR ELECTRIC VEHICLES
2.4 Data collection methods
The major obstacle to modelling charging behaviour is the insufficiency of real-world data because of the limited amount of electric vehicles in the transport network. Revealed charging data could be used to calibrate policy-sensitive discrete choice models of charging behaviour. Detecting the existing few drivers and convincing them to participate in surveys or even to be monitored for some months is a rather inefficient and challenging approach.
Consequently, collective actions and partnerships between various stakeholders (car
manufacturers, power utilities, universities etc.) are required in order to design EV trials and find volunteers to take part in them.
Electric vehicle trials around the world are essential in order to generate datasets not only on charging behaviour but on battery and component performance as well. Sensing and communication technology developments like GPS devices create new opportunities to collect sophisticated spatiotemporal travel information. Participant privacy and data ownership issues, in general, slow down the research progress of the electro-mobility area but significant steps have been made during the last few years. Representative data that can be collected with these trials are:
Power flows across the battery terminals
SOC
EV position from GPS
Ignition position
Brake pedal state
Ambient temperature
Ancillary load state
A notable example is the EV Project in the U.S. with 18 cities participating, and a target to enrol 8,300 vehicles (LEAFS and Volts) and install 14,000 Level 2 and DC fast-charging units (combined residential and commercial) (Smart and Schey, 2012). Participants had a charging unit installed at their home for free, in exchange for data collection both from their vehicles and the charging unit. Also, the project was crucial for the deployment of public infrastructure in strategic locations, following an instalment plan that would be beneficial for the users, the owner and the community as a whole.
Another demonstration project, dedicated only to PHEVs, was launched from the Idaho National Laboratory (INL) through the U.S. department of Energy’s Advanced Vehicle Testing Activity (AVTA) (Smart et al. 2010). The trial took place in 26 U.S. states, 3 Canadian provinces and Finland and data was collected from onboard loggers in 290 Ford Escape Hybrids and Toyota Prius that were converted to PHEVs. The sample was divided in two (one for commercial use and one for personal use) and all charging events were carried out with a Level 1 rate.
The Ilmenau University of technology in Berlin monitored the charging events of 50 BMW MINI-E’s for one year and observed that some of the drivers were not charging their vehicles
every day (Westerman et al., 2010). This behaviour is either in line with the “not-every-day”
charging scenario described in subsection 2.3.7, or it can be attributed to the fact that those drivers were not using their vehicle every day. Similar MINI-E trials were carried out by the BMW Group in other countries around the world (450 vehicles in New York and Los Angeles and 40 vehicles in the United Kingdom). The UK trial reached similar conclusions for the frequency of charging and provided evidence that time-of-use pricing can effectively shift charging load peaks (BMWGroup, 2011).
Cenex, the UK’s first Centre of Excellence for low carbon and fuel technologies, has cooperated with organisations in the North East of England by providing them with electric smart Fortwo passenger cars to evaluate the integration of these vehicles into fleets and the opportunities of targeting at this direction for early adoption (Carroll, 2010). In order to achieve the successful deployment and management of those fleets, Cenex has established partnerships with several councils in this area. The aim of the trial was to collect both qualitative data (from drivers and fleet managers) and quantitative data from logging devices in the EVs. Trip-related information was collected from the vehicle’s CAN (Control Area Network) bus and GPS instruments and it was transmitted to a central server every time the driver was turning off the ignition. The trial lasted for 6 months and the results have shown that there was a positive feedback from fleet managers after their experience with the vehicles and the rating for this experience was even higher for organisations with access to dedicated charging infrastructure.
Switch EV, another EV trial that took part in the North East of England between November 2010 and May 2013, covered equipment and installation costs for various charging point technologies (3,7, 22 and 50KW outlets) in public and workplace facilities with the hosts providing free electricity to the drivers as an exchange. Several EV commercial models were used including the Nissan Leaf and the Peugeot iOn. One uncommon characteristic of this trial is the collection and availability of information regarding payment transactions, by monitoring both the vehicle loggers and the charging devices (Hubner et al., 2013). It is also one of the first trials to introduce a membership scheme and the real-time information map with charging post availability.
In Birmingham and Coventry, 108 EVs were monitored for the purposes of the CABLED trial. Half of the participants had smart meters installed at home and they were reimbursed for charging their vehicles off-peak (Robinson et al., 2013). Out-of-home charging stations
were also installed, with prices varying from free up to some amount that was levied for parking.
Congested urban areas with great diversity in land use and extended public transport networks, like Central London, are of high interest for testing EV use. The Technology Strategy Board (TSB) has already conducted similar trials in London. Two examples are: a) The Mercedes Smart trial with 60 EVs deployed to residential customers and smart meters installed by EDF to evaluate charging behaviour and users’ response to tariff incentives and b) The Toyota Plug-in Hybrid trial with 20 second-generation PHEVs, targeting major business customers like TfL and Sainsbury’s (Marantes, 2009).
Last but not least, an EV trial was undertaken as part of Low Carbon London (LCL), a project of UK Power Networks (UKPN). The aim of the LCL was to analyse the impact of various low carbon technologies on London’s distribution network. In the trial, there was a mix of residential and commercial participants. Among the 72 residential participants, the 47 were already EV drivers, while the remaining 25 were leased a Nissan Leaf for the purposes of the trial. Apart from the onboard loggers of the EVs, data was collected from Source London infrastructure that was used for recharging. Moreover, socio-economic information and driving patterns were collected from questionnaires that have been designed from the Centre of Transport Studies at Imperial College. Some notable conclusions from the subsequent analysis were that peak demand for residential users occurred around 09:00 pm in the evening and that public charging infrastructure was used mainly as an “insurance” policy and not as part of the drivers’ routine (UKPN, 2014).
The characteristics and the key findings of the EV trials described above are summarised in Table 2.3.
A qualitative comparison of the observed charging patterns from EV trials and the charging scenarios presented in 2.3.7 suggests that the existing modelling techniques can capture the aggregate charging load both in the presence and in the absence of differentiated electricity tariffs. Nevertheless, without a disaggregate analysis of charging behaviour it is difficult to link the various charging scenarios with the idiosyncratic characteristics of each area, and hence, it is possible to overestimate or underestimate the magnitude of charging demand.
The need to capture heterogeneity in charging choices is supported by the work of Franke and Krems (2103c) who showed that the starting point of a charging event is affected by the user-battery interaction context.
Table 2.3: Characteristics and key findings of representative EV trials
EV Trial Location Vehicles Charging
infrastructure References Key findings
England Smart Fortwo 13A/ 240V Carrol, 2010
Positive feedback from fleet
and 50kW outlets Hubner et al., 2013
PHEV trial London Toyota Prius
PHEV Same as above Marantes,
Apart from directly monitoring the driving and charging behaviour of EV owners, there is a great value in talking with them by conducting face-to-face interviews. Kurani et al. (2009) explain how they employed this narrative analysis and how hearing personal stories from
the drivers contributed to synthesising the general framework and explaining the meanings behind their actions. Although these interviews can be guided based on previously collected personal information (e.g. travel patterns or vehicle design preferences), they have an open-ended character so that a wide range of perspectives is represented and emerging themes are highlighted.
Kurani et al. (2007) examined the early stage experience of PHEV use by interviewing 23 drivers, including also questions regarding their perceptions of drawbacks and benefits as well as their suggestions for future designs. Perceptions and attitudes towards the essential characteristics of the new technology (acceleration, driving style, environmental feel and noise) were analysed for the Smart Move trial (Carroll, 2010) after the completion of a questionnaire by the fleet users. Tal et al. (2014) implemented an innovative approach to collect self-reported travel and charging data, with the use of a web map (Figure 2.4). The major advantage of this method is its cost-efficiency by overriding costly travel diary administration and installation of monitoring equipment. Additionally, the nature of the survey tool allows them to infer the subjective needs of the drivers and their willingness to pay.
Stated preferences (SP) exercises are also very important in order to explicitly model the behaviour of respondents when they are faced with hypothetical situations. In the area of electro-mobility, most of the SP studies are oriented around customer’s intentions to purchase electric vehicles rather than their preferences around the everyday use of the vehicle (Hidrue et al., 2011). One common problem with designing such a choice experiment is the lack of knowledge regarding the underlying technology and the complexity in processing several unfamiliar concepts (e.g. range anxiety, combination of driving modes for PHEVs, etc.) in a short period. Drivers have no constructed preferences for these attributes and, therefore, it is difficult for them to foresee how they would use the new technology and why they should buy it.
Moreover, their choices are affected by their current experience with conventional vehicles, and this asymmetry in experience might lead to biased estimation of parameters that have significant differences (e.g. driving range). For this reason, the use of traditional SP methods in electric vehicle preferences has been criticised (Turrentine et al., 1992; Kurani et al., 1996).
Turrentine et al. (1992) tried to assess the adaptability of households to limited vehicle range trough purchase intention and range simulation games (PIREG). In these games, respondents provided their activity-travel diary and then they were asked to adjust it in order to cope with the hypothetical limitations from the use of an electric vehicle. Gaming and Simulation (GS) approaches like this one have the disadvantage of increased complexity and, as a result, it is more difficult to achieve large samples for model estimation.
Figure 2.4: Web map for self-reported driving and charging habits. Reproduced from (Tal et al., 2014)
Kurani et al. (1996) were the first that attempted to enrich their stated preference instrument with in-depth information for EVs and at the same time to address the sample size problem in what they describe as a “reflexive design”. Along the same line, Axsen and Kurani (2008) used information from a travel diary that they administered to their respondents in order to visualise their recharge potential. Then for the exercise, they provided them with this visual material as well as with details for PHEV upgrade components to help them take a more informed decision. In contrast with typical choice exercises, this was a design exercise because instead of a choice set the respondents were able to design their own vehicle of preference out of a design envelope. Their target population was comprised from regular
drivers that were likely to buy a new vehicle in the short-term future. Both studies, by using exercises prior to the SP choices intend to mitigate the negative effects from the lack of experience in the valuation of the vehicle attributes.
The combination of SP data with real-world trials has enabled the comparison of user preferences before and after experiencing EVs. Jensen et al. (2014) have developed a panel survey and observed that even though perceptions for driving performance improved after the trial, the drivers were more concerned regarding their ability to maintain their mobility level.
Parsons et al. (2014) investigated the preferences of 3029 randomly selected respondents in the U.S. regarding V2G-enabled electric vehicles. The complexity of their choice experiment is even higher than for the ones described earlier because not only there is the need to make choices for an unfamiliar topic (i.e. electric vehicles), but these choices are combined with a completely new and obscure element, this of selling power back to the grid. Their strategy to overcome this complexity was to follow a sequential, two-step choice process.
Respondents had first to decide between conventional gasoline and typical electric vehicles, and only when they have raised awareness of the attributes of the new technology, were they introduced to the concept of V2G contracts. The comprehension of this methodological approach was supported by conducted focus groups. Another interesting aspect of this study is the treatment of the “yea-saying” effect (i.e. the tendency to overestimate the EV choice because some respondents present a more environmental-friendly aspect of themselves), which might have a significant impact in several adoption studies.
In the area of charging behaviour, the only SP survey known to the authors is this of Daina (2014). Of course, the difficulties are similar with the design of choice experiments for the adoptions of EVs because there is a low degree of familiarity with the hypothetical charging situations presented to the respondents. The author suggests that there are two levels of imagination that the respondents have to go through when they are making their charging choices: the process of using an electric vehicle itself and the fact that recharging their car might demand more planning than just plugging it in. For this reason, they are first presented with a stated adaptation section where they are allowed to configure the charging settings and observe the effects this could have on their daily activity schedule. Then they complete the SP exercise, by choosing between different charging attributes (i.e. final SOC, driving range after charging, duration and cost of charging operation) for a given initial SOC and a
starting charging point tailored to the individual’s completed travel diary. A typical choice card from this survey looks like the one in Figure 2.5.
Figure 2.5: Example of hypothetical charging choice scenario. Reproduced from (Daina, 2014)
2.5 Summary
During the last decade, a great variety of electric vehicle technologies emerged in the automotive market (e.g. BEVs, PHEVs, E-REVs etc.). Among these technologies, PHEVs seem as the most promising alternative to conventional ICEs, at least for the initial stage that drivers will hesitate to rely on pure BEVs. The capability of PHEVs to switch their operation mode from electricity to gasoline when the battery is running low can reduce the effect of
“range anxiety”, one of the main psychological obstacles in the adoption of electric vehicles.
As it was described in section 2.2, the limited range is only one of the numerous parameters that might affect the decision of an individual to purchase an EV.
Several choice modelling frameworks were developed in order to understand the main motives of the drivers and to predict the future sales of EVs. Nevertheless, little attention was given to the choices that take place during the actual use of these vehicles. The prolonged duration of charging events, the uncertainty about future electricity prices or the difficulty in processing complex price signals, as well as the lack of information for public infrastructure availability are some of the problems that an EV driver could encounter on an everyday basis. Therefore, modelling and understanding of charging behaviour is necessary not only to ensure the proliferation of the new technology but also to design new charging
services that will facilitate the use of EVs and turn them into a competitive alternative to conventional vehicles.
The frequency and location of charging events are probably the most prominent characteristics of charging behaviour. Revealed preferences from real world trials differ in this aspect, according to the area of study and the distinctive attributes of each trial (e.g.
recharge potential at home, tariff programmes etc.). In modelling studies, it is usually assumed that charging is directly proportional to parking opportunities, and hence travel data are analysed to estimate the dwelling periods for the individuals.
Moreover, there is a lot of variability in modelling the required energy levels and the SOC that triggers the initiation of a charging event. For simulations, it is typical assumed that individuals start their day with a fully charged battery and hence it is required to reach the 100% SOC overnight. Stochastic methods have been implemented to capture the uncertainty in the time elapsed between two consecutive charging events. The “safety” battery level that drivers would prefer to have available has been also examined from a psychological perspective, highlighting the multidimensionality of the problem. For this reason, attitudinal questions were included in the online survey (Chapter 3) in order to identify the effect of unobserved factors on charging behaviour.
The amount of electricity that is required for a charging event is also a function of driving distance, driving style (e.g. aggressiveness, acceleration etc.), technical characteristics of the charger and vehicle type. Both disaggregate and aggregate modelling of EV recharging effects, strongly depend on the assumptions made for the factors above. Different approaches have been presented, but a more comprehensive summary can be found in Appendix C.
As it will be explained later, the level of complexity of the designed SP survey did not allow the estimation of users’ sensitivity to V2G-related attributes. Nevertheless, considering the analysis for V2G scenarios that is undertaken in Chapter 6, it was necessary to make the appropriate assumptions for the associated demand. The work in this area is still in a premature stage, yet, some guidelines for the relationship types between customers and operators (i.e. contractual and non-contractual) and the respective benefits for the former were discussed.
Out-of-home charging choices are strongly interrelated with the choices for parking (type and location) and this is an area where gaps have been identified in the existing literature.
Future parking policies have to be adapted to the increasing deployment of charging
infrastructure. For example, charging employees for plugging-in their vehicles at workplaces is not a straightforward task and further research is required. Nicholas and Tal (2013) argue that low charging rates (i.e. 1.2kW) are sufficient for 50%-80% of charging needs at the workplace and that higher rates should be more expensive than electricity at home.
The main attributes that affect parking choices have been analysed so that they can be properly integrated with the joint charging/parking choice-modelling framework of this thesis. Finally, the advantages of applying reservation systems in the parking industry were discussed, since this is a vital component of the suggested revenue management model.
In terms of pricing, charging in private facilities entails a combination of parking costs and electricity costs. The latter might fluctuate based on spatiotemporal energy demands and the importance of DSM strategies in order to avoid peak loads is highlighted. Furthermore, it is
In terms of pricing, charging in private facilities entails a combination of parking costs and electricity costs. The latter might fluctuate based on spatiotemporal energy demands and the importance of DSM strategies in order to avoid peak loads is highlighted. Furthermore, it is