by Merel Steenbrink, Elenna Dugundji and Rob van der Mei, CWI

At the Stochastics research group at CWI we use socio-economic features of neighbourhoods to predict the demand for charging stations for electric vehicles. Based on a large set of behavioural data, a discrete choice model is estimated, and the utility of charging stations in every neighbourhood is calculated. In this way the demand can be predicted and the municipality of Amsterdam can proceed proactively.

The municipality of Amsterdam would like to stimulate the use of electric vehicles in the city. To this end, a substantial number of charging stations [L1] have been constructed in the city. Over 1,000 stations are already in operation, but the city wants to further increase this number to 4,000 by 2018  [L2, L3, L4]. The policy is quite encouraging: if you buy an electric vehicle, and there’s no charging station within a 200 metre radius of your residence, you can apply for one in front of your door. However, the process of investigating suitable places is a time-consuming task. If it were possible to predict where the demand for charging stations will be, the scaling-up process could take place more effectively.

In this research a Discrete Choice Model was estimated to predict the demand for charging stations. Every charging activity can be seen as a choice for a certain station over other stations. In this way the utility of every station can be determined as a function of characteristics of the stations. When this function is estimated, the utility of future stations can be determined, based on their characteristics. In this way the municipality can focus on potential neighbourhoods, where the utility is high, and already investigate the appropriateness of these locations even before any requests have been made.

To estimate the model, behavioural data from 2012, 2013 and 2014 were used. The data describe the charging activity at every station, which reveals how often, how much and how long a poll has been used. To be able to use different characteristics, and to decrease the size of the choice set, the choices have been aggregated to the area of neighbourhoods. Every choice for a charging station is a choice for a neighbourhood. Socio-economic data from the Central Bureau for Statistics (CBS) have been incorporated into the model. The model takes into account: average number of cars per household, percentage “Western” inhabitants (persons with nationality from Europe, North America, Indonesia or Japan), the percentage of apartments. Furthermore it considers the mean percentage of time per day that the stations in each neighbourhood were used. Because the data were aggregated there was a compensating term with the number of stations per neighbourhood.

We estimated a Discrete Choice Model which calculates the probability of an electric car driver choosing a charging station in a certain neighbourhood [1],[2],[3]. We calculated this probability for every neighbourhood and in this way obtained a ranked list of neighbourhoods with high and low probabilities. We thereby are able to advise the municipality in which neighbourhoods to focus their expansion activities.

Until now we have incorporated into our model socio-economic characteristics for which data were readily available from the CBS. Further work coupling with other sources of data is likely to identify other relevant variables - for instance, proximity to malls or large office buildings may turn out to be important. These factors will be added to the model as we develop it. The model also does not currently differentiate between day and night time charging sessions. This consideration could reveal additional variation in the usage patterns of the charging stations in each neighbourhood.

Some neighbourhoods with low population density have no aggregate socio-economic statistics reported by CBS to protect the privacy of these residents. These neighborhoods have been currently excluded from this investigation due to lack of descriptive data. However, many of these neighbourhoods – largely industrial areas or sporting facilities – may require charging stations. Additional data on land use and the built environment may allow us to include these neighbourhoods.
Finally, when we have ascertained which neighbourhoods need charging stations, the next step is to determine the exact locations of these charging stations in each neighbourhood.

Links:
[L1] https://chargemap.com/city/amsterdam
[L2] http://kwz.me/U3
[L3] http://kwz.me/U5
[L3] http://kwz.me/U6

Refernces:
[1] F. Goetzke, R. Gerike, A. Páez, E.R. Dugundji: “Social Interactions in Transportation: Analyzing Groups and Spatial Networks”, Transportation, 42 (5), 2015, 723-731.
http://dx.doi.org/10.1007/s11116-015-9643-9
[2] M. Maness, C. Cirillo and E.R. Dugundji: “Generalized Behavioral Framework for Choice Models of Social Influence: Behavioral and Data Concerns in Travel Behavior”, Journal of Transport Geography, 46, 2015, 137–150. http://dx.doi.org/10.1016/j.jtrangeo.2015.06.005
[3] E.R. Dugundji, J.L. Walker: “Discrete Choice with Social and Spatial Network Interdependencies: An Empirical example using Mixed Generalized Extreme Value Models with Field and Panel Effects”, Transportation Research Record, No. 1921, 2005, 70-78. http://dx.doi.org/10.3141/1921-09

Please contact:
Elenna Dugundji
CWI, The Netherlands
This email address is being protected from spambots. You need JavaScript enabled to view it.

Next issue: July 2018
Special theme:
Human-Robot Interaction
Call for the next issue
Image ERCIM News 105 epub
This issue in ePub format

 
Get the latest issue to your desktop
RSS Feed