Synthetic electric vehicles time series for energy system modelling (as part of the SEDOS project)

Electric vehicle profiles are a complex matter to create and therefore require sophisticated tools. To increase transparency, there are open tools such as venco.py. However, even with open tools, transparency is often not guaranteed, as the data is not openly available due to personal data protection regulations. For this reason, as part of the SEDOS project, the proprietary data from the MOP study (German Mobility Panel) was synthetically reproduced using a machine learning approach without distorting its character. This data set includes both the synthetic input data and the resulting time series from venco.py.

There are four kinds of time series included in the dataset in order to describe and limit the transport sectors dispatch for energy system modells:

Name
Type
Unit

demand_timeseries
Fixed
Normalized sum 1

sto_min_timeseries
Lower
Normalized between 0 and 1 as fraction of SOC

sto_max_timeseries
Upper
Normalized between 0 and 1 as fraction of SOC

availability_timeseries_fixed / availability_timeseries_max
Fixed/Upper
Normalized to maximum

All vehicles have to serve a given demand profile (demand_timeseries) that describes the normalized fraction of the scalar annual demand (demand_annual) that is given in the tra_scalar table. For cars and trucks, this demand profile is a driving profile, as it is assumed that the refueling of the fuel vehicles is balanced across the fleet and takes place during the driving time. For the flexibly modeled e-vehicles, the driving profile specifies when the storage tank is discharged and is a boundary condition for optimization. Overhead line trains also have a driving profile as a demand time series, as the energy demand occurs at the same time as the vehicle is in motion. For all other vehicles (buses, ships, airplanes, special vehicles), however, a tank profile is specified, as it is assumed that the driving process is decoupled from the tank process in terms of time. The vehicles are refueled at the depot after the working day. Refueling is therefore more concentrated in the afternoon.

The flexibly modelled vehicles (cars and trucks) are furthermore constrained. The SOC of the vehicle batteries are restricted by a band (sto_max_timeseries/sto_min_timeseries) that represents the maximum/minimum SOC that the fleet should be able to provide in order to fulfill user demands.

The charging power of an electric vehicle is limited by the maximum output capacity of the wallbox when charging a single vehicle. It furthermore varies over time when considering a whole fleet since the number of vehicles connected to the grid is not constant during the day. To take this circumstance into account the maximum power a single car can be charges with is given exogeniously with the parameter capacity_p_unit, describing the maximum power of a wallbox. The percentage of the fleet that can charge with this maximum power at each point of time is represented by the time series availability_timeseries_max. In order to calculate the maximum power the fleet can charge at a certain point of time, the time normalized time series needs to be multiplied by capacity_p_unit (individual wallbox capacity) as well as with the vehicle number.

As an exception, the inflexible charging vehicles (e.g. tra_road_mcar_inflex_uni) are not limited by a maximum charging capacity, but the charging capacity is specified by a fixed time series (availability_timeseries_fixed). There is no optimization intended. Otherwise, the principle is identical. All BEV-profiles are calculated with the DLR open source tool venco.py. Since the needed input data by the Mobility Panel MoP (individual trips) is restricted and not openly available, the data was recreated using a ML approach. 

For more information about the project, please visit: https://sedos-project.github.io/organization/