search

RAMP-project / RAMP-mobility

A novel application of the RAMP main engine for generating bottom-up stochastic electric vehicles load profiles.
European Union Public License 1.2
22 stars 13 forks source link

Occasional use entails lower-than-expected total energy consumption #33

Open cdemulde opened 1 year ago

cdemulde commented 1 year ago

Hi, I was bumping into this issue: I have been generating load profiles for different numbers of cars, but I have noticed that the total amount of energy used over a year is a lot lower for the generated profiles than what I would expect.

An example: for 26 cars, driving on average 15 km/d and consuming 18 kWh/100km (see for example https://ev-database.org/cheatsheet/energy-consumption-electric-car), electricity use should be around 25 MWh on a yearly basis. When I adjust the necessary input file (d_tot.csv, where I adjust the driven distances for France from 50 and 60 to 15 and 18, resp.), I end up with a profile that over the whole year consumes about 10 MWh, so a lot lower. Am I missing something?

@FLomb mentioned rounding effects on Gitter, playing a bigger role when using smaller amounts of cars (configuring 26 cars for example leads to actual simulation of only 20 cars). Not sure if the whole difference (in the example I gave 25 vs 10 MWh) can be explained by this rounding effect though. Maybe I am making some wrong assumption somewhere?

FraSanvit commented 1 year ago

We witnessed already the issue you are highlighting. After several checks, we realised that mismatch is related to the mobility pattern. Indeed, due to the occasional_use parameter, trips in the free-time windows (main windows have higher occasional_use values) are not generated. This affects the amount of average daily distance per vehicle and so the resulting charging demand. We are working on that. For some hints, you can check this link; https://github.com/FraSanvit/RAMP-mobility/tree/master

cdemulde commented 1 year ago

That indeed explains the large discrepancy in my case I think, I was mostly using free-time windows for the generation of my profiles, so the impact of those was proportionally a lot larger. Thanks for the pointer @FraSanvit, looking forward to a final solution!