Transport model: overestimated flexibility?

Hey @nworbmot, all,

we employ p-e-s in the eGon project (website, github) to create future scenarios which are subsequently used to parametrize our more detailed models on different spatial scales. One part of p-e-s is the transport model, we would like to use for the motorized individual travel.

We've tried to reconstruct what's happening using the model and the paper and have questions how this model can represent real EVs. We'd appreciate if you could help us here to get a better understanding:

The p-e-s model

Let's assume we activate DSM only, so without V2G the model basically looks like this:

[bus_el] --> [bus_ev] --> [charging load]
         (a)    ^     (b)
                |
                ˅
             [store] (c)

with the following constraints for the nodes:

(a)
- constrained by car availability
- max. nom. power given by charging infrastructure
(b) fixed transport charging load calculated from (partially) time shifted transport load (BASt data).
(c) 1) battery 4 flex - the usable flex. is defined as a fixed share of the battery cap. ->participating cars 2) constraint applies: min. available bat. cap. at certain time of day

I see 2 problems resp. things which I do not understand:

(I) Overcharging

From my understanding, a physical EV consists of (b) and (c). Let's consider the following example:

2 BEVs: 100kWh battery each, initial SOC(t=0)=50kWh for each car, 10h standing time (grid-connected), target SOC: 100% at t=9
charging point: max. nom. power = 100kW
constant charging assumed -> total charging power at (b): P(t=0..9)=10kW
1 BEV (50%) participating in DSM (bev_availability in cfg) -> storage capacity: 100kWh

The model allows to fill the storage additionally to the constant load of 10kW (which is ok as long as the energy drawn from [bus_el] to [bus_ev] (a) does not exceed a max. of 100kWh). But In the end 200 kWh could have been drawn, right? Then the cars "leave" (no load at (b)) but the storage part of the energy remains available for e.g. other cars, doesn't it?

This leads us to

(II) Energy transfers between EVs

The model allows energy transfers between EVs that are not necessarily connected to the grid. While the max. charging load is restricted by EV availability at (a), the discharging of the storage (c) isn't. So even when an EV isn't available i.e. not connected to the grid, it is possible to transfer energy from storage (charged before when EV was present) to load (e.g. to another EV).

Due to these two aspects I'd expect the available flexibility for the system to be overestimated, but I'm not sure whether we've missed something here. I guess the model wasn't made for just a few EVs and effects might get less significant due to the highly aggregated data on European level? Are there any further implicit assumptions (e.g. in the data)?

Yes, the model assumes a very high level of aggregation, e.g. hundreds of thousands of BEVs at every node, so there is a significant smoothing effect between the behaviour of individual BEVs. Note that the availability constraints are never binding in this model (because the charging power is so massive compared to average consumption). What drives the results is the storage and the requirement to fill the SoC in the morning (which prevents storing for multiple days). The main problem in my eyes is the model assumes the cars are connected to the grid during the day at workplaces, which allows it to absorb solar PV peaks. If there is no workplace charging, the flexibility for solar PV is overestimated. I think Martin Klein did a survey of good EV modelling approaches at some point which might be better for a smaller number of vehicles.

PyPSA / pypsa-eur