[lenzim97]

Clip e_initial of e-mobility stores to their SoC band

Fixes #846

Problem

e_initial of the battery_storage stores refers to 1 January 00:00, but it was used for every start snapshot and every rolling horizon window. At other snapshots it is often outside the SoC band, which leads to infeasibilities in highly resolved networks or requires (negative) load shedding.

Changes

• New function adjust_e_initial_emob() in tools/utilities.py: clips e_initial of all battery_storage stores to [e_min_pu, e_max_pu] * e_nom at the first snapshot of the optimization.
• The function is called before lopf/grid optimization, the pre-market model, sclopf and every rolling horizon window.
• Rolling horizon: from the second window on, battery_storage stores take e_initial from the result of the previous window, like the other non-seasonal stores.
• Removed the previous fix in disaggregate/temporal.py, which only clipped e_initial to e_max at the first snapshot of the temporal disaggregation. This is now covered by the rolling horizon.

Testing

Run with eGon2035, snapshots 240–408, 30 AC clusters, market optimization with rolling horizon (168/120) and grid optimization, linopy:

• e_initial is within the SoC band in the pre-market model, the market model and the grid optimization.
• No jumps of the state of charge at the rolling horizon boundaries.

Not tested with an higher resoluted grid

[ClaraBuettner]

This function is called in that many functions, is it really needed like this?

[lenzim97]

Transfered into adjust_network() now

[ClaraBuettner]

The checks with black are failing in this file. Could you run black again?

[lenzim97]

just applied for the files i changed