Myopic Implementation

[lfreese]

Feature Request

Develop a myopic implementation of foresight rather than perfect.

Suggested Solution

Solution is in progress, modifying planning_horizons

Additional Info

No response

[lfreese]

Two continuing issues in myopic approach (all in solve_network.py):

• 1) I've updated the n.add() for each component to be less verbose as n.add(nm, df_idx, **df.loc[df_idx]) [line 1558], but this leads to Generators missing passed attirbutes from the df such as land_region, heat_rate, etc. because of the warning "WARNING:pypsa.components:Generator has no attribute land_region, ignoring this passed value."

The only solution I've found to this is to re-add in a n.madd() for only the Generators, with each individual passed value being named. n.madd() is deprecated in future pypsa versions so this is not ideal to have

• 2) Line 1594 n.storage_units.loc[:,'state_of_charge_initial'] = n.storage_units_t.state_of_charge.loc[planning_horizon].sum() leads to an infeasible solve. The purpose of this was to carry over the state of charge from the previous time period and set that as the state_of_charge_initial for the next time period.

Relevant code in https://github.com/lfreese/pypsa-usa/tree/issue-518

[lfreese]

Problem 2 above was solved and has been pushed. Needed to be: n.storage_units.loc[:,'state_of_charge_initial'] = n.storage_units_t.state_of_charge.loc[planning_horizon].iloc[-1] to get the final snapshot state of charge

[lfreese]

@ktehranchi have you pushed your changes to the opts (the SAFE/SAFER) ones you discussed? If so, I'll pull those and add the sns_horizon to them so that they work with myopic. If those are still being developed, I think this might be ready for a pull request. It doesn't do quite as well in the full year snapshot recreating the perfect foresight as it did in a 2-3 day snapshot, so I'm not sure if we feel like that should preclude pulling this, or if that's reasonable.

Figures here show examples of the difference in production and capacity additions between a 2030 myopic and 2030 perfect foresight. Config file attached.

config.myopic.txt

The key gaps in production are coming in with coal, OCGT, and CCGT, for example:

But the profiles are looking quite similar.