feat: add storage expansion results to Grid

[danielolsen]

Pull Request doc

Purpose

• Bugfixes to storage data preparation (follow-up to #101).
  • some sloppy refactoring on my part, plus
  • we had not added the Switch storage module to the list of modules to import, so Switch was ignoring all the storage data we were providing, so we did not realize that we were not providing the full set of required information.
• Add storage expansion results to Grid (serves #49).

What the code is doing

• In const.py: the previously mentioned bugfix constants.
• In grid_to_switch.py: the previously-mentioned bugfix logic
• In helpers.py:
  • We add a recover_storage_buses function which translates Switch storage expansion candidates into bus IDs (following the pattern from #101)
  • We add a split_plant_existing_expansion function which separates a list of plant IDs into real and hypothetical (this was needed to fix a bug that popped up within add_gen_upgrades_to_grid, when existing generators with Pmax = 0 are present in the input data).
• In switch_to_grid.py:
  • extract_build_decisions is expanded to return generation, transmission, and storage energy decisions.
  • in add_gen_upgrades_to_grid, we make some small changes to indexing to ensure that we only process generator expansions (not storage, not existing generators).
  • We add a new function add_storage_upgrades_to_grid, which does what's expected following the name.

Testing

Tested manually on results files generated both with and without storage candidates.

Code used to prepare inputs with storage candidates:

>>> import os
>>> import pandas as pd
>>> from powersimdata import Scenario
>>> from switchwrapper.prepare import prepare_inputs
>>> scenario = Scenario(599)
>>> grid = scenario.get_grid()
>>> profiles = {
...     "demand": scenario.get_demand(),
...     "hydro": scenario.get_hydro(),
...     "solar": scenario.get_solar(),
...     "wind": scenario.get_wind(),
... }
>>> timepoints = pd.read_csv("timepoints_input_v2.csv", index_col=0)
>>> timestamps_to_timepoints = pd.read_csv("slicing_recovery.csv", index_col=0).squeeze()
>>> prepare_inputs(grid, profiles, timepoints, timestamps_to_timepoints, "prepared", storage_candidate_buses={2010001, 2010002})
Please enter base study year (normally PowerSimData scenario year): 2030
Please enter the number of investment stages: 1
Single stage expansion identified.
Please enter investment period year, separate by space: 2030
Please enter start year for each period, separate by space: 2030
Please enter end year for each period, separate by space: 2030
>>>

To analyze the results:

>>> import pickle
>>> from powersimdata import Scenario
>>> from switchwrapper.switch_to_grid import construct_grids_from_switch_results
>>> filename = "path_to_results_file\results.pickle"
>>> with open(filename, "rb") as f:
...     results = pickle.load(f)
...
>>> grid = Scenario(599).get_grid() # the grid used as input to this Switch run
>>> all_grids = construct_grids_from_switch_results(grid, results)
>>>

For this example, there are storage results, as verified via:

>>> import re
>>> sorted({re.search(r"(.*)\[", v).group(1) for v in results.solution._list[0]["Variable"]})
['BuildGen', 'BuildLocalTD', 'BuildStorageEnergy', 'BuildTx', 'ChargeStorage', 'DispatchBaseloadByPeriod', 'DispatchGen', 'DispatchTx', 'GenFuelUseRate', 'StateOfCharge', 'WithdrawFromCentralGrid']

But no storage capacity is build, so there is no storage added to the resulting Grid:

>>> all_grids[2030].storage
{'gen': Empty DataFrame
Columns: [bus_id, Pg, Qg, Qmax, Qmin, Vg, mBase, status, Pmax, Pmin, Pc1, Pc2, Qc1min, Qc1max, Qc2min, Qc2max, ramp_agc, ramp_10, ramp_30, ramp_q, apf, mu_Pmax, mu_Pmin, mu_Qmax, mu_Qmin]
Index: [], 'gencost': Empty DataFrame
Columns: [type, startup, shutdown, n, c2, c1, c0]
Index: [], 'StorageData': Empty DataFrame
Columns: [UnitIdx, InitialStorage, InitialStorageLowerBound, InitialStorageUpperBound, InitialStorageCost, TerminalStoragePrice, MinStorageLevel, MaxStorageLevel, OutEff, InEff, LossFactor, rho, ExpectedTerminalStorageMax, ExpectedTerminalStorageMin]
Index: [], 'genfuel': [], 'duration': None, 'min_stor': None, 'max_stor': None, 'InEff': None, 'OutEff': None, 'LossFactor': None, 'energy_price': None, 'terminal_min': None, 'terminal_max': None}

Time estimate

30 minutes.

[BainanXia]

Correct me if I'm wrong: there are two parameters that we need to extract for each storage expansion candidate: 1) capacity, which is specified within the group of generators, i.e. BuildGen, and it gives the maximum input/output power of a storage device 2) energy, which is specified within the particular group "BuildStorageEnergy", and it gives the maximum energy that a specific storage device could store. What is the relationship of the same entry between different investment years? Are they cumulative values that we could take directly or deltas that we need to do a cumulative sum?

[danielolsen]

Correct me if I'm wrong: there are two parameters that we need to extract for each storage expansion candidate: 1) capacity, which is specified within the group of generators, i.e. BuildGen, and it gives the maximum input/output power of a storage device 2) energy, which is specified within the particular group "BuildStorageEnergy", and it gives the maximum energy that a specific storage device could store.

This is correct.

What is the relationship of the same entry between different investment years? Are they cumulative values that we could take directly or deltas that we need to do a cumulative sum?

This is a very good question! I don't know, but we have the same concerns for generation and transmission. Currently, we are processing the Switch results as if they are cumulative, but that could be wrong. It will need further investigation.

[danielolsen]

I added a couple more commits so that we can successfully call get_pg even when there are some storage candidates in the inputs, and therefore storage dispatch in the results:

• match_variables and parse_timepoints now have a new parameter value_name, rather than hard-coding that the returned variable values are labaled "capacity" (as we did when we originally set up extraction of grid build decisions).
• We pass value_name="dispatch" within SwitchExtract._get_parsed_data
• We index into this MultiIndex (`("capacity", foo), ("capacity", bar), ...) to only return the PG values of the plants, not the storage devices.

[BainanXia]

Looks good to me. We can merge this once @YifanLi86 is happy.