Skip to content

Latest commit

 

History

History
182 lines (124 loc) · 8.88 KB

todo.md

File metadata and controls

182 lines (124 loc) · 8.88 KB
Raw

todo:

(- means to do, + means done)

  • make sure balancing_areas() works with new generator tables

  • update save_mip_results code to use final capacity from GenCapacity.csv or gen_cap.csv instead of calculating it internally (or at least take account of economic retirements and --retire early flag)

  • resolve possible issues where build_year coincides with a model year (Switch may treat these as being built at the start of the period or end or something, but we need some treatment that will work across steps of myopic models)

  • Best effort​ (all of the above)

  • Write a define_scenarios.py script to auto-generate scenarios.txt (for all possible cases), depending on whether it is a pre-chain year, post-chain year, which transmission and co2 costs are used, whether retirement or ramp limits are enabled, etc. (maybe define these in switch_params.yaml somehow?)

    • maybe: grapple somehow with the fact that powergenome thinks it's setting up scenario dirs, but really it's data dirs, and then a scenario consists of a data dir plus some other settings

low-priority:

  • make dirs a more readable color in ls on hpc

  • setup hpc to share history right away between sessions

  • add code to powergenome.util.init_pudl_connection to give an error if the PG_DB file doesn't exist, instead of creating a new empty one

done:

  • propagate ramp rate limits and minimum load from PG to switch input files (see notes below)

  • chain retirement decisions between myopic model stages

  • prevent creation of duplicate rows in the hydro tables and maybe others (these get one per generator model year; maybe generate from gen_info instead?)

  • make sure switch can run with gen_build_predetermined after start of study (may cause problems in post solve); if not, drop any unusable gens in pg_to_switch

  • add retire_early and retire_late options to gen_build_suspend and set that in options.txt; will also need to add the gen_can_retire_early flag, so we can use this module full-time

  • finish implementing retirement logic: prevent plants from doing age-based retirement if they don't already have a firm retirement date. (We did this by implementing PowerGenome's approach for the retirement cases, which is to set no retirement age, which then gets translated into a 500 year life, and also to look at the New_Build and/or Can_Retire flags to see which ones are allowed to have economic retirement.)

  • add hydrogen as a fuel

  • add gen_can_suspend and gen_can_retire_early settings (both default to 0) to gen_info.csv for use by gen_build_suspend (if gen_can_retire and not gen_can_suspend, require later suspensions >= previous suspensions for each gen/vintage)

  • add --retire {early, mid, late} flag for gen_build, to control whether retirement (and construction) are moved to the start, middle or end of the period when they occur (which affects how we decide what is online during that period)

  • resolve cases where gen_build_predetermined for 2040 has capacity that wasn't in the 2030 file (e.g., with date of construction in 2025) or has less capacity than the 2030 file (e.g., BASN_natural_gas_fired_combined_cycle_1,1994 is 141 Mw in 2030 and 53 MW in 2040). Could the second part be something about clustering plants that are already scheduled to retire? Do we not use a uniform retirement age for each cluster?

  • get fixed O&M to show up for existing plants so there is money to save by retiring them

  • update transmission scenario definitions to all branch off base_short instead of using separate dirs for most

  • setup .slurm scripts to run the scenarios as array jobs with different jobs for different years and dependencies between them

  • eventually: more cases, e.g., longer strips of weeks and maybe tx/carbon cost cases)

  • implement transmission-limit cases in pipeline (caps)

  • generate case data: base_short, base_short_current_policies, base_short_no_ccs

  • create a module to limit transmission additions on each corridor to a specified amount

create a script to setup the alternative cases (alternative input files for base_short; flags in scenarios.txt):

  • base:
    • need to add a script or module to generate next-stage gen_build_predetermined from prior stage
  • transmission limits
    • update pg_switch.py: use powergenome.GenX.network_max_reinforcement to find expansion limits per line per period
  • carbon slack
    • update pg_switch.py: generate multiple carbon price files at end, maybe with settings from switch_params.py
  • Early plant retirement​
  • Foresight with sample weeks​
  • save results in MIP_results_comparison

  • Unit commitment​ (ramp limits)

    • update pg_switch.py (Ramp_Up_Percentage, Ramp_Dn_Percentage and maybe Min_Power returned by add_misc_gen_values)
    • give these names and implement the percent ramp rules in a side module gen_ramp_limits (should these be a percentage of capacity committed in previous timepoint or current timepoint?)

  • run all the models

old notes:

put everything in Switch-USA-PG dir/repo

  • PowerGenome subpackage (points to develop branch for now)
  • MIP_results_comparison subpackage

Use git commit --author="Rangrang Zheng [email protected]" to set the author to Rangrang for the iniial commits

Create a switch directory.

Move SWITCH_INPUTS_rr_east/self_defined_modules to switch/study_modules

*** Make sure Rangrang can run with only Switch 2.0.7 (not pre-release) and study_modules. (Or else with the next_version branch from the main switch repo.)

Add download_pg_data.py script to download corrected-20z-resource-groups and PowerGenome Data Files into a pg_data folder.

Add a module to study_modules to chain the myopic models (needs to know where the next stage model will be, e.g., --next-stage-inputs-dir in/this_scenario_2050), based on to_myopic.py. This will then copy gen_build_predetermined.csv and gen_build_costs.csv to the other dir, keeping anything already in those files in the other dir from later than the end of this study (but not earlier, because we need to transfer building plans from this study and any inherited from prior studies).

Update pg_to_switch.py instructions to use settings from MIP_Results_comparison/case_settings/26-zone.

Update pg_to_switch.py to

  • put the switch inputs in switch/in/*
  • write scenarios_2030.txt, scenarios_2040.txt, scenarios_2050.txt and scenarios_foresight.txt to run the scenarios. These should put outputs in switch/out/*.

Find script that converts model outputs to standard form for comparisons (line 1 to 380 of to_myopic.py), and move it to switch/. Add module that runs that script automatically to create standard model outputs in MIP_compare_results when the model finishes.

Update settings in MIP_compare_results to generate blank production cost model(s)

Add script to MIP_compare_results to setup operational simulation stage:

  • add alternative gen_build_predetermined.csv and gen_build_costs.csv files for each model output summary
  • create a scenarios.txt file to run for each of these

old notes:

hard to see a clean way through this: there's a lot of study-specific code and data in the pg_to_switch pipeline, which can't be easily split into generic US stuff plus study-specific choices

  • e.g., maybe we should have a settings file for the Switch version of PowerGenome, where we specify the gen_info columns we want (?) and then the Switch plugin adjusts the settings as needed to get the right upstream generator columns from PowerGenome, then uses that to produce the required output columns; maybe our plugin could have a big list of final tables.columns, source columns.tables they depend on and conversion function to produce them. Then we call powergenome to get the required tables and columns, then call the converter functions to translate them.
  • also, what to do with data that doesn't exist from a public source and is currently specified in settings.yml or hard-coded? e.g., retirement ages for equipment
  • note: it looks like settings.yml should mainly be static for Switch models, though it may vary depending on the features requested, so e.g., requesting advanced spinning reserves in Switch may require changing some entries in settings.yml, but the user shouldn't know what those are and we do. On the other hand, we should let the user specify zones, resolution, etc. So we should have a switch_settings.yml (including module selection?) that we convert into settings for powergenome.

repos:

  • powergenome (just gets installed)
  • PG_TO_SWTICH_refactor (overlaps MIP, but we want this to evolve into freestanding repo)
  • MIP (PG_TO_SWTICH_refactor)

collect MIP powergenome data files together into one folder (maybe in the PowerGenome/data folder in the PowerGenome dev branch that we are using, similar to Rangrang's email on 11/8/23?)

collect public powergenome data into a different location that can be bundled or configured automatically for public use

update conversion_functions.py to create Switch 2.0.7 outputs instead of 2.0.6