FINITO Link by mmowers · Pull Request #128 · ReEDS-Model/ReEDS

mmowers · 2026-06-23T23:06:38Z

Summary

This PR provides a linkage to the FINITO model, which is currently a private repository at https://github.com/NatLabRockies/FINITO). The linkage allows the full electricity sector detail of the ReEDS model to be used in combination with the industrial (and other) sector representation of the FINITO model.

The linkage is turned off by default, so this PR itself should not have impact on ReEDS results (although very small differences were observed in out years, presumably because the ReEDS model itself was touched, even if the matrix is identical).

For a high-level overview of FINITO and the linkage points see this slide deck.

See companion FINITO PR: https://github.com/NatLabRockies/FINITO/pull/438

Technical details

Turning on the linkage adds the constraints of the FINITO model to those of the ReEDS model, and adds the two objective functions together.

To turn on the linkage, clone the FINITO model and checkout the bs/reeds_link2 branch, unless this branch has already been merged into main, at which point main.

Configure these switches:

GSw_FINITO_Link: This switch turns on the linkage
finito_dir: This switch needs to point to the FINITO directory as a complete, absolute path
finito_cases_file: This switch should be the suffix of the cases file in the FINITO directory, which contains the corresponding FINITO scenario configuration
finito_case: This switch is the specific case in finito_cases_file to be run
first_year_finito: This switch is the first year that the linkage will be implemented

The reeds conda environment should be used for the linked model.

Linked runs result in the the same set of ReEDS outputs, but new outputs are also added from FINITO in the finito subdirectory of the run directory.

Relevant sources or documentation

FINITO documentation is at https://github.com/NatLabRockies/FINITO

Validation, testing, and comparison report(s)

A comparison of default (Ref) was made berfore and after this change, showing no changes, as expected:
reeds_report_ref_2026-08-05.zip

Running the USA_fast case with the linkage turned on results in moderate generation differences to the ReEDS standalone "USA_fast", but similar amonts of fossil and renewables by 2050:

Electricity load differences show that FINITO endogenous industrial electricity demand is lower than the reference demand that was removed from the EER load profiles, even in the near term:

Gas price differences align with gas generation differences by 2050, although alignment isn't perfect for the years before:

See more results of the comparison here:
reeds_report_finito_vs_usa-fast_plus4_2026-08-08.zip

Also, a more thorough comparison of the reference industrial electricity load that was removed from ReEDS load profiles, as compared to the endogenous finito load that was added back in, is here:
finito_load_check_2026-08-06.zip

Checklist for author

Details to double-check

General information to guide review

Zero impact on results of default case
No large data file(s) added/modified
No substantive impact on runtime for full-US reference case
No substantive impact on folder size for full-US reference case
No change to process flow (runreeds.py, reeds/core/solve/solve.py)
No change to code organization
No change to package requirements (environment.yml or Project.toml)

Did you use LLM tools (chatbot or copilot) in the preparation of this PR? If so, describe how

Tag points of contact here if you would like additional review of the relevant parts of the model

…/finito_link

…NITO output

…O for resource adequacy calcs

…aint marginal

bsergi · 2026-08-13T13:47:32Z

+* remove any FINITO scaling, and then apply ReEDS scaling
+$ifthene.linked_objective Sw_FINITO_Link==1
+        + cost_scale * ( Z_finito.l(t)$t_finito(t) * deflator('2018') / obj_scale )
+$endif.linked_objective 


The absolute value of the error check gets larger even after accounting for the FINITO objective function. It's still small but I'm not sure why this is happening yet, so would be interested in any suggestions folks have on this.

I think it might have to do with your fuel cost reporting. You have $tfuel(t) on biopower and gas price reporting, but not for nuclear and coal.

wesleyjcole

This was interesting to see. Nice work on helping to get this where it is.

You do use "linked model" a lot throughout. It might be nice to be more specific with the language there because we have had other linked models, so those comments can get confusing.

wesleyjcole · 2026-08-19T20:15:59Z

 "h2_usage(r,allh,t)",metric tons/hour,total H2 usage by hour (H2-CT/CC and non-electric H2 consumption),,,
 "load_cat(loadtype,r,t)",MWh,Annual exogenous load by category,,,
 "load_rt(r,t)",MWh,Annual exogenous load,,,
+"loadvar_rt(r,t)",MWh,Annual LOAD during rep periods,,,


This says "during rep periods," but you sum over all h, so it includes stress periods as well.

We restrict h to rep periods at the top of report.gms though:

ReEDS/reeds/core/terminus/report.gms

Lines 121 to 122 in bc80431

* Restrict operational outputs to representative timeslices and seasons

h(h)$[not h_rep(h)] = no ;

But if you want to be extra careful (in case we decide to change that down the road and miss it here), you could add a $h_rep(h) to the sum.

wesleyjcole · 2026-08-19T20:17:21Z

 RE_gen_price_nat(t),2004$/MWh,marginal cost of the national RE generation constraint,,,
 "repbioprice(r,t)",2004$/MMBtu,highest marginal bioprice of utilized bins for each region,,,
 "repgasprice(cendiv,t)",2004$/MMBtu,highest marginal gas price of utilized gas bins for each census division,,,
+"repgasprice_finito(cendiv,allh,t)",2004$/MMBtu,highest marginal gas price of utilized gas bins by hour for each census division when using linked model,,,


It's not clear from the comment whether this is the price from ReEDS sent to FINITO or from FINITO sent to ReEDS.

wesleyjcole · 2026-08-19T20:24:53Z

+* TODO: activate after FINITO pricing PR is merged
+*       + smax{cfp$[gasp(cfp)$map_cf_fe(cfp,'NG')$valcft(cfp,t)], cf_markup(cfp,'NG','Electric_Power',cendiv,t) }


Can this be left out of the code and added as an issue instead?

wesleyjcole · 2026-08-19T20:27:24Z

+* remove any FINITO scaling, and then apply ReEDS scaling
+$ifthene.linked_objective Sw_FINITO_Link==1
+        + cost_scale * ( Z_finito.l(t)$t_finito(t) * deflator('2018') / obj_scale )
+$endif.linked_objective 


I think it might have to do with your fuel cost reporting. You have $tfuel(t) on biopower and gas price reporting, but not for nuclear and coal.

wesleyjcole · 2026-08-19T20:28:57Z

+            shutil.copy(os.path.join(reeds_path,f), os.path.join(case,f))
        shutil.copy(f, Path(case, f))


Should this be here twice in a row like this? Is line 107 needed?

wesleyjcole · 2026-08-19T20:58:51Z

+        #TODO: why so much shifting? validate the dynamics here with FINITO team
+        #TODO: also check against reference quantity


More TODOs to remove.

wesleyjcole · 2026-08-19T21:00:21Z

 demandscen,Demand scenario (see inputs\load),N/A,AEO_2026_baseline,
 dr_shedscen,DR shed scenario,demo_data_January_2025,demo_data_January_2025,
 evmcscen,EVMC scenario for load and supply curve and flexibility,Baseline,Baseline,
+GSw_FINITO_Link,Switch to enable linkage with FINITO,0; 1,0,


I think we should stick with convention and keep all the GSw_* switches together, so I think this is worth moving down with the rest of them.

wesleyjcole · 2026-08-19T21:01:26Z

-pras_unitsize_source,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,r2x
-pras_vre_combine,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,1
-pras_samples,,,,,,10,10,10,,,,,,,,,,,,,,,,10,10,,,,,,
+,Default Value,Pacific,USA_defaults,Mid_Case,USA_decarb,github_Pacific,github_Everything,github_MA_county_CC,Pacific_CC,Pacific_weks,Pacific_full_year,Interday_storage,Pacific_2020,Pacific_rep15,WY_county,WECC_county,PJM_county_CC,NYVT_mixed,OR_water,MonteCarlo_Random,MonteCarlo_LHS,Everything,Simple,USA_fast,USA_faster,MultiMetricRA,Pacific_DR,Pacific_MGA,Pacific_LoadSite95,MARICTNYNJPAOH_Offshore,R2P,FINITO,FINITO_Pacific


Why add two FINITO cases here instead of just one?

Agreed, I would keep it to a single small test case.

wesleyjcole · 2026-08-19T21:05:19Z

+    load_finito = load_finito.rename(columns=p2zone).T.groupby(level=0).sum().T
+
+    # allocate annual load to hours, assuming flat demand
+    # TODO: should this use h_weight_finito?


Another TODO.

wesleyjcole · 2026-08-19T21:22:41Z

+        load_hourly_finito = hourly_load.get_hourly_finito_load(inputs_case)
+        load = load + load_hourly_finito
+
    ### Subset to keep_modelyear if provided
    if keep_modelyear:
        load = load.loc[keep_modelyear].copy()
    ### load.h5 is busbar load, but b_inputs.gms ingests end-use load, so scale down by distloss


If I'm reading things right, get_hourly_finito_load returns end-use load, but this comment says load.h5 is bubar load, so I think you are adding end-use load to bus-bar load.

It also would be good to add docstrings to the functions to clarify which this kind of thing.

patrickbrown4

Thanks, it's cool to see this coming together.

I guess I have a set of general concerns about maintainability - questions like:

What's the proposed approach for keeping the linked models working together? It seems hard when one is not open source.
- How do we know when a change to ReEDS (variable/parameter names, indices, concepts) breaks the FINITO linkage? (We've had such issues with reeds2plexos/r2x; running the r2x test as part of the github runner seems to help a lot.)
There's a lot of duplication of ReEDS data/code within the FINITO repo; what happens when the ReEDS side changes?

I don't think those concerns should keep us from moving forward, but it might be helpful to discuss as a team what the expectations are. For example, the PR template specifies to run the full cases_test.csv for "changes to model structure, spatial/temporal resolution, or other large changes"; since a FINITO case is now included in cases_test.csv, does that mean everyone who makes big changes to ReEDS should get set up to run FINITO? Or do we leave that to periodic checks by a FINITO developer (in which case maybe we should move it from cases_test.csv into a separate file like cases_examples.csv)?

I do think we should consider moving the new python code to a dedicated reeds.finito module to help with testing/maintenance/organization, though I can't think of a good way to do it for the GAMS code.

patrickbrown4 · 2026-08-19T23:06:54Z

+        # Collect all arguments for initialize_price_response.py
+        initialize_price_response_args = f" -c {casedir} -b {BatchName} -cr {caseSwitches['GSw_Trade_PriceResponse_RefScen']} -l {caseSwitches['GSw_FINITO_Link']}"
+        # Call initialize_price_response.py file before starting the runs
+        os.system('python ' + initialize_price_response_path + initialize_price_response_args)


Can you use subprocess.run() instead? It's recommended over os.system() for new code, and is what we use for running PRAS etc.

patrickbrown4 · 2026-08-19T23:08:06Z

+
+    # copy directories
+    os.makedirs(casedir_finito, exist_ok=True)
+    shutil.copytree(os.path.join(caseSwitches['finito_dir'], 'inputs'),os.path.join(casedir,'finito', 'inputs'))


(nitpick) Sorry to quibble but in my opinion, Path makes these a bit more readable since it's shorter

Suggested change

shutil.copytree(os.path.join(caseSwitches['finito_dir'], 'inputs'),os.path.join(casedir,'finito', 'inputs'))

shutil.copytree(Path(caseSwitches['finito_dir'], 'inputs'), Path(casedir, 'finito', 'inputs'))

patrickbrown4 · 2026-08-19T23:11:38Z

+def get_load(inputs_case, sw, keep_modelyear=None, keep_weatheryears=[2012]):
    """
    """
    ### Subset to modeled regions
    load = reeds.io.read_file(os.path.join(inputs_case,'load.h5'))
+
+    ### When running the linked model (GSw_FINITO_Link=1) we add reference load estimates
+    ### for FINITO load back in to use when selecting representative periods


(nitpick) sw is so fast to load that I leave it out of required arguments if case/inputs_case is already included; that simplifies testing and ensures the inputs are in sync. (Would need to adjust the call in reedsplots.plot_seed_stressperiods() too.)

Suggested change

def get_load(inputs_case, sw, keep_modelyear=None, keep_weatheryears=[2012]):

"""

"""

### Subset to modeled regions

load = reeds.io.read_file(os.path.join(inputs_case,'load.h5'))

### When running the linked model (GSw_FINITO_Link=1) we add reference load estimates

### for FINITO load back in to use when selecting representative periods

def get_load(inputs_case, keep_modelyear=None, keep_weatheryears=[2012]):

"""

"""

### Subset to modeled regions

load = reeds.io.read_file(os.path.join(inputs_case,'load.h5'))

### When running the linked model (GSw_FINITO_Link=1) we add reference load estimates

### for FINITO load back in to use when selecting representative periods

sw = reeds.io.get_switches(inputs_case)

patrickbrown4 · 2026-08-19T23:16:55Z


+    ### Rewrite the  finito scalar as GAMS-readable definitions
+    if int(sw['GSw_FINITO_Link']) == 1:
+        pd.read_csv(os.path.join(os.path.dirname(inputs_case),'finito','inputs','scalars.csv')).iloc[:, :-1].to_csv(os.path.join(os.path.dirname(inputs_case),'finito','inputs','scalars.csv'),index=False)


Agreed, it seems risky to me to rely on position. I would explicitly specify the entries to be dropped (ignoring errors if they aren't there).

patrickbrown4 · 2026-08-19T23:20:06Z

 "h2_usage(r,allh,t)",metric tons/hour,total H2 usage by hour (H2-CT/CC and non-electric H2 consumption),,,
 "load_cat(loadtype,r,t)",MWh,Annual exogenous load by category,,,
 "load_rt(r,t)",MWh,Annual exogenous load,,,
+"loadvar_rt(r,t)",MWh,Annual LOAD during rep periods,,,


We restrict h to rep periods at the top of report.gms though:

ReEDS/reeds/core/terminus/report.gms

Lines 121 to 122 in bc80431

* Restrict operational outputs to representative timeslices and seasons

h(h)$[not h_rep(h)] = no ;

But if you want to be extra careful (in case we decide to change that down the road and miss it here), you could add a $h_rep(h) to the sum.

patrickbrown4 · 2026-08-21T17:44:46Z

+* ----- ReEDS-FINITO temporal parameters -----
+*=============================================
+$ifthene.linked_finito_temporal_params %GSw_FINITO_Link% ==1
+$include finito/model/linked_finito_temporal_params.gms


At least I assume this change is required for Windows, but if not we should get rid of %ds% everywhere (in a separate PR)

Suggested change

$include finito/model/linked_finito_temporal_params.gms

$include finito%ds%model%ds%linked_finito_temporal_params.gms

patrickbrown4 · 2026-08-21T17:45:57Z

+*** Fix FINITO decision variables to their optimized levels for this solve year
+$ifthene.linked_finito_varfix %GSw_FINITO_Link% == 1
+* Industrial variable fix
+$include finito/model/finito_varfix.gms


Suggested change

$include finito/model/finito_varfix.gms

$include finito%ds%model%ds%finito_varfix.gms

patrickbrown4 · 2026-08-21T17:51:26Z


 *gas used from each bin is the sum of all gas used
-eq_gasused(cendiv,h,t)$[tmodel(t)$((Sw_GasCurve=0) or (Sw_GasCurve=3))]..
+eq_gasused(cendiv,h,t)$[tmodel(t)$tfuel(t)$((Sw_GasCurve=0) or (Sw_GasCurve=3))]..


Note that I didn't try to check for cases where tfuel(t) should be applied but is not

patrickbrown4 · 2026-08-21T17:59:21Z

+    # Validation check: FINITO reference load should not exceed baseline load.
+    # If it does, clip to zero and report


Should this raise an exception instead?

patrickbrown4 · 2026-08-21T18:29:34Z

+    with open(cases_linked_path, 'r', newline='') as csvfile:
+        reader = csv.reader(csvfile)
+        header = next(reader)


Both are fine but a one-line would be pd.read_csv(cases_linked_path, header=None).iloc[0].values

bsergi and others added 22 commits April 15, 2026 14:35

branch transfer finito linkage

436ade0

branch transfer finito linkage

0424508

Merge branch 'bs/finito_link' of github.com:ReEDS-Model/ReEDS into bs…

3596316

…/finito_link

Merge remote-tracking branch 'origin/main' into bs/finito_link

e76819c

Fix windows bug in outputs for linked ReEDS-FINITO runs

5fd4848

Merge branch 'main' into bs/finito_link

a2c1935

Update to autopop_set.py call

99147d1

Merge remote-tracking branch 'origin/main' into bs/finito_link

c06a183

Merge remote-tracking branch 'origin/main' into bs/finito_link

e74e2c4

Update region switches in cases_finito.csv

6c98dde

Fix bug in report.gms for linked FINITO runs

046b21d

Add comment to report.gms

8cecf3c

Convert USE_ELE_FINITO to busbar load when using in ReEDS

b0720c9

error handling for missing finito gdx file

bf47a9b

error handling for finito gdx

eaaaa7e

Merge with main

32c5b48

Fix bug in standalone ReEDS outputs by putting if statement around FI…

877b591

…NITO output

Update to z54 by default in cases_finito.csv

86c4093

Remove round statement to align with main

74443a6

Cleaning up unnecessary changes for cleaner comparison to main

9688064

Remove old switch from cases_finito.csv

93bd58b

Remove call to removed finito file

d687d47

mmowers marked this pull request as ready for review July 2, 2026 15:50

mmowers added 4 commits July 2, 2026 12:31

Revert to main for Canada switch behavior

5aa7540

Revert to main for rounding of parameters

ea246a9

Add FINITO case to cases_test.csv

88ff50d

Update FINITO case

6310db7

bsergi changed the title ~~Finito Link~~ FINITO Link Jul 6, 2026

bsergi and others added 2 commits July 6, 2026 16:02

clean up finito_dir entry

02f14b8

For now, use only rep period load (not stress period load) from FINIT…

58f1227

…O for resource adequacy calcs

bsergi and others added 19 commits July 31, 2026 11:00

some comment and formatting changes

9f94ef4

add FINITO column to cases_test.csv

2e604e7

fix bug in FINITO cases file setup

df894f1

updates to gas and biomass pricing (need to finalize)

5f3bf1b

Add FINITO_Pacific case to cases_test.csv

d0738d6

Fix bug: Add missing .l suffix

48b4a06

Fix FINITO gas price aggregation bug in report.gms

3f6e960

Fix divide by zero in ReEDS-FINITO repbioprice output

7ee8166

Fix repgasprice_finito weighting

ac23719

Fix repgasprice overwrite for finito-linked years

d5c230e

Fix ng price in finito-linked runs by using the correct finito constr…

49b2f31

…aint marginal

Add finito_load_check.py

d899f5f

Add missing reference lines on finito_load_check.py

52cfb69

Fix gas price reporting for finito-linked runs

b2adca4

Convert finito electricity load from end-use to busbar

386381c

Output annual LOAD.l

f5babb4

clean up old comment

99f61b8

generalize FINITO path in cases.csv

2ca82c0

separate gas price reporting for supply curves and detailed fuels

9ed9a74

bsergi removed the data_changes label Aug 11, 2026

bsergi requested review from patrickbrown4 and wesleyjcole August 11, 2026 20:09

mmowers and others added 2 commits August 11, 2026 17:13

Remove finito_load_check.py script

6ee34b8

Merge remote-tracking branch 'origin/main' into bs/finito_link

5685ec2

github-actions Bot added data_changes postprocessing labels Aug 12, 2026

bsergi reviewed Aug 13, 2026

View reviewed changes

Merge remote-tracking branch 'origin/main' into bs/finito_link

73a23fa

wesleyjcole reviewed Aug 19, 2026

View reviewed changes

patrickbrown4 reviewed Aug 21, 2026

View reviewed changes

	* Restrict operational outputs to representative timeslices and seasons
	h(h)$[not h_rep(h)] = no ;

		* TODO: activate after FINITO pricing PR is merged
		* + smax{cfp$[gasp(cfp)$map_cf_fe(cfp,'NG')$valcft(cfp,t)], cf_markup(cfp,'NG','Electric_Power',cendiv,t) }

		shutil.copy(os.path.join(reeds_path,f), os.path.join(case,f))
		shutil.copy(f, Path(case, f))

		#TODO: why so much shifting? validate the dynamics here with FINITO team
		#TODO: also check against reference quantity

	shutil.copytree(os.path.join(caseSwitches['finito_dir'], 'inputs'),os.path.join(casedir,'finito', 'inputs'))
	shutil.copytree(Path(caseSwitches['finito_dir'], 'inputs'), Path(casedir, 'finito', 'inputs'))

	$include finito/model/linked_finito_temporal_params.gms
	$include finito%ds%model%ds%linked_finito_temporal_params.gms

	$include finito/model/finito_varfix.gms
	$include finito%ds%model%ds%finito_varfix.gms

		# Validation check: FINITO reference load should not exceed baseline load.
		# If it does, clip to zero and report

Uh oh!

Conversation

mmowers commented Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Technical details

Relevant sources or documentation

Validation, testing, and comparison report(s)

Checklist for author

Details to double-check

General information to guide review

Did you use LLM tools (chatbot or copilot) in the preparation of this PR? If so, describe how

Tag points of contact here if you would like additional review of the relevant parts of the model

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

wesleyjcole left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

patrickbrown4 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

mmowers commented Jun 23, 2026 •

edited

Loading