Rolling horizon dsm by Manish-Khanra · Pull Request #804 · assume-framework/assume

Manish-Khanra · 2026-04-30T12:49:45Z

Related Issue

Closes #804
If no issue exists, delete this line.

Description

This PR implements rolling-horizon optimization for the DSM units, enabling reactive bidding while maintaining inter-temporal feasibility. Previously, DSM units (steel plants, buildings, hydrogen plants, steam generation plants) optimized over the full simulation horizon. This feature allows units to re-optimize shorter look-ahead windows after each market round while carrying forward component states (e.g., storage SoC, operational status).

Checklist

Documentation updated (docstrings, READMEs, user guides, inline comments, doc folder updates etc.)
New unit/integration tests added (if applicable)
Changes noted in release notes (if any)
Consent to release this PR's code under the GNU Affero General Public License v3.0

Additional Notes (optional)

Key Features

Rolling-Horizon Mode: Configure via dsm_optimisation_config with horizon_mode, look_ahead_horizon, commit_horizon, and rolling_step keys
Three Steel Plant Strategies: cost-optimized (default), profile-guided (tracks normalized load profile), and min-demand (enforces hourly minimums)
Fully Extensible: Removed hardcoded technology-specific logic; new DSM unit types can enable rolling-horizon by setting 3 class attributes and optionally 1 method override
Production-Ready: All 53 tests passing; backward compatible; comprehensive documentation

…o rolling_horizon_dsm

…o rolling_horizon_dsm Co-authored-by: Copilot <copilot@github.com>

codecov · 2026-04-30T12:54:10Z

Codecov Report

❌ Patch coverage is 76.00000% with 108 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.09%. Comparing base (6f2857a) to head (ae32acd).

Files with missing lines	Patch %	Lines
assume/units/dsm_load_shift.py	74.26%	87 Missing ⚠️
assume/units/steel_plant.py	82.00%	9 Missing ⚠️
assume/strategies/naive_strategies.py	58.33%	5 Missing ⚠️
assume/scenario/loader_csv.py	82.60%	4 Missing ⚠️
assume/common/forecaster.py	88.88%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #804      +/-   ##
==========================================
+ Coverage   82.09%   83.09%   +1.00%     
==========================================
  Files          56       56              
  Lines        9081     9518     +437     
==========================================
+ Hits         7455     7909     +454     
+ Misses       1626     1609      -17

Flag	Coverage Δ
pytest	`83.09% <76.00%> (+1.00%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

maurerle

I did not review the dsm_load_shift.py - but all other files.
There are still a few things to do.

I guess that the way in which units are added to the DSMFlex is not that good, as the code seems to not know which parameters of the class exist.
The result is the need for various hasattr checks, as you never know if a param exists or is initialized.
I suggest using planning mode and a larger context window when using an LLM to help with this..

maurerle · 2026-04-30T15:28:39Z


    # Set the congestion_indicator in the instance
+    # Delete the old component if it exists to avoid Pyomo warnings
+    if hasattr(instance, "congestion_indicator"):


I don't like this check - can't we test for something like instance.flexibility_measure == "congestion_management_flexibility" in which case the congestion_indicator is set?
This is more like a general thing

Co-authored-by: Copilot <copilot@github.com>

maurerle · 2026-05-13T08:30:54Z

@Manish-Khanra please rebase/merge and fix the conflicts - then this can be merged :)

…o rolling_horizon_dsm

Manish-Khanra · 2026-05-18T08:14:44Z

@Manish-Khanra please rebase/merge and fix the conflicts - then this can be merged :)
@maurerle Has been successfully rebased

maurerle · 2026-05-18T08:36:27Z

        demand_value = self.steel_demand
-        if "_remaining_demand" in self.components:
-            demand_value = self.components["_remaining_demand"]
+        if hasattr(self, "_rh_window_remaining_demand"):


why not initialize _rh_window_remaining_demand in init instead of using hasattr here?

maurerle

Hi @Manish-Khanra I reviewed this PR with @reinecfi
we have some recommendations and changes requested.

If something is not easily understood, we can have a call as well.

maurerle · 2026-05-21T07:28:35Z

+        return "price_from_cleared_history"
+
+    # Otherwise treat as algorithm name and return as-is
+    return value


Shouldn't this happen automagically?
Via forecaster _registries:

assume/assume/common/forecaster.py

Line 657 in 7575332

congestion_signal_update_algorithm = self._registries["update"].get(

I guess we currently can only decide if we want to use a self calculated forecast or the existing from forecasts_df.csv
This is still the case with this function (as forecasts_df is always preferred).

I wonder if this function is actually needed. @reinecfi

maurerle · 2026-05-21T07:39:59Z

+                        market_prices=initial_market_prices
+                        if initial_market_prices
+                        else unit.get("market_prices"),


Suggested change

market_prices=initial_market_prices

if initial_market_prices

else unit.get("market_prices"),

market_prices=initial_market_prices or unit.get("market_prices"),

what is market_prices here?

maurerle · 2026-05-21T07:40:59Z

+
+                    if isinstance(price_update_resolved, pd.Series):
+                        # Direct column: use as initial market prices
+                        initial_market_prices["EOM"] = price_update_resolved


hardcoding "EOM" here does not look right?

But I guess this is consistently weird with the DSM unit and DSM forecaster as well.

maurerle · 2026-05-21T07:54:21Z

-            unit.optimisation_counter = 1
+        if unit.horizon_mode == "rolling_horizon":
+            current_market_time = product_tuples[0][0]
+            did_reoptimize = unit._check_and_reoptimize_rolling_window(


This should be calling update_forecasts_if_needed instead.
This logic should then go into this update_forecasts_if_needed function.

But I think that we need to specify this further together with @reinecfi

…o rolling_horizon_dsm

This is still very different from other units and includes many workarounds.

maurerle · 2026-05-29T12:51:14Z

+    Returns:
+        dict: Updated forecast with adaptive prices learned from clearing history.
+    """
+    unit = kwargs.get("unit")


where is this unit set when running this in the simulation (not in the test)

maurerle

Did you test that everything works well with your latest commit as intended?
I think that some relevant parts were removed there..? For example adding an "adaptive" forecast algorithm function.

The general idea is to keep the rolling horizon out of the loader_csv and implement the update function in the forecast_algorithm instead of the loader_csv so that it can be reused?

But maybe I am mistaken and this is intended?

…o rolling_horizon_dsm

Manish-Khanra · 2026-06-03T20:16:57Z

Did you test that everything works well with your latest commit as intended? I think that some relevant parts were removed there..? For example adding an "adaptive" forecast algorithm function.

The general idea is to keep the rolling horizon out of the loader_csv and implement the update function in the forecast_algorithm instead of the loader_csv so that it can be reused?

But maybe I am mistaken and this is intended?

@maurerle Yes removing the adaptive forecast algorithm was intentional. And you're right that it's a relevant piece, which is exactly why I am moving it to its own PR rather than dropping it. I originally added the adaptive update algorithm in the same commit as rolling horizon, but they are really two separate features. The adaptive algorithm is a forecast update mechanism that refreshes the price forecast for future use. Since rolling horizon doesnot depend on it, I have stashed the forecast algorithm to keep this PR focused on the rolling-horizon implementation and I will submit it as a dedicated PR.

…Python versions other than 3.14

Manish-Khanra · 2026-06-03T22:24:07Z

@maurerle The 3.14 checks fail because pypsa <= 0.35.2 (pinned in #808, see #807) becasue the network extra cant install on 3.14. Until the pin can be dropped, I have guarded the notebook 10 with pypsa + network test steps with if: matrix.python-version != '3.14'; 3.10/3.12 are unchanged. This notebook has its own fix now bu I dont know if this temporarily fix like this is appropriate. If not, could you please handle this part?

…ents

Manish-Khanra and others added 4 commits April 7, 2026 16:31

rolling horision fucntion and forecast first version

87ba266

Merge branch 'main' of https://github.com/assume-framework/assume int…

ffa070b

…o rolling_horizon_dsm

test and documentation

e6043ab

Merge branch 'main' of https://github.com/assume-framework/assume int…

b7ca3eb

…o rolling_horizon_dsm Co-authored-by: Copilot <copilot@github.com>

maurerle self-requested a review April 30, 2026 13:09

maurerle reviewed Apr 30, 2026

View reviewed changes

changes as per reviewer's remarks

6e63cb5

Co-authored-by: Copilot <copilot@github.com>

Manish-Khanra requested a review from maurerle April 30, 2026 20:26

Manish-Khanra added 2 commits May 18, 2026 09:56

Merge branch 'main' of https://github.com/assume-framework/assume int…

d1f3732

…o rolling_horizon_dsm

Unknown DSM component technology fix

6310d09

Manish-Khanra closed this May 18, 2026

Manish-Khanra reopened this May 18, 2026

maurerle reviewed May 18, 2026

View reviewed changes

replaced hasattr with is not None

7575332

maurerle requested changes May 21, 2026

View reviewed changes

Manish-Khanra and others added 3 commits May 22, 2026 17:23

Merge branch 'main' of https://github.com/assume-framework/assume int…

b95a4fc

…o rolling_horizon_dsm

Revision 2.0

fa4365c

loader_csv: refactor steelplant registration

52da6f6

This is still very different from other units and includes many workarounds.

Manish-Khanra requested a review from maurerle May 23, 2026 12:23

maurerle reviewed May 29, 2026

View reviewed changes

maurerle and others added 2 commits May 29, 2026 15:40

update changes from todays discussion

ec47e0a

removed all reduncancies

809bb16

Manish-Khanra requested a review from maurerle June 1, 2026 19:08

maurerle requested changes Jun 2, 2026

View reviewed changes

Comment thread assume/scenario/loader_csv.py Outdated

Merge branch 'main' of https://github.com/assume-framework/assume int…

3146b92

…o rolling_horizon_dsm

Remove dead adaptive-price logic from DSM loader

4b54d25

Add network notebook execution and pypsa dependency installation for …

6ed0de6

…Python versions other than 3.14

Manish-Khanra requested a review from maurerle June 3, 2026 22:24

Add conditional pytest tests for Python 3.14 to skip network requirem…

ae32acd

…ents

Conversation

Manish-Khanra commented Apr 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Related Issue

Description

Checklist

Additional Notes (optional)

Key Features

Uh oh!

codecov Bot commented Apr 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

maurerle left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

maurerle Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

maurerle commented May 13, 2026

Uh oh!

Manish-Khanra commented May 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

maurerle May 18, 2026

Choose a reason for hiding this comment

Uh oh!

maurerle left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

maurerle May 21, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

maurerle May 21, 2026

Choose a reason for hiding this comment

Uh oh!

maurerle May 21, 2026

Choose a reason for hiding this comment

Uh oh!

maurerle May 21, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

maurerle May 29, 2026

Choose a reason for hiding this comment

Uh oh!

maurerle left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Manish-Khanra commented Jun 3, 2026

Uh oh!

Manish-Khanra commented Jun 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Manish-Khanra commented Apr 30, 2026 •

edited

Loading

codecov Bot commented Apr 30, 2026 •

edited

Loading

Manish-Khanra commented May 18, 2026 •

edited

Loading