Flexibility Analysis #1398

michaelbynum · 2024-04-23T15:10:46Z

Summary/Motivation:

This PR adds a module for formulating flexibility analysis problems.

Legal Acknowledgement

By contributing to this software project, I agree to the following terms and conditions for my contribution:

I agree my contributions are submitted under the license terms described in the LICENSE.txt file at the top level of this directory.
I represent I am authorized to make the contributions and grant the license. If my employer has rights to intellectual property that includes these contributions, I represent that I have received permission to make contributions and grant the required license on behalf of that employer.

…tion

michaelbynum · 2024-04-24T20:34:33Z

Thanks!

andrewlee94 · 2024-04-25T18:25:07Z

requirements-dev.txt

@@ -29,6 +30,7 @@ jsonschema
 jupyter_contrib_nbextensions
 snowballstemmer==1.2.1
 addheader>=0.2.2
+highspy


Would SCIP be sufficient for these tests? We already have SCIP available (conditionally) for use in the Diagnostics tests, and we do try to keep our dependencies to a minimum.

SCIP would work, but it makes the tests way slower. I'm solving a bunch of LPs with changes only to parameters. The Pyomo interface to HiGHS is a persistent interface, and is much faster than the interface to SCIP for that type of workflow.

We might want to see if it makes sense to do something like @lbianchi-lbl did for SCIP then so that it does not need to be a dependency in requirements-dev.txt. If I recall, we basically have a check that sees if SCIP is available and otherwise skips the tests. We then just make sure we have it installed on the test server (via GHA setup) so the test will run on the server.

Sounds good to me. As long as it gets installed in GHA so the tests run, I'm happy.

Not sure if it applies, but would the other way around work? i.e. is it possible to use HiGHS instead of SCIP for our existing tests for which SCIP is currently being used?

If so, it might make sense to consider having highspy as a dev dependency and just use it for testing for both cases (of course, this is purely on the side of streamlining dependencies and very much solver-domain-ignorant, so feel free to add why this wouldn't work).

I actually need both. SCIP also solves nonlinear problems, while HiGHS does not...

Ah, that makes sense then, thanks for clarifying!

lbianchi-lbl · 2024-04-25T18:25:23Z

idaes/apps/flexibility_analysis/_check_relu_dependencies.py

+omlt, nump_available = attempt_import("omlt")
+if not tensorflow_available or not nump_available:


Suggested change

omlt, nump_available = attempt_import("omlt")

if not tensorflow_available or not nump_available:

omlt, omlt_available = attempt_import("omlt")

if not tensorflow_available or not omlt_available:

Good catch! Thanks!

andrewlee94

Here are some comments on the first file (I'll try to get through the rest tomorrow).

One general comment; could you add yourself to the top-level CODEOWNERS file as the maintainer for this code (or a nominee).

andrewlee94 · 2024-04-25T19:15:36Z

idaes/apps/flexibility_analysis/_check_dependencies.py

+"""
+This module is only used to check dependencies for unit tests.
+"""
+import unittest


Seeing as this is just for testing, should it be moved into the tests folder?

Great point. I'll move this.

andrewlee94 · 2024-04-25T19:16:10Z

idaes/apps/flexibility_analysis/_check_relu_dependencies.py

+"""
+This module is only used to check dependencies for unit tests.
+"""
+import unittest


andrewlee94 · 2024-04-25T19:17:42Z

idaes/apps/flexibility_analysis/check_optimal.py

+works for both APPSI solver interfaces and non-appsi solver interfaces.
+"""
+import pyomo.environ as pe
+from pyomo.contrib import appsi


A minor comment (so does not need to be fixed), but this seems like it would be better located in Pyomo (probably in pyomo.contrib with the appsi code seeing as it related to that).

Great point.

andrewlee94 · 2024-04-25T19:20:50Z

idaes/apps/flexibility_analysis/decision_rules/dr_config.py

+        visibility=0,
+    ):
+        super().__init__(
+            description=description,


Is the purpose of this just to change the default values for the implicit and implicit_domain arguments (whilst still leaving them as arguments)?

I don't think I changed any defaults. This class is just meant to be a base class for other, related config classes. I guess it is not really necessary. I think I just created it for type hinting.

andrewlee94 · 2024-04-25T19:23:12Z

idaes/apps/flexibility_analysis/decision_rules/linear_dr.py

+            visibility=visibility,
+        )
+        self.solver = self.declare(
+            "solver", ConfigValue(default=pe.SolverFactory("ipopt"))


Should this also have a doc attribute, just in case someone tries to query the ConfigDict?

Absolutely. I'll add one.

andrewlee94 · 2024-04-25T19:33:47Z

idaes/apps/flexibility_analysis/decision_rules/relu_dr_config.py

+            implicit_domain=implicit_domain,
+            visibility=visibility,
+        )
+        self.n_layers: int = self.declare(


These might be obvious to people with more experience with neural networks, but it might be good to have some short doc strings for these arguments for less experiences users.

andrewlee94 · 2024-04-25T19:36:01Z

idaes/apps/flexibility_analysis/decision_rules/relu_dr.py

+
+    config: ReluDRConfig = config()
+    if config.batch_size > n_samples:
+        config.batch_size = n_samples


It might be good to log a message for the user here to let them know they set a batch size greater than the number of samples and thus this value was changed.

andrewlee94 · 2024-04-25T19:38:50Z

idaes/apps/flexibility_analysis/decision_rules/relu_dr.py

+    for _ in range(config.n_layers - 1):
+        nn.add(keras.layers.Dense(config.n_nodes_per_layer, activation="relu"))
+    nn.add(keras.layers.Dense(len(outputs)))
+    if config.learning_rate is None:


I assume this is needed as keras does not accept None as a value for learning_rate (i.e. if you provide the argument it must have a meaningful value). An inline comment might help to make sure no-one tries to "fix" this later.

andrewlee94 · 2024-04-25T19:39:00Z

idaes/apps/flexibility_analysis/decision_rules/relu_dr.py

+        training_output,
+        batch_size=config.batch_size,
+        epochs=config.epochs,
+        # verbose=0,


Commented line.

andrewlee94 · 2024-04-25T19:40:55Z

idaes/apps/flexibility_analysis/decision_rules/tests/test_linear_dr.py

+# All rights reserved.  Please see the files COPYRIGHT.md and LICENSE.md
+# for full copyright and license information.
+#################################################################################
+from idaes.apps.flexibility_analysis import _check_dependencies


Should there be tests for the relu_dr as well?

ksbeattie · 2024-05-09T18:39:29Z

@michaelbynum any update here?

andrewlee94 · 2024-05-17T14:35:05Z

@michaelbynum Something you might want to think about for the docs here: I have received a number of questions this week about this PR and how it is distinguished from robust optimization and Pyros (and thus when you would want to use one versus the other). It would probably be good to have some discussion of this somewhere in the docs to guide new users to the best tool for what they want to do. On the plus side, there is already some interest in tools like this.

ksbeattie · 2024-05-23T18:37:08Z

@michaelbynum I'm moving this to the Aug release as the May release will be created next week.

ksbeattie · 2024-06-06T18:50:06Z

@michaelbynum any update on this?

michaelbynum · 2024-06-10T11:42:38Z

Apologies. I'll be pushing updates to address reviews early next week.

ksbeattie · 2024-08-01T18:26:44Z

Apologies. I'll be pushing updates to address reviews early next week.

@michaelbynum I know you're busy... any news on this?

ksbeattie · 2024-08-08T18:38:24Z

@michaelbynum, I'm taking this off the Aug release board as it doesn't appear work is progressing here.

michaelbynum and others added 30 commits March 2, 2022 08:28

working on a module for flexibility analysis

84c832f

reformatting files in flexibility analysis

63d2e82

flexibility: updating imports

051db2b

flexibility analysis cleanup

d27ecde

flexibility analysis cleanup

73c4326

flexibility analysis cleanup

3922932

flexibility analysis: working on tests

dc77dd6

flexibility analysis: updating imports

dae913f

Merge branch 'main' into flex

3ccd861

flexibility analysis: minor updates

94cee4a

flexibility analysis: better initialization for sampling

95b354f

flexibility analysis: better initialization for sampling

f8a1c03

flexibility analysis: better ordering of sampled points

de53a44

flexibility analysis: cleaning up sampling code

d778ceb

flexibility analysis: adding config options for active constraint method

fb236f2

flexibility analysis: minor updates

2312395

flexibility analysis: option for total violation instead of max viola…

d81c314

…tion

Merge branch 'main' into flex

4c63cbb

update tests

53fc67b

update tests

5fc5458

update flexibility example

d5bf854

update example

66b66ce

Merge branch 'main' into flex

4b17072

working on docs for flexibility analysis

91b8051

flexibility analysis docs

5970504

flexibility analysis docs

accd15d

Merge remote-tracking branch 'main_fork/main' into flex

f675b34

working on docs

4c5a410

Merge remote-tracking branch 'main_fork/main' into flex

78dd253

merge main into flex

751e788

debugging ci

fce9af4

michaelbynum added 8 commits April 24, 2024 14:35

revert some changes

79d4b6a

speed up flexibility tests

7f51740

fix tests

6004260

debugging ci

8cdbde9

fix tests

4a78054

revert some changes

013fa66

fix docs

6743fb1

more flexibility tests

7cb6ee8

ksbeattie requested review from jsiirola, andrewlee94 and radhakrishnatg and removed request for dangunter, lbianchi-lbl, bpaul4 and ksbeattie April 25, 2024 18:21

ksbeattie assigned michaelbynum Apr 25, 2024

ksbeattie requested a review from lbianchi-lbl April 25, 2024 18:22

andrewlee94 reviewed Apr 25, 2024

View reviewed changes

lbianchi-lbl reviewed Apr 25, 2024

View reviewed changes

ksbeattie added the Priority:Normal Normal Priority Issue or PR label Apr 25, 2024

andrewlee94 reviewed Apr 25, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Flexibility Analysis #1398

Flexibility Analysis #1398

michaelbynum commented Apr 23, 2024 •

edited

Loading

michaelbynum commented Apr 24, 2024

andrewlee94 Apr 25, 2024

michaelbynum Apr 26, 2024

andrewlee94 Apr 26, 2024

michaelbynum Apr 26, 2024

lbianchi-lbl Apr 26, 2024

michaelbynum Apr 26, 2024

lbianchi-lbl Apr 26, 2024

lbianchi-lbl Apr 25, 2024

michaelbynum Apr 26, 2024

andrewlee94 left a comment

andrewlee94 Apr 25, 2024

michaelbynum Apr 26, 2024

andrewlee94 Apr 25, 2024

andrewlee94 Apr 25, 2024

michaelbynum Apr 26, 2024

andrewlee94 Apr 25, 2024

michaelbynum Apr 26, 2024

andrewlee94 Apr 25, 2024

michaelbynum Apr 26, 2024

andrewlee94 Apr 25, 2024

andrewlee94 Apr 25, 2024

andrewlee94 Apr 25, 2024

andrewlee94 Apr 25, 2024

andrewlee94 Apr 25, 2024

ksbeattie commented May 9, 2024

andrewlee94 commented May 17, 2024

ksbeattie commented May 23, 2024

ksbeattie commented Jun 6, 2024

michaelbynum commented Jun 10, 2024

ksbeattie commented Aug 1, 2024

ksbeattie commented Aug 8, 2024

		omlt, nump_available = attempt_import("omlt")
		if not tensorflow_available or not nump_available:

Flexibility Analysis #1398

Are you sure you want to change the base?

Flexibility Analysis #1398

Conversation

michaelbynum commented Apr 23, 2024 • edited Loading

Summary/Motivation:

Legal Acknowledgement

michaelbynum commented Apr 24, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andrewlee94 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ksbeattie commented May 9, 2024

andrewlee94 commented May 17, 2024

ksbeattie commented May 23, 2024

ksbeattie commented Jun 6, 2024

michaelbynum commented Jun 10, 2024

ksbeattie commented Aug 1, 2024

ksbeattie commented Aug 8, 2024

michaelbynum commented Apr 23, 2024 •

edited

Loading