Add grid model: event-constrained power flow on IEEE 14-bus

[dovallev]

Summary

This PR adds a new GDP benchmark model for event-constrained optimal power flow on the IEEE 14-bus network.

Model description

The model minimizes the total capacity slack added to generators and transmission lines so that power balance constraints are satisfied across a set of Monte Carlo load scenarios (default: 500 samples drawn from a multivariate normal distribution).

Rather than requiring all equipment to stay within limits in every scenario (a joint chance constraint), an event constraint with ATLEAST logic requires that at least active_gens generators and active_lines transmission lines simultaneously satisfy their capacity bounds. This event must hold in at least 90% of scenarios (\alpha = 0.9), approximated via Sample Average Approximation.

The GDP formulation encodes whether each capacity constraint is satisfied or violated per scenario using disjunctions, and connects them to a per-scenario Boolean event variable via a logical equivalence (equivalent + land + atleast).

Reference

Ovalle, D., Mazzadi, S., Laird, C. D., Grossmann, I. E., & Pulsipher, J. L. (2025). Event constrained programming. Computers & Chemical Engineering, 199, 109145. https://doi.org/10.1016/j.compchemeng.2025.109145

Problem size (default: active_gens=4, active_lines=20, num_samples=500)

Problem	vars	Bool	bin	int	cont	cons	nl	disj	disjtn
grid	47525	35000	0	0	12525	52000	0	35000	12500

Changes

• gdplib/grid/grid.py — new model with build_model(active_gens=4, active_lines=20, num_samples=500)
• gdplib/grid/__init__.py — exposes build_model following library conventions
• gdplib/grid/README.md — model documentation with problem description, reference, and size table
• gdplib/__init__.py — registers gdplib.grid
• tests/test_comprehensive_coverage.py, tests/test_model_structure.py, tests/test_module_imports.py — added "grid" to the manual module lists

[bernalde]

Blocking issues:

1. build_model() resets NumPy's process-global RNG.
2. The new grid model only has generic smoke coverage; it needs grid-specific tests for the public behavior and transformation path.

Nonblocking issues:

1. gdplib/grid/grid.py does not pass the repo Black check.
2. The branch currently conflicts with main after #98 added multiperiod_blending; rebase/merge main and keep both modules in the import and test lists.
3. Consider tightening validation for integer parameters and num_samples > 0.

Questions:
None.

Tests run and outcomes:

• pytest tests/test_module_imports.py::TestModuleImports::test_build_model_callable[grid] tests/test_comprehensive_coverage.py::TestComprehensiveCoverage::test_model_construction_and_execution[grid] tests/test_model_structure.py::TestModelStructure::test_build_model_has_docstring[grid] -v --tb=short: passed, 3 tests.
• python -c "from gdplib.grid import build_model; from pyomo.environ import TransformationFactory; m=build_model(num_samples=3); TransformationFactory('core.logical_to_linear').apply_to(m); TransformationFactory('gdp.bigm').apply_to(m)": passed.
• pytest tests/ -v --tb=short: passed, 212 passed, 1 skipped, 4 warnings, under local Python 3.13.5.
• black gdplib/grid/grid.py --check --diff -S -C: failed; Black would reformat the new file.
• flake8 gdplib/grid/grid.py --count --select=E9,F63,F7,F82 --show-source --statistics: passed.
• typos --config ./.github/workflows/typos.toml gdplib/grid/README.md gdplib/grid/grid.py: not run because typos is not installed locally.
• Supported-version local pytest was attempted with Python 3.12, but that interpreter has neither pytest nor pyomo installed here. GitHub Actions reports test-python-3.9, 3.10, 3.11, 3.12, coverage-report, and test-pip-installation passing; lint/style-and-typos failing at Black Formatting Check.

This PR should not be merged as-is. I would not merge this until the blocking issues above are addressed.

[dovallev]

1. Global rng fixed.
2. Tests grid-specific tests were added using num_samples=10.
3. Black formatting applied to grid.py and all files from the merge.
4. sympy>=1.0 added to requirements.txt and pyproject.toml (Sergey's comment)
5. Parameter validation tightened (integers + num_samples > 0)
6. Upstream merged — grid and multiperiod_blending coexist in all import/test lists

[bernalde]

Blocking issues:

1. Runtime dependency metadata and the committed Pixi lock are inconsistent after adding sympy.
2. The new model is not registered with the generated model-size reporting workflow.

Nonblocking issues:

1. The PR introduces trailing whitespace/line-ending churn in gdplib/kaibel/main_gdp.py.
2. Focused flake8 on the new grid files reports avoidable lint debt.

Questions:
None.

Tests run and outcomes:

• pixi run --locked pytest tests/test_grid.py -v --tb=short: failed immediately with lock-file not up-to-date with the workspace.
• pixi run pytest tests/test_grid.py -v --tb=short: passed, 20 tests, after Pixi resolved the missing lock entries locally.
• pixi run test: passed, 313 passed, 1 skipped, after Pixi resolved the missing lock entries locally.
• pixi run lint: passed.
• git diff --check origin/main...HEAD: failed on trailing whitespace in gdplib/kaibel/main_gdp.py.
• pixi run flake8 gdplib/grid tests/test_grid.py --max-complexity=10 --max-line-length=88 --statistics: failed on unused imports, long lines, ambiguous l, and complexity.
• Small solve smoke test with build_model(num_samples=5), core.logical_to_linear, gdp.bigm, and GLPK: optimal, objective 0.0.

This PR should not be merged as-is. I would not merge this until the blocking issues above are addressed.

[dovallev]

Addressed the blocking and nonblocking items from the latest review:

• Added sympy>=1.0 to pixi.toml and setup.py; regenerated pixi.lock so pixi run --locked works correctly.
• Registered grid in generate_model_size_report.py, generated gdplib/grid/model_size_report.md, and added the grid column to the root README.md size table.
• Removed unused imports (Constraint, Disjunction in grid.py; Disjunct in test_grid.py); renamed ambiguous loop variable l to ln.

[bernalde]

Pushed f490890 (Address PR 101 review cleanup) to dovallev:add-grid-model.

Main changes made:

• Applied Black formatting to the PR-touched files that were causing the current lint CI failure.
• Removed the remaining Kaibel whitespace/line-ending artifacts so the PR diff is clean under git diff --check.
• Wrapped the remaining long grid docstring lines and added a targeted # noqa: C901 for the expected large declarative build_model() body; the focused grid flake8 check now passes.
• Verified the force-pushed branch already addressed the prior functional review items: local NumPy RNG use, grid-specific tests, parameter validation, SymPy dependency alignment, Pixi lock update, and generated model-size report integration.

Tests and checks run:

• pixi run --locked pytest tests/test_grid.py -v --tb=short: 20 passed.
• pixi run --locked flake8 gdplib/grid tests/test_grid.py --max-complexity=10 --max-line-length=88 --statistics: passed.
• pixi run --locked lint: passed.
• pixi run --locked test: 317 passed, 1 skipped.
• git diff --check origin/main...HEAD: passed.

Comments intentionally not addressed:

• None. The remaining unresolved review threads appear addressed by the force-pushed branch plus f490890.
• I did not mark threads resolved, per request.

Remaining risks or follow-up:

• GitHub Actions reran on f490890 and all checks passed.
• No solver-backed default optimality run was added; coverage remains construction, validation, and GDP transformation focused, keeping optional solvers out of default tests.

[bernalde]

Fresh maintainer review at f490890.

I do not see remaining blocking or nonblocking issues in the current diff. The previous review comments appear adequately addressed: RNG isolation, grid-specific tests, parameter validation, dependency and Pixi lock alignment, generated model-size reporting, whitespace cleanup, and focused grid lint cleanup are all in place.

Verification:

• Inspected the current PR diff, commits since the prior reviews, generated docs, dependency manifests, tests, and surrounding registration code.
• GitHub Actions on f490890 are all passing.
• git diff --check origin/main...HEAD: passed.
• pixi run --locked pytest tests/test_grid.py tests/test_generate_model_size_report.py -v --tb=short: 26 passed.
• pixi run --locked flake8 gdplib/grid tests/test_grid.py --max-complexity=10 --max-line-length=88 --statistics: passed.
• Focused model-size check for gdplib.grid.build_model() matched the committed report values.
• Small transformed GLPK smoke test with num_samples=5: optimal, objective 0.0.

Residual risk: no solver-backed default optimality evidence for the full benchmark instance was added, which is appropriate for this repository because optional solvers should stay out of default CI.