pyDAC 3.0.0

pyDAC 3.0.0

First public release of pyDAC, a simulator of adsorption-based direct air capture (DAC).
Version 3 is the first one published; earlier versions were internal.

What pyDAC does

• Solves the coupled mass, energy and momentum balances of a packed sorbent column in 1D or
  2D (axisymmetric), with CVODE (SUNDIALS) and a finite-volume discretization (pymrm).
• Runs adsorption cycles composed from configurable steps (adsorption, blowdown, heating,
  purge, steam- or wall-heated desorption, cooling, pressurization), by default a
  steam-assisted temperature-vacuum swing adsorption (TVSA) cycle: step by step, or repeated
  until cyclic steady state (CSS).
• Reports process key performance indicators: CO2 purity and recovery, productivity,
  equivalent work and its breakdown, and the O2 and hot-bed check.
• Optimizes operating conditions with NSGA-II (pymoo), with checkpoints and exact
  continuation of an interrupted campaign.

How you use it

• One command, pydac, with steps, css, nsga2, validate and get-started, and the same
  functions from Python (pydac.load_case, run_steps, run_css, run_nsga2, resume,
  report).
• A case is a JSON document. Start from the built-in defaults and change settings with
  --set, or save a complete case (--export-case) and run it exactly as saved with
  --case. A JSON Schema is included for editor validation and completion.
• Every run folder is named by the fingerprint of the complete case and saves the case it
  ran (resolved_config.json) and a provenance record: package version, Python and the
  numerically relevant dependency versions. Runs are reproducible without Git.
• pydac get-started FOLDER copies a starter kit into a new folder: README, licence, citation, the
  user guide and seven tutorials (cases, notebooks with their saved outputs, and the
  validation data).
• Saved runs can be continued (--resume) or reported again (--report). A continuation
  needs the pyDAC version that wrote the run; reading, replaying and reporting it work with
  any later 3.x release.

Requirements and installation

• Python 3.11 to 3.14 on Linux (x86_64). Windows is being tested: the full test suite passes
  on Windows with Python 3.14 in CI, but it is not supported yet. macOS is neither tested nor
  supported.
• CI runs the full test suite on Linux with Python 3.11 (lowest allowed dependency
  versions), 3.12 (the exact lock file, and the base dependencies only), 3.13 and 3.14
  (newest dependency versions), and on Windows with Python 3.14.
• Install pyDAC in a dedicated environment with python -m pip install pydac3. It pins some
  dependencies exactly (pymrm==2.1.10, scikit-sundae 1.1.x, pandas<3); a shared
  environment can break. The distribution is pydac3; the import and the command are
  pydac. pip install pydac installs an unrelated project.
• To change pyDAC itself, install it editable from a clone: see CONTRIBUTING.md.

Known limitations

• pymrm 2.1.10 declares documentation and test tools (Sphinx, pytest, numba, ...) as
  runtime dependencies, so the installation is larger than pyDAC needs.
• A few tests need the repository files next to tests/; run the test suite from a clone,
  not from the installed package.
• The H2O GAB isotherm is fitted up to 90 % relative humidity. Above it pyDAC extrapolates
  and prints a warning; the default cycle reaches it briefly during pressurization.
• The 2D model is the costly one: start with the 1D examples of the guide.

Documentation

• Start with the tutorials: examples/notebooks/, from 01_first_steps.ipynb in order
  (pydac get-started copies them).
• README.md: installation and a first run.
• docs/guide/: the user guide (installation, quick start, cases, running, optimization,
  results, reproducibility, extending).
• docs/manual/: the reference manual (framework map, inputs, core, discretization,
  equations, cycle steps, post-processing, NSGA-II).
• CONTRIBUTING.md: working in a clone, tests, extending the model.

Citation and license

MIT license. Please cite pyDAC with the metadata in CITATION.cff. Zenodo archives this
release with its own DOI; the DOI enters CITATION.cff and the README in 3.0.1.