PT60

Public observations → reusable AC power-flow cases → a verifiable dataset and toolchain.

PT60 connects Portuguese public network, asset and operating records into reproducible AC power-flow cases for the 60, 130, 150, 220 and 400 kV network. The retained network contains 3,783 buses, 4,943 lines and 228 transformers. Two seasonal weeks provide 336 hourly samples of 15-minute operating intervals, with archived inputs, primary results and spatial-allocation comparisons.

Four deliverables

The dataset is a release candidate; permanent repository deposit and DOI are pending. The paper is a manuscript, not a published article. PyPI publication is pending account credentials; the installable wheel below is available now.

Install and load

Use Python 3.13 in a virtual environment. Until PyPI publication:

python -m pip install https://grid.jczw.xyz/downloads/pt60_tools-0.4.0-py3-none-any.whl
pt60 fetch https://grid.jczw.xyz/download --output work/PT60-v2.1.0-rc2 --sha256 07727842080141300c4ddf180f82a4e7f9f66e3a444a30297d650f79a3486913
export PT60_DATA="$PWD/work/PT60-v2.1.0-rc2"
pt60 verify
pt60 info
pt60 view

The browser opens at http://127.0.0.1:8050. The local viewer works without external map services. The hosted map uses the same dataset and stable IDs.

from pt60 import Dataset

data = Dataset("work/PT60-v2.1.0-rc2")
case = data.cases("SUMMER")[0]
inputs = data.load_input(case["case_id"])
results = data.results(case["case_id"])

Build and reproduce cases

Install solver dependencies only when calculating new results:

python -m pip install "pt60-tools[solve] @ https://grid.jczw.xyz/downloads/pt60_tools-0.4.0-py3-none-any.whl"
pt60 init --output work/my-input
# Edit the generated input files.
pt60 solve --input work/my-input --output work/my-result
pt60 replay --case PT60_2025_SUMMER_WEEK_JUL07_13_H000 --output work/replayed-case
pt60 view --result-dir work/my-result

pt60 batch --full --spatial --output work/reproduction reproduces the full seasonal experiment and its controls. Outputs must be outside the frozen release. See the usage guide for fields and API. The core installation requires NumPy, without PyTorch or GridSFM.

Development and provenance

pt60/ is the installable package; portuguese_hv_network/src/ builds the underlying network; portuguese_hv_network/site/ serves the website; paper/ contains the current manuscript and figure sources. Install from this checkout with python -m pip install -e '.[solve,test]' and run python -m pytest tests/test_pt60_tools.py.

The acquisition pipeline compiles the static network from source records. The distributed toolchain starts from that compiled network and archived observations to reproduce cases. Source roles, inferred quantities and model assumptions are retained in the dataset. Convergence demonstrates numerical consistency; it does not establish agreement with an operator state estimate.

Historical releases remain under data/releases/ for provenance. They are not alternative current downloads. Exploratory OPF and neural work is retained in experiments/opf_neural/ and output/opf/, outside the core package and paper. See the archive index for older material.

License

Source code is MIT licensed. Data retain source-specific terms, including E-REDES CC BY 4.0 and OpenStreetMap ODbL obligations. See DATA_LICENSE.md and the dataset's attribution files.