Format Conversion with Arrow Intermediate

gat convert format lets you bridge MATPOWER, PSS/E, CIM, pandapower, and PowerModels.jl case files by writing a temporary Arrow dataset and then running the appropriate exporter. The CLI auto-detects the source when you omit --from, and the Arrow intermediate keeps downstream tooling stable even when you move between formats.

Supported Formats

GAT supports bidirectional conversion between six power system data formats:

Usage

gat convert format \
  --input case14.m \
  --to matpower \
  --output case14_roundtrip.m

When --from is omitted the command auto-detects the input format; provide it explicitly when an extension is ambiguous (e.g., .json could be pandapower or PowerModels.jl). --force overwrites existing directories or files.

Target Formats

• --to arrow: keep the temporary Arrow dataset for DuckDB, Polars, or downstream GAT solvers.
• --to matpower: export a .m file with bus/gen/branch/gencost data for MATPOWER tools.
• --to psse: emit a PSS/E RAW file with bus, load, generator, and branch sections.
• --to cim: write a minimal CIM RDF/XML graph for round-tripping through the CIM importer.
• --to pandapower: serialize bus, load, gen, line, and trafo tables into pandapower JSON.
• --to powermodels: export PowerModels.jl JSON format for Julia optimization tools.

Examples

Basic Conversion

# MATPOWER to Arrow
gat convert format --input case14.m --to arrow -o case14.arrow

# Arrow to MATPOWER
gat convert format --input case14.arrow --to matpower -o case14.m

# PSS/E to pandapower
gat convert format --input ieee118.raw --to pandapower -o ieee118.json

PowerModels.jl Workflows

# Export to PowerModels.jl for Julia optimization
gat convert format --input case30.m --to powermodels -o case30_pm.json

# Import from PowerModels.jl
gat convert format --input network.json --from powermodels --to arrow -o network.arrow

# Convert PGLib benchmarks to MATPOWER
gat convert format --input pglib_opf_case14.json --from powermodels --to matpower -o case14.m

Industrial Format Interchange

# CIM to Arrow for analysis
gat convert format --input grid_model.rdf --to arrow -o grid.arrow

# Arrow to CIM for SCADA systems
gat convert format --input grid.arrow --to cim -o grid_export.rdf

Data Preservation

Each exporter reloads the Arrow dataset (load_grid_from_arrow) before writing the final target, ensuring conversions stay deterministic even when chained.

Preserved Fields

• Bus: ID, name, voltage (kV, p.u.), vmin/vmax limits, area/zone
• Generator: Bus, P/Q output, min/max limits, cost model
• Load: Bus, active/reactive power demand
• Branch: From/to bus, R/X, charging B, tap ratio, ratings, status
• Cost models: Polynomial (c0 + c1*P + c2*P^2) and piecewise linear

Cost Model Handling

Generator cost models are preserved across all formats:

• Polynomial: Coefficient arrays (GAT stores [c0, c1, c2])
• Piecewise linear: (MW, cost) breakpoint pairs

Note: PowerModels.jl uses reversed coefficient order [c2, c1, c0]. GAT handles this automatically.

Format-Specific Notes

PowerModels.jl

• Dictionary-based JSON with baseMVA at root level
• Component indices as string keys: "bus": {"1": {...}, "2": {...}}
• Compatible with PGLib-OPF benchmarks
• Auto-detected by checking for baseMVA + bus without pandapower markers

pandapower vs PowerModels.jl

Both use .json extension. GAT disambiguates by content:

• pandapower: Contains pandapowerNet or _module keys
• PowerModels.jl: Contains baseMVA at root without pandapower markers

Use --from pandapower or --from powermodels to force the parser.

Metadata

Export metadata from the Arrow manifest (source file, timestamp, and GAT version) is written into the MATPOWER, PSS/E, CIM, pandapower, and PowerModels outputs for provenance tracking.

See Also

• Arrow Schema - Table-by-table Arrow dataset documentation
• pandapower Schema - Column mappings for pandapower
• OPF Guide - Running optimal power flow after conversion