Search run: 2026-08-15 formulation and graph-model landscape

Page status: generated search-run record.

Protocol: 0.2.0 Run date: 2026-08-15 Purpose: expand the scoping-review seed matrix from network-reduction papers to graph, port, and circuit-formulation precedents. This is still a seed run, not an exhaustive systematic review or a double-coded release.

Queries

The following portable queries were executed as web searches and then checked against the primary record or publisher/official landing page where available:

  1. "power network" graph representation port factor hypergraph multigraph
  2. "modified nodal" network analysis voltage source branch current
  3. "sparse tableau" power flow node breaker optimal power flow
  4. "port-Hamiltonian" graph interconnection electrical network
  5. typed graph transformation" electrical network rewrite provenance

The search strings are preserved in review/search-strings.md; database-specific translations, export files, deduplication, and citation chasing remain future work.

Records added

The run added six formulation/framework rows to review/evidence-matrix.csv:

  • EV-0005 — Ho, Ruehli, and Brennan, modified nodal analysis;
  • EV-0006 — Hachtel, Brayton, and Gustavson, sparse tableau;
  • EV-0007 — Park, Holzer, and DeMarco, sparse-tableau node–breaker OPF;
  • EV-0008 — Baez and Fong, compositional passive linear networks;
  • EV-0009 — van der Schaft and Maschke, port-Hamiltonian systems on graphs;
  • EV-0010 — Ehrig et al., typed algebraic graph transformation.

All rows are include and single_coded. They are coded as scoped precedents, not as evidence that one formulation dominates the others or that the book's source architecture is uniquely canonical.

Matrix effects

The evidence matrix now contains ten included single-coded records: four network-reduction/topology seeds, three circuit-formulation records, and three graph/port/transformation-framework records. scripts/check_evidence_matrix.py passes the controlled vocabulary and schema checks.

Limitations and next actions

  • No Scopus, Web of Science, Compendex, MathSciNet, or IEEE Xplore export was available in this run.
  • Search-result pages were used only to locate records; full-text coding should be independently repeated for every included row.
  • Deduplication, backward/forward citation chasing, and a second coder remain open.
  • The next evidence pass should add information-model and engineering-compiler records, then test whether the graph-family taxonomy is missing a recurring power-system representation.