Rotormaps Katana

What's New

• v1.6.19 — 2026 grid-capacity data, now live in the app: available-capacity figures published by all nine TSOs for 2026 are now used across the map, popups and analysis. 7,009 transmission lines and substations carry the new figures, of which 5,310 had no operator capacity data at all before and 1,517 were previously showing 2025 numbers. Provenance is now explicit: from 2026 the operators publish forecast flow (予想潮流) rather than available capacity, so most headroom is derived rather than published — every figure is labelled with which it is, because a derived number is a screening estimate and not an operator commitment. Data integrity: capacity records withdrawn after a matching review are retained and flagged rather than deleted, so any figure previously quoted stays reconcilable, and the app never displays a withdrawn value. Capacity matched to a circuit only by an ambiguous equipment number can fill a gap but can no longer overwrite a confidently-identified figure.
• v1.6.18 — Per-user logins & admin console: the shared access password is replaced by individual accounts — every user signs in with their own email and password. An in-app Admin page lets an administrator create users, enable/disable access instantly, reset passwords, and remove accounts. Disabling a user revokes their session on the very next request. Q-Headroom Pro: a new behind-the-login map adds a 50 MW datacentre hosting-feasibility layer alongside the BESS Q-headroom layer. Thermal accuracy fix: branches with no real rating in the dataset no longer report phantom overloads — the hosting verdict now counts only branches that carry a genuine rating.
• v1.6.17 — Engine generalisation: the convergence scoping is now device-agnostic, so a datacentre-only analysis (no BESS) is judged on its own electrical sub-network just like a BESS solve. Preparatory work for the Q-Headroom Pro datacentre layer.
• v1.6.16 — Datacentre large-demand what-if: drop a datacentre pin to model a large constant P/Q demand (up to 200 MW contracted) at a chosen utilisation. The datacentre connects like a generation pin — nearest/chosen substation, line tap, or a drawn dedicated line — and is solved as a steady-state PQ load bus. The engine also surfaces a heuristic in-rush & fault-duty screen at the connection point, derived from the solved network impedance. Limitation: Katana is a steady-state (fundamental-frequency) solver, not a transient/EMT model, so in-rush and fault duty are order-of-magnitude screening estimates only — always labelled as such, and never a substitute for a transient study. The CPAT export carries the datacentre load, and the results panel badges the demand scenario with the steady-state-only caveat.
• v1.6.15 — Major release. Nodal analysis: solver accuracy groundwork for the CRIEPI/CPAT gold-standard load-flow benchmark — the Newton–Raphson solver now honours PV/slack generator voltage setpoints, models EHV line charging (the Ferranti effect), exports per-branch reactive flow, and respects the caller's convergence tolerance. Validated end-to-end against the IEEJ WEST 10-machine daytime benchmark: Katana's solution matches the published IEEJ load-flow map to within 0.005° and is identical to an adapter-free raw-PyPSA control run (zero adapter effect). Identical solver results to v1.6.14 for all normal platform use — verified across five random seeds and the 89-test engine suite. Grid Code compliance verdict: a new tier-aware badge next to Converged now separately states whether the solved voltages are Voltage Compliant, in Warning, or NOT Compliant — convergence and grid compliance are no longer conflated. CPAT export: any regional analysis can now be exported as Y-method data files for CPATFree (CRIEPI's L-method solver), with a live CPAT Ready badge in the header and an Export to CPAT button in the results panel. Data enrichment: 2,400+ substations in the TEPCO and Chugoku service areas now carry operator-published available-capacity (akiyouryou) data, merged alongside REI capacity with additive, latest-wins precedence. Performance: warm startup ~6× faster via orjson parsing and a topology-graph pickle cache. Internal: engine pre-solve pipeline extracted to a single canonical module and all modelling constants consolidated into one tunable registry — no behavioural change, but calibration is now a one-file job.