How much more can a transmission line carry when you actually look at the weather?

A Dynamic Line Rating (DLR) demo: free NWP forecast data → IEEE 738 conductor ampacity along an 8-span demo corridor near Boise, Idaho — live, per span, per hour, as MCP tools any AI agent can call.

get_dlr_rating(hours_ahead=0) — sample output, July 8, 2026, 6 AM MT

"Line rated 816 A vs 725 A static — +12% uplift, limited by span 4 (wind nearly parallel to wire)."

+12%capacity above the static rating in this snapshot — connect for the live number

8 spansBoise valley → foothills, each with its own wire bearing

0 keysstateless — Open-Meteo + NWS, fetched live per call

Why this exists

Static line ratings assume the worst day of the year: 40 °C air, 2 ft/s of wind, full sun. Real weather is almost always better — and wind direction matters as much as speed, because crosswind cools a conductor far better than parallel wind. This server turns that physics into per-span, per-hour ampacity so the headroom (and the limiting span) is visible at a glance.

This is the productization layer around forecasts — data plumbing, IEEE 738 heat balance, forecast verification, and a clean tool interface. It is deliberately not the science layer (mesoscale modeling, CFD, line sensors).

The five tools

Tool	Returns
`get_dlr_rating`	Per-span IEEE 738 ampacity, limiting span, % uplift vs static rating
`get_corridor_forecast`	Per-span hourly 80 m wind, 2 m temperature, solar radiation (HRRR / GFS / ECMWF)
`get_forecast_skill`	Forecast vs KBOI observations: bias / MAE / RMSE for temp and wind
`get_station_observations`	Latest ASOS/METAR ground truth
`get_fire_weather_alerts`	Active NWS alerts at the corridor midpoint

Verification, not vibes

A rating is only as good as its forecast, so the server verifies itself. Sample run (July 8, 2026): last 3 days of Open-Meteo forecasts vs 500 hourly KBOI observations — the get_forecast_skill tool recomputes this live on every call:

Variable	n	Bias	MAE	RMSE
Temperature (°C)	500	+0.02	0.91	1.21
Wind speed (m/s)	485	−0.29	1.02	1.28

Physics validated against the published IEEE 738 worked example for 795 kcmil ACSR "Drake" — this implementation reproduces 1025 A with matching heat-balance components (q_c ≈ 82 W/m, q_r ≈ 39 W/m).

Connect

MCP endpoint (Streamable HTTP):
https://gridweather.publicmcp.org/mcp

Claude Code:
claude mcp add --transport http gridweather https://gridweather.publicmcp.org/mcp

Then ask: "What's the line rating on the corridor right now, and how does tomorrow afternoon look?"

Stack

Python · FastMCP · httpx — one file, ~400 lines, no database, no scheduler
Open-Meteo forecast API (HRRR / GFS / ECMWF; free non-commercial) + NWS api.weather.gov
IEEE 738 simplified steady-state heat balance; conservative fallbacks when fields are null
Hosted on the PUBLICMCP infrastructure, but private: unlisted, not in the registry