Overview
National grid state, distributed nodes, and the physics-supervised model at a glance.
LEM thesis
The Large Energy Model observes grid signals, detects instability, coordinates distributed autonomous nodes through PAVLINA, and improves orchestration policy over time using physics-supervised learning.
Demand vs generation — last 24h
Generation mix (latest)
Live Elexon FUELINST metered generation only. Embedded (distribution-connected) solar and wind are not included yet — total will under-report actual UK output by ~5–8 GW on sunny/windy days. Full three-source merge coming.
Grid frequency
Nominal 50 Hz ± 0.05. Wider excursions surface as events for PAVLINA to consider.
Physics-supervised vs baseline (demand forecast)
The clean green trace is the output of a model whose loss function already penalises conservation-law violations at training time — not an ordinary forecaster with a post-hoc constraint filter bolted on. The amber baseline is that same architecture trained without the physics term; it stays calm until it hits a conservation-violating sample, where it distorts into a sharp ECG-style spike and pulses a red ring. Visualization is stylised — underlying series is seeded demo data, not a real-time feed.
Data provenance
- Mock seeded dataActive (live)
Historical grid, carbon, node telemetry, events, PAVLINA and training series seeded into the demo database.
- Simulated autonomous nodesActive (live)
12 representative nodes stand in for a synthetic fleet of 10,000 distributed assets.
- Live upstream feedsUpstream error
Elexon BMRS (metered generation) + NG ESO carbon + Open-Meteo, refreshed every 20 min via cron. Embedded solar/wind (PV_Live, NESO) disabled pending source-aligned merge. carbonintensity_uk: ok · elexon_demand: ok · elexon_generation: ok · neso_embedded_wind: disabled · openmeteo_weather: error · pvlive_solar: disabled.
- Live control layerPlanned
Future PAVLINA outbound actuation to real DER, EV and BESS operators via signed control APIs.
