Source: InfluxDB MARAGON_Generators, field power_avg (kW), max() per gen_id over range(start: -24h), window ending 2026-08-24 18:06 UTC. Roster: 123 gens in master_config.json, excluding sites Spares and Out of Service.
| Bucket | Gens (n) | Group avg peak (kW) | (330 − avg) × n | MW vs 330 assumption |
|---|---|---|---|---|
| >330 (over nameplate) | 9 | 331.85 | −16.6 kW | −0.02 MW (exceeds 330) |
| 320–330 | 65 | 326.83 | 206.0 kW | 0.21 MW |
| 310–320 | 11 | 314.18 | 174.0 kW | 0.17 MW |
| 300–310 | 4 | 307.25 | 91.0 kW | 0.09 MW |
| <300 | 23 | 228.67 | 2,330.5 kW | 2.33 MW |
| Total (112 running) | 112 | 305.13 | 2,784.9 kW | 2.78 MW |
"Missing power" = (330 kW − group's actual average peak) × gens in that group, i.e. the MW each group would add if every one of its gens peaked at the 330 kW nameplate cap (max_gen_kw in master_config) instead of its observed 24h peak. Summed across the 112 running gens: 2.78 MW short of a 330 kW-per-gen assumption. The >330 group already exceeds 330 (transient overshoot, up to 337 kW), so it contributes a small negative (−0.02 MW), not a shortfall — 9 gens briefly ran above their own nameplate rating, worth a look but not a power gap.
Excluded from "running" (not in the 112, not in the chart):
• 4 gens had telemetry in the window but never rose above 0 kW: 22-24-243, 22V-25-027, 22V-25-020, 22V-25-475 — stopped the whole 24h, not a data gap.
• 7 gens had zero power_avg points at all in MARAGON_Generators for the last 24h: 22-24-325, 22-24-313, 22V-25-468, 22V-25-037, 22V-25-464, 22-24-326, 22V-25-568 — this is a telemetry gap, not a claim they were stopped or running. Check status_api live_stale/Iridium fallback for these before treating them either way.
Method note: "peak" is each gen's own individual 24h max — since gens don't necessarily hit their peak at the same instant, the 34.18 MW total is not the same as "MW the fleet produced at any one moment" (that figure would come from status_api / a single-timestamp snapshot, and is normally lower, ~29–30 MW per the live running-fleet method). Treat this report as a capability/degradation view, not a live output figure.