Construction ~4 kW avg / 7 kW peak, ~55 kWh/day. Measured at Kalaeloa.
Basecamp ~30 kW active / ~43 kW peak, ~540 kWh/day. Anchored to a film-power vendor's basecamp rule of thumb: ~300 A average / ~420 A peak (film amps — summed across the three 120V legs). Real power = 120 V × total A × PF: 36 kVA avg / 50 kVA peak → ≈30 / 43 kW at 0.85 PF (mixed HVAC + motor loads). Trailers, catering, 24/7 HVAC and holding.
Lighting ~85 kW active day / ~115 kW night, ~1,100 kWh/day. A large electric package: sits between indie draws (~80–150 kWh/day) and the installed capacity of a big feature, which runs two ~750 A gensets. In film terms 750 A is the per-leg rating, so one genset = 120 V × (3 × 750) ≈ 270 kW and two ≈ 540 kW installed (a touch less at the ~0.9 PF of ballasted HMI/LED) — heavily diversified in practice. The night case is the hard one for solar.
Each 15-minute step serves load in this order: solar → grid → generator → battery, and charges the battery solar surplus → grid → generator. Grid is placed ahead of the diesel everywhere it appears, for both serving load and recharging — so wherever grid capacity exists, it's used before a generator fires. On Oahu grid carbon (1.49 lb/kWh) is cleaner than a lightly-loaded diesel; the generator is the last resort.
The generator only starts when grid + battery can't meet the load, or SOC hits its floor and grid can't recharge it. When it does run it serves the load directly and run-hard charges the battery toward its ceiling, then shuts off. Running the engine to charge the battery and load-following it directly are both modeled; the controller never discharges the battery while a generator is running to charge it.
Multiple generators stage independently: the controller brings on only the fewest units (or, for mixed sizes, the size-appropriate combination) needed for the moment's demand, each loaded efficiently — not all units at once. Each running unit burns fuel on its own curve. An earlier version lumped N generators onto one curve, which under-counted partial-load fuel; staging each unit separately is less optimistic and more accurate, so modeled multi-genset reductions are lower than before — and more defensible.
Construction: the two DCA70s that actually ran, idling at the 25% load-bank floor through production hours — 816 gal. What the job really used, not a strawman.
Basecamp / Lighting: one diesel sized to the load's peak, running continuously and load-following but never below the 25% floor. Conservative — in practice gaffers often bring far larger generators than the load needs (a film-power vendor notes basecamps drawing ~300–420 A are routinely fed by 1400 A sets), which idle at low load and burn more. Modeling a right-sized diesel understates the real diesel baseline rather than inflating it.
SGA-60: 60 kW array, modeled at 82% of nameplate peak (~49 kW), ~315 kWh on a clear day.
Per-day sky derate: clear 100% / partly 70% / overcast 40%. "Typical" = 13 clear / 6 partly / 2 overcast across 21 days (leeward Oahu) — a representative pattern, not weather-station data. We default to Typical, not Clear, so numbers aren't cherry-picked.
RPS150 = 30 kW discharge / 132 kWh, and absorbs up to 28 kW per unit when charging; RPS1200 = 240 kW / 1042 kWh. Round-trip efficiency modeled at ~90% (0.95 × 0.95). Standby/parasitic draw 0.5 kW (RPS150) / 1.5 kW (RPS1200). The per-unit charge cap bounds how fast solar, grid, or a generator can refill the pack — so undersized storage curtails (wastes) midday solar it has no room to hold. Sizing matters: a battery sized for a construction load can't bank enough daytime sun to carry a heavy night-lighting load, which is why the night cases need more storage (the tool shows the curtailment when it happens). The efficiency and parasitic figures sit on the optimistic side.
Deployed & measured: 2× RPS150 + DCA70, DEIF-controlled, on a real production.
Proposed (not yet deployed at scale): the SGA-60 solar asset, RPS1200-class storage, and the full peer-to-peer DEIF ASC/AGC control mesh shown in the single-line. The diagram shows the proposed architecture — not an as-built system at every scale.
Optional crew-EV load: up to 3 Level-2 chargers at 40 A / 9.6 kW each, charging Tesla Model 3 Long Range from 30→90% (~47 kWh to the car, ~53 kWh from the bus at ~90% onboard-charger efficiency, ~5.5 h per session). Sessions are randomly staggered through the work window, queued to the 3 chargers, averaging the selected charges/day with ±1.5 day-to-day variability. Level-2 AC charging is modeled as a flat 9.6 kW with no 30→90% taper — the taper people picture is a DC-fast-charge (Supercharger) behavior; on AC the car's 11.5 kW onboard charger isn't the bottleneck below ~95% SOC. Charging is deliberately not solar- or SOC-aware — it's random demand the system has to absorb, which is why adding EVs lowers the reduction %. The same EV load is added to the diesel-only baseline, so the comparison stays fair.
A representative smooth daily shape (flat base + active block), not a per-device simulation. The spiky transient behavior (peak:average up to ~8×) lives in the PTLC moment, not the energy run — so the dispatch view shows energy balance, not instantaneous peak adequacy against real inrush.
"Load served" is a model output — the genset always backstops the load. It is not an uptime or availability guarantee; maintenance, derate, and failures aren't modeled.
Diesel 22.4 lb CO₂/gal; Oahu grid 1.49 lb CO₂/kWh (eGRID HIOA). Operational emissions only — equipment lifecycle carbon isn't counted. The diesel factor cross-checks against the industry figure of 2.6 kg CO₂/L (≈21.7 lb/gal).
Load magnitudes informed by published film-power and sustainability material: Valid Mfg (basecamp & lighting draw), LMU/SFTV production handbook (feature lighting generators), RMI & Sustainable Production Alliance / Sustainable Entertainment Alliance (production fuel as the dominant emissions source), Anton Bauer (diesel CO₂ factor). Film "amps" are summed across the three 120V legs, so kW = 120 V × total A × PF (equivalently √3 × 208 × A/leg for a balanced load). Figures are representative, not any single production's metered data.