The model · assumptions
Every assumption, disclosed
The model turns decisions into a fleet and reports what it would produce and cost. Nothing here is settled by a citation. This page lists every input: its default, whether it rests on a published source or is a bare site assumption, and how sensitive the outputs are to it.
Base data: EIA 2024. Fleet reconciles to 1230 GW national net-summer capability, -1.71% off. ← Back to the model
1 · Base fleet & data
| Input | Default | Source | Sensitivity |
|---|---|---|---|
| Operating fleet | EIA-860M, 2024-12, net-summer capacity | cited | Sets the whole starting point. Reconciled to EIA's national total within 2%. |
| Capacity factors | Empirical, base-year generation ÷ (net-summer capacity × 8,760 h) | cited | Drives generation from capacity. Uses year-end capacity, so fast-growing solar reads low. |
| Gas CC vs peaker CF | peaker 0.12 (assumed), CC = residual | site assumption | National data can't split gas generation; peaker CF is assumed, CC reconciles the total. |
| Load & VRE hourly shapes | EIA-930, base-year US48, seasonal 24-hour profiles | cited | Sets dispatch timing — when solar and demand actually land. |
| Demand base | 4295.5 TWh (base-year generation) | cited | Grown by the demand-growth knob; the single biggest driver of unserved energy. |
2 · Retirement & construction
| Technology | Service life (yr) | Lead time (yr) |
|---|---|---|
| Coal | 50 | 4 |
| Gas (CC) | 35 | 3 |
| Gas (peaker) | 40 | 2 |
| Nuclear | 60 | 7 |
| Wind | 25 | 2 |
| Solar | 30 | 1 |
| Hydro | 100 | 6 |
| Battery | 15 | 1 |
| Oil | 50 | 3 |
| Biomass | 45 | 3 |
| Geothermal | 40 | 4 |
| Other | 40 | 3 |
site assumption Service lives are typical, not measured per plant; lead times are roughly Lazard's published construction times. Plants already past their nominal life at the start are retired on a 8-year ramp (oldest first) rather than all at once, since they are demonstrably still operating in the base year. Sensitivity: high. Shorter lives or the “end-of-life” policy retire the thermal fleet faster, opening a firm-capacity gap that build rates must fill.
3 · Hourly dispatch
| Assumption | Default | Source | Sensitivity |
|---|---|---|---|
| Merit order (marginal $/MWh) | Gas (CC) $30, Coal $32, Gas (peaker) $60, Oil $130 | site assumption | Decides which flexible plant runs first. Affects the dispatch mix, not total unserved energy. |
| Availability derate | Gas (CC) 0.87, Coal 0.85, Gas (peaker) 0.9, Oil 0.9 | site assumption | Max hourly output share for firm plant. Directly sets firm capacity and reserve margin. |
| Storage duration | 4 h per MW | site assumption | How long batteries sustain output. Longer duration cuts evening shortfalls. |
| Storage round-trip efficiency | 0.85 | site assumption | Energy kept per charge/discharge cycle. Second-order for reliability. |
| Baseload treatment | Nuclear, hydro, geothermal, biomass run flat at their CF | site assumption | Simplification: hydro's peaking flexibility is not credited (conservative for reliability). |
4 · Feedbacks
| Assumption | Default | Source | Sensitivity |
|---|---|---|---|
| Solar learning rate | 20% per doubling | cited | Published solar experience curve. Sets how far solar capital falls with deployment. |
| Battery learning rate | 18% per doubling | cited | Published battery experience curve. Adjustable in the controls. |
| Solar / battery base capital | $1100/kW, $1400/kW | site assumption | NREL ATB-class reference. Anchors the capital integral; moves cumulative-capital totals. |
| ELCC of solar | 0.51 at low penetration → 0.12 at high | cited | Fitted to the published marginal ELCC spread across MISO, CAISO, SPP, PJM, ERCOT, NYISO. |
| ELCC of wind / battery | wind 0.4→0.15, battery 0.95→0.1 | cited | Same framework; a 4-hour battery saturates the net-peak quickly, so its value falls fastest. |
| Reference capital, other techs | nuclear $7000/kW, gas CC $1250/kW … | site assumption | NREL ATB / Lazard-class. No learning applied; used only for the capital integral. |
5 · Uncertainty & impact coefficients
| Assumption | Default | Source | Sensitivity |
|---|---|---|---|
| Per-TWh coefficients (deaths, CO₂, land, cost) | Identical to the descriptive site | cited | OWID, IPCC AR5 Annex III, van Zalk & Behrens 2018, Lazard LCOE+ v19.0. Never re-forked here. |
| Horizon uncertainty widening | +2 pts of half-band width per year | site assumption | Structural uncertainty beyond the fixed coefficient band. Base year unwidened; ~+52% band at 2050. |
| Fossil pollution-control anchor | Global central (adjustable: stringent / moderate / limited) | cited | Re-anchors fossil death rates within their published range, as on the country pages. |
6 · Out of scope — stated, not modeled
The model answers “what would this fleet produce and cost,” not “could this happen.” It does not model transmission and interconnection queues, electricity prices or market clearing, siting and permitting, manufacturing supply-chain limits, or policy feasibility. A scenario the model runs cleanly may be impossible for reasons entirely outside it.