Geothermal power-cost model.
Compare 1,080 cost scenarios using the retained lifetime-output models. Change installed plant cost, subsurface cost and financing to calculate what a target electricity price requires. The scenarios are assumptions, not supplier quotes.
Published field interpretation · SLB / Utah FORGE · 28 August 2024
Follow the evidence downhole.
Null samples remain gaps. Depth is not time. Downhole barrel rates are not surface barrel rates.
Where the surface flow is attributed.
Trace is censored, not exact zero. Hatched bar: unresolved deep aggregate. Individual rates reconcile to the published total.
Do not choose an intervention from this profile alone. The deep aggregate is unresolved; thermal capacity, connectivity and valve authority are not identified. Those require further measurements.
Source quality and attribution
SLB / Leonid Kolomytsev; University of Utah / Casey Chadwick; Utah FORGE. GDR 1668 · DOI 10.15121/2473673 · CC BY 4.0. Processed external field evidence, not Zemanium hardware measurements.
SIMULATED · GEOPHIRES-X 3.18.3 · fixed assumed geometry · 25-year horizon
Export, not gross output.
Export = gross − 5% of gross − pumping. Negative years remain visible; no low-cost claim is computed for an importing operating year.
What can the design afford?
Additional installed capital above the assumed $107.5m, with 2%/year incremental O&M. Negative means capital must fall. No supplier price or savings is invented.
Model boundary, parameter warnings and fair comparison
24 cases separate temperature, flow, diameter and productivity index. Geometry is assumed: 1 km³, 20 fractures, 900 m height. Fixed capital across different flows is a controlled comparison, not optimized plant sizing. Financial stress uses 2,000 common draws: 4–12% real discount; 0.8–1.2 capital multiplier; $0–10m replacement in years 8–17. Neither sensitivity nor sample count validates geology.
Registered failure cases
Model limits.
Smaller active volume
Keep the 240 °C / 100 kg/s / 12-inch / PI 50 scenario, but reduce assumed reservoir volume from 1 to 0.014 km³. This is an adverse design scenario, not a FORGE lifetime forecast.
REJECTED: finite inventory crossed; negative gross generation and importing years. No economic result is released from this case.
Required next evidence: active heat-exchange volume, recharge/contact bounds and a reservoir/conversion model valid through depletion. The public FORGE heat assessment motivates the volume challenge; it does not validate this geometry.
Claims the record does not establish
- “36% of flow is wasted.” It is unresolved in this interpretation, not proven wasted.
- “Larger wells always win.” Heat loss, flow and completion cost can reverse the decision.
- “The 240 °C case proves cheap power.” The engine warns its ORC correlation may be invalid above 200 °C.
Source hashes make this replay checkable. They do not authenticate the original sensors or prove an intervention.
Campaign status · SIMULATED records · frozen gates
Lifetime output and superhot duty
Lifetime net electricity per well pair
The current control search stopped at its frozen headroom gate: 0 of 12 development cases cleared 25% gain against both strong oracles. No causal controller was promoted.
The perfect-information schedule behind the diagnostic is neither deployable nor a global optimum. The 48-case evaluation set was not run. This stopped search does not prove the $30/MWh mission impossible; it names the next measurement.
Superhot instrument duty
Target: 500 operating hours at nominal 450 °C gauge-segment temperature, five-point accuracy within ±5 °C and housing temperature at or below 100 °C, independently witnessed. Proposed bench duty; not a current tool rating.
Blind echo acquisition runs. Matched-duty amplitude calibration and waveform-bias correction remain the next physical measurements. No Zemanium hot-duty hours are measured in this release.
Two independent evidence gates: success on either track cannot certify the other. Public campaign gates are in the technical dossier's sensing campaign (optional reading; every tab and calculation on this page works offline).
Two SOTA targets. No SOTA result claimed today.
SIMULATED / assumption scenarios · peak-sized plant · 25-year horizon
Shared capital budget.
Design to the joint ceiling.
The block ceiling reserves the selected subsurface and other costs. These allowances are alternatives, not additive savings. Signed negative ceilings remain visible: even a free block cannot close that case.
Inputs for the next review.
- Operator: constrain sustainable flow, pressure, temperature decline, chemistry and contact volume.
- Plant supplier: quote one fixed machine and its off-design map, including auxiliaries, cooling and reinjection constraints.
- Completion lead: cost the actual bore and lifetime workovers. Reconcile every excluded scope before spending the budget.
Calculation boundary and primary sources
No new physics was generated by this cost sweep. The retained GEOPHIRES-X 3.18.3 trajectories assume a reservoir geometry and 90% utilization; no firm-capacity guarantee follows. Export deducts pumping and 5% of gross auxiliaries. Annual costs: 2% of installed capital, source make-up water and $8/exported MWh. Explicit events: $5m in year 12 and $2m in year 25. All capital is at time zero; construction interest is not modeled. Installed block, subsurface and other scope are unquoted allowances in constant real USD, not scope-complete estimates.
The design preset requires 294 kg/s, 12-inch internal bore and long-lived contact in an assumed 1 km³ reservoir; it does not establish these in a real field. The 240 °C ORC extrapolations and importing/finite-inventory failures receive no price in this tab.
NLR 2025 ATB: whole-plant costs and resource assumptions · DOE GETEM: coupled plant effectiveness and cost · GEOPHIRES-X: conversion-model boundaries. Whole-project benchmarks are not block-only prices. Source highlights and extracted primary text are retained in the QA research records.
Reproduction and evidence hashes
Rebuild from the retained QA records with tools/build-confidence-demo.py. Model runs: tools/confidence-campaign.py. Field replay: tools/forge-profile.py. The supplied page works offline without accounts, a backend or live model inference.