Heat delivery

Injected tonnage and steam temperature give an enthalpy budget. A heat-efficiency term penalises depth and rewards a shorter, higher-rate injection window, because a slow injection loses more heat to overburden and thief zones.

Soak and mobility

Viscosity reduction follows a saturating curve in soak duration: strong gains in the first days, negligible after roughly a week, with an explicit penalty for deferred production beyond that.

Inflow and decline

Peak rate scales sub-linearly with cycle size, then declines exponentially from soak exit. Cycle length is itself a function of the design rather than a fixed calendar rule.

Lift envelope

Theoretical pump displacement from speed and stroke is compared against inflow. The ratio gives fillage; the pair gives rod loading. Over-pumping is therefore visible as a modelled quantity, not a hunch.

Risk and economics

Failure risk rises with rod loading and with sustained low fillage. Net value nets illustrative oil revenue against steam cost and a risk charge, so a scenario that pumps hard for marginal oil scores poorly.

Guardrails and assurance

Any recommendation carries guardrails, a confidence figure, feature-level explanation and an accept/defer/reject audit trail. Data-quality and drift screens sit alongside so a stale tag cannot silently drive advice.

Mobility gain saturates near day seven. Past that, the model still shows a small rate benefit but net value falls because production is deferred without a matching gain.

Peak rate scales with roughly the 0.6 power of cycle size, so steam-oil ratio degrades steadily as tonnage grows. This is the mechanism behind the steam-trim recommendations inside the demo.