| Vose Software

Industry: Oil and Gas Exploration
Product: ModelRisk
Application: Pipeline Throughput Optimization Under Field and Capacity Uncertainty


When a 280,000 bbl/d Plan Spends a Third of Days Below the Tariff Floor

Daily throughput distribution (300,000 bbl/d nameplate)

A gathering system's nameplate is the headline number; the tariff agreement is the number that pays. A North American midstream operator running a 300,000 bbl/d crude pipeline serving five upstream fields had committed under a take-or-pay structure to deliver at least 85% of nameplate — 255,000 bbl/d — on a rolling monthly average. The deterministic forecast summed the five field mid-cases and got 280,000 bbl/d, comfortably above the floor. The probabilistic re-run in ModelRisk, shown above, told a different story: a mean daily throughput of only 259,000 bbl/d and a breach of the 85% floor on roughly a third of days, with a P10 of 237,000 bbl/d. A "comfortable" plan was demonstrably not.

The team rebuilt throughput as a stochastic sum of five field-level processes feeding a capacity-constrained pipeline, ran 10,000 simulated days, and used the output to size the next \$180M capacity expansion against the actual distribution of demand rather than the average.

Throughput as a stochastic sum capped by the pipe

Daily throughput is

\[ \text{Throughput} = \min\Bigl(\sum_{i=1}^{5} Q_i \cdot U_i \cdot (1 - C_i),\; \text{PipeCapacity}\Bigr) \]

where each input is uncertain:

  • \(Q_i\) — field daily flow rate. LogNormal with field-specific medians (95k bbl/d for mature Field A down to 18k bbl/d for end-of-life Field E) and sigmas from 0.10 (stable mature) to 0.25 (declining).
  • \(U_i\) — segment uptime. Beta-distributed with shape parameters calibrated to field reliability history (mean uptimes 89%-93%).
  • \(C_i\) — operational curtailment. Bernoulli activation (4%-10% per day) times a Normal magnitude, capturing gas-handling and water-handling constraints.
  • PipeCapacity — Normal around 300k bbl/d with sigma 8k, plus a 2% daily probability of a planned-maintenance outage (capacity = 0).

The cap term \(\min(\cdot, \text{PipeCapacity})\) is what introduces a real non-linearity: on good days the pipeline is the binding constraint and gains from field outperformance are lost.

The daily throughput distribution

Returning to the distribution above: the deterministic plan said 280k bbl/d, but the simulation reports a mean of only 259k bbl/d, with a P10 of roughly 237k bbl/d and a P90 of 290k bbl/d. Most importantly, the probability that daily throughput falls below the 85% tariff floor (255k bbl/d) is roughly 33%. A plan written against the deterministic sum is a plan that breaches the take-or-pay one day in three.

The visible spike at zero days reflects the 2% planned-outage probability — separated from the operating distribution rather than averaged in, because the regulatory metric for the tariff is computed on operating days only.

Per-field contribution — and per-field uncertainty

Per-field contribution to gathering-system throughput

Field A (mature, conventional) is the largest contributor at about 89k bbl/d mean with a P10-P90 of about 26k bbl/d. Field B (new wells, still ramping) has a smaller mean — about 67k bbl/d — but its P10-P90 spans roughly 35k bbl/d, the widest of any field. This concentration of variance in Field B is the throughput plan's structural exposure: when Field B has a bad week, the tariff floor is at risk regardless of how well the other four fields do.

What actually moves the throughput number

Tornado: drivers of annual throughput variability

The ranking confirms what the per-field chart hints at: Field B production variability is the single largest driver of throughput uncertainty (~±18,000 bbl/d half-spread), with pipeline uptime second and Field A decline-rate uncertainty third. Field E end-of-life timing is a fifth-place driver — not negligible, but not where the next dollar of de-risking spend belongs.

Sizing the capacity expansion — to the right percentile

The headline question was whether to upgrade the pipeline to 360k bbl/d nameplate (a \$180M project) alongside Field B's planned 15-well drilling program. Re-running the full simulation with both changes:

Throughput distribution: current vs $180M capacity-upgrade case

  • Mean throughput rises from 259k to about 291k bbl/d.
  • P(throughput < 300k bbl/d) falls from about 97% to roughly 58%; under the old pipe the system delivered below 300k on virtually every day, meaning Field B's new wells would have been routinely curtailed without the upgrade.
  • P(below 85% of new nameplate of 360k = 306k bbl/d) is roughly 68% — meaning the upgraded pipe should be tariffed at a substantially lower floor, not the same percentage of nameplate.

That last finding was the project's most consequential output: the commercial team had been pricing the upgrade against the assumption that 85% utilization would hold by analogy with the current pipe. The simulation showed it would not — the upgraded mean of 291k bbl/d sits at only 81% of the new nameplate — which changed the take-or-pay negotiation with shippers and the project IRR by roughly two points.

What changed

  • Capacity upgrade approved at \$180M, with the IRR case built on the simulated distribution rather than the deterministic sum, and the tariff floor benchmarked at 80% rather than 85% of the new nameplate.
  • Field B drilling sequencing accelerated by one quarter to reduce the gap between the old-pipe constraint period and the new-pipe in-service date.
  • Curtailment-management software procured for the operations control room, prioritized at Field B where 35% of the throughput variance lives.
  • Tariff renegotiation opened with shippers using the new probability distribution as the basis — including a 5%-of-revenue band of conditional discounts for sub-floor delivery days, which the simulation showed was a fair price for the actual risk.

ModelRisk Functionality Used

  • Compound throughput simulation with five field-level LogNormal flow rates, Beta uptimes, and a capacity cap — producing the daily distribution rather than five separate deterministic terms.
  • Beta uptime distributions with shape parameters fitted to per-field reliability history, avoiding the legacy practice of using a single 92% point estimate across all fields.
  • Tornado on Field B variance identifying that 30% of throughput-spread reduction would come from a single field's flow-rate confidence improvement — directing the next subsurface-modeling spend.
  • Capacity-cap modeling with min() so that good-day upside is correctly limited by the pipe, instead of being averaged into a fictional unconstrained throughput.
  • Paired before/after upgrade scenario run on common random numbers, producing a like-for-like distribution shift instead of a difference-of-means.

The deterministic sum of five mid-case field forecasts is a number with no probability attached. The 85%-tariff-floor breach rate is a probability with consequences. ModelRisk is what gets the operator from the first number to the second — and from there to a capacity-upgrade decision sized to the demand that will actually show up.