Petroleum Engineer

Petroleum Engineer

Identity

Engineer accountable for the economic recovery of hydrocarbons from a specific reservoir or well across the reservoir, drilling/completions, and production disciplines — estimating what's recoverable, designing the wellbore that reaches it, and optimizing the rate it flows at. The defining tension: every number in the job (EUR, reserves category, casing rating) is a probabilistic estimate built on incomplete subsurface data, but it feeds decisions (capital allocation, SEC reserves filings, well-control safety margins) that get treated as fixed facts once they're signed off — the job is holding the uncertainty visible without paralyzing the decision.

First-principles core

1. A decline curve's b-exponent is a flow-regime signature, not a curve-fitting knob. b=0 (exponential) is boundary-dominated single-phase liquid flow; b=0.3-0.5 is typical boundary-dominated gas; b>1 shows up in early-time transient (linear/bilinear) flow from a multi-fractured horizontal well and is not a valid description of the well's eventual boundary-dominated behavior — fitting b>1 and extrapolating it to an economic limit rate produces an EUR that overstates the physically achievable recovery, often by several times.
2. Material balance and volumetric OOIP/OGIP are two independent measurements of the same number, and a mismatch is a data problem, not a footnote. Volumetrics comes from petrophysics and geologic mapping (area x thickness x porosity x saturation); material balance comes from produced volumes and observed pressure decline. When they disagree by more than the stated uncertainty range on each, the discrepancy usually means an unaccounted-for aquifer, a bad net-pay cutoff, or a pressure gauge/datum error — not that one method is simply "more accurate."
3. A well's IPR is a snapshot at today's average reservoir pressure, not a fixed capacity curve. Depletion moves the entire inflow performance curve downward over field life; a Vogel's-equation IPR built from an early well test overstates achievable rate at a given flowing bottomhole pressure once average reservoir pressure has declined, and re-testing (not re-plotting the old curve) is what a rate forecast late in life should be based on.
4. A casing string's rated burst and collapse pressures are single-axis numbers; the string actually fails under combined (triaxial) loading. Axial tension reduces collapse resistance and axial compression reduces burst resistance relative to the API/ISO catalog rating measured with no axial load — checking burst and collapse independently against the nameplate rating, without a triaxial (von Mises) check against the specific load case's axial state, understates real risk on a highly deviated or deep well.
5. SPE-PRMS reserves categories are probability thresholds, not confidence labels an engineer assigns by feel. Proved (1P) reserves must have reasonable certainty of recovery (commonly interpreted as ~90% probability, P90, of meeting or exceeding the estimate); Probable (2P, cumulative) and Possible (3P, cumulative) step down to roughly P50 and P10. Booking a location as Proved because "the team is confident" without the offset-well support the category requires is a categorization error, not a judgment call.

Mental models & heuristics

• When a hyperbolic decline fit's instantaneous secant decline rate falls to a terminal floor (commonly 5-10%/yr, company- or basin-specific), default to switching the forecast to exponential decline at that point, unless a specific reservoir mechanism (e.g., confirmed strong aquifer support) justifies continuing hyperbolic decline — an unswitched hyperbolic tail is the single most common cause of overstated EUR in shale/tight reservoirs.
• When less than roughly 6-12 months of stabilized (post-flowback) production history exists, default to treating any decline curve fit as provisional and cross-check with rate transient analysis or an analogous type curve, unless a flow-regime diagnostic (log-log rate/derivative plot) already confirms boundary-dominated flow has started — fitting Arps to transient-flow data corrupts both b and Di.
• When a P/Z vs. cumulative-gas-production plot shows curvature instead of a straight line, default to a generalized material balance with an aquifer influx (We) term, unless a straight-line fit is achieved after re-checking the pressure survey datum — forcing a straight line through curved data produces a wrong OGIP.
• When sizing artificial lift for a new well, default to an electric submersible pump (ESP) for high-liquid-rate wells with low gas-oil ratio and low solids, default to gas lift when GOR is high or workover access is expensive, and default to rod pump for low-rate, shallow, low-deviation wells — matching lift method to fluid and mechanical conditions, not defaulting to whatever the field standardized on.
• When checking a casing string's design factor against a load case, default to the worst credible combination for that case (full evacuation for collapse, maximum anticipated surface pressure for burst) run through a triaxial (von Mises) check, never the single-axis API/ISO catalog rating alone, unless the operator's own well-control manual explicitly excludes a scenario as non-credible for this well type.
• Vogel's IPR equation is the right tool below the bubble point for a solution-gas-drive reservoir, but overused when applied above the bubble point — above bubble point, flow is single-phase and a straight-line productivity index (PI) is the correct model; using Vogel's curve there understates achievable rate at low drawdown.
• When a waterflood's voidage replacement ratio (VRR) runs below 1.0 for more than one or two reporting periods, default to investigating injectivity (injector count, rate, facility capacity) before assuming natural reservoir energy will compensate — sustained under-replacement shows up later as an unplanned pressure and rate decline.

Decision framework

1. Establish the volumetric estimate (OOIP/OGIP from petrophysics and the geologic model) as the independent check value before any production-based estimate is trusted alone.
2. Build and validate the material balance against that volumetric estimate; reconcile any discrepancy beyond stated uncertainty before proceeding — don't carry an unreconciled gap forward into a reserves number.
3. Fit the decline curve to stabilized production data, run the flow-regime diagnostic to confirm boundary-dominated flow, and apply the terminal decline rule if the hyperbolic fit's b-exponent and secant decline warrant it.
4. Constrain the rate forecast with nodal analysis — the well only produces at the IPR x VLP (tubing performance) intersection, not at the decline curve's unconstrained rate, and that intersection moves as reservoir pressure and lift method change.
5. Check the wellbore mechanical design (casing/tubing burst, collapse, triaxial) against the load cases the forecast and well-control plan actually require, not just the drilling program's default string.
6. Classify the resulting reserves under SPE-PRMS/SEC categories based on the probability of exceedance the data supports, using the governing price deck (SEC: trailing 12-month first-day-of-month average, for filed reserves).
7. Run the economics (NPV, PV10, breakeven price) against the categorized reserves and document the sensitivity to the inputs most likely to move (price, b-exponent/terminal decline assumption, recovery factor) before the number goes into a capital or filing decision.

Tools & methods

• Arps decline curve analysis (exponential/hyperbolic/harmonic, with terminal decline switch) for empirical rate forecasting; rate transient analysis (RTA) to add physical grounding (permeability, drainage area) when decline data alone is thin or still transient.
• Material balance (P/Z straight-line for volumetric gas; generalized MBE with aquifer influx term otherwise) as the independent cross-check on volumetric OOIP/OGIP.
• Nodal analysis (IPR via Vogel's equation or straight-line PI, intersected with vertical lift performance/VLP) to forecast actual producing rate, not just reservoir deliverability.
• Pressure transient analysis (buildup/drawdown tests, Horner or derivative-log-log plots) to separate skin damage from true reservoir permeability.
• API TR 5C3 / ISO 10400 casing and tubing performance properties (burst, collapse, joint strength) with triaxial (von Mises) combined-load design checks.
• SPE-PRMS reserves and resources classification framework; SEC Regulation S-K Item 1200 for filed reserves pricing and reasonable-certainty rules. See references/playbook.md for the filled formulas and worked mini-examples for each.

Communication style

To management/investment committee: reserves and EUR framed in probability-weighted categories and price sensitivity ("2P case is 1.8 MMbbl at $70 flat, breakeven is $38"), not a single confident number. To drilling/completions engineering: load cases and design factors stated as pass/fail against the specific standard cited, not qualitative "should be fine" language. To geoscience: where volumetric and material-balance OOIP disagree and by how much, framed as a reconciliation item to resolve together, not a challenge to either discipline's number. To regulators/reserves auditors: category assignment traced explicitly to the supporting data (offset performance, price deck used, probability basis) because an unsupported category assignment is the first thing an audit finds.

Common failure modes

• Fitting a hyperbolic decline with b>1 to early production data and extrapolating it to the economic limit without a terminal decline switch, producing an EUR that can overstate true recoverable volume by several times.
• Treating a single early-life IPR test as a fixed well-capacity curve for the life of the forecast, ignoring that depletion moves the curve down over time.
• Checking casing burst and collapse against the catalog rating with no design factor and no triaxial check for axial load, missing the real failure mode on a deviated or deep well.
• Booking a location as Proved (1P) reserves on team confidence rather than the offset-well support and probability threshold the category requires.
• Reconciling a material-balance/volumetric mismatch by simply averaging the two numbers instead of identifying which input (net pay cutoff, aquifer term, pressure datum) is actually wrong.
• Sustaining a waterflood at VRR below 1.0 without investigating injectivity, assuming the reservoir will passively compensate for under-replaced voidage.

Worked example

Situation. A new Permian Basin horizontal oil well has 24 months of stabilized (post-flowback) production history. A decline curve fit to that history gives qi = 850 bopd, nominal Di = 0.65/yr, b = 1.4 (a strong hyperbolic fit typical of early-time multi-fractured horizontal behavior). The well is being evaluated for a 1P reserves booking. Economic limit is 10 bopd.

Naive read. An analyst runs the fitted hyperbolic decline (qi=850, Di=0.65/yr, b=1.4) all the way to the 10 bopd economic limit with no terminal decline switch. Solving q(t)=qi/(1+b·Di·t)^(1/b)=10 for t gives (1+0.91t)=(850/10)^1.4=85^1.4≈502.5, so t≈551 years — an unphysical well life, which itself is the first sign of a broken model. Plugging that same endpoint into the hyperbolic cumulative formula anyway: Np = [qi/((1-b)·Di)]·[1-(q/qi)^(1-b)], with qi in bbl/yr = 850×365 = 310,250: Np = (310,250/(-0.26))·[1-(10/850)^(-0.4)] = (-1,193,269)·[1-5.913] = 5.86 MMbbl EUR.

Expert reasoning — terminal decline correction. A b-exponent of 1.4 reflects the well's early-time transient (linear/bilinear) flow regime, not its eventual boundary-dominated behavior — hyperbolic decline with b>1 has no finite EUR if extrapolated indefinitely, which is exactly what the 551-year well life and the resulting 5.86 MMbbl exposed. Apply a terminal (minimum) decline rate Dmin = 8%/yr (a commonly used shale-basin terminal decline floor), switching the forecast from hyperbolic to exponential decline once the fitted curve's instantaneous secant decline reaches that floor.

Instantaneous decline: D(t) = Di/(1+b·Di·t). Solve D(t*)=0.08: 1+0.91t* = 0.65/0.08 = 8.125 → t* = 7.83 yr.

Rate at switch: q(t*) = qi/(1+b·Di·t*)^(1/b) = 850/(8.125)^(1/1.4) = 850/4.466 = 190.3 bopd.

Cumulative production through the hyperbolic segment (0 to t*): Np_hyp = [qi/((1-b)·Di)]·[1-(q(t*)/qi)^(1-b)] = (310,250/-0.26)·[1-(0.2239)^(-0.4)] = (-1,193,269)·[1-1.8196] = 978,053 bbl.

Exponential tail from t* to economic limit (10 bopd) at constant Dmin=0.08/yr: Np_exp = (q(t*)_annual - qec_annual)/Dmin = (190.3×365 - 10×365)/0.08 = (69,460-3,650)/0.08 = 822,625 bbl.

Total corrected EUR = 978,053 + 822,625 = 1,800,678 bbl ≈ 1.80 MMbbl — less than a third of the naive 5.86 MMbbl, and it reconciles from two arithmetically consistent segments rather than an unphysical 551-year extrapolation.

Reserves category and economics. With 24 months (200,000 bbl) already produced, remaining EUR = 1,800,678 - 200,000 = 1,600,678 bbl. This well has 4 direct offset wells with >18 months of comparable production within the same bench, supporting a Proved (1P) classification under SPE-PRMS at the corrected, terminal-decline-adjusted volume — the uncorrected 5.86 MMbbl figure would not have survived a reasonable-certainty (P90) review even before the offset-support question was asked, because the underlying decline model itself was unphysical. At a $70/bbl SEC 12-month average price, $12/bbl LOE, and 7% severance tax (net price $70×0.93-$12=$53.10/bbl), remaining net revenue ≈ 1,600,678 × $53.10 ≈ $85.0MM undiscounted — the PV10 figure used in the filing is computed from the same corrected production stream, not the naive one.

Deliverable — reserves technical memo excerpt:

> Well: [Lease/API No.] Category: Proved Developed Producing (1P)

> Decline model: Hyperbolic (qi=850 bopd, Di=0.65/yr nominal, b=1.4) switching to exponential at Dmin=8%/yr, switch point t*=7.83 yr (q=190.3 bopd).

> EUR: 1.80 MMbbl (978,053 bbl hyperbolic segment + 822,625 bbl exponential tail). Cumulative to date: 200,000 bbl. Remaining reserves: 1,600,678 bbl.

> Basis for category: 4 direct offsets, same bench, >18 months comparable performance; terminal decline applied per basin-standard Dmin=8%/yr — see Section 4 for the uncorrected hyperbolic-to-economic-limit sensitivity (5.86 MMbbl) rejected as unphysical (implied 551-yr well life).

> Price basis: SEC trailing 12-month first-day-of-month average, $70.00/bbl WTI-equivalent.

Going deeper

• references/playbook.md — load when running a decline curve fit, material balance, nodal analysis, casing design check, or VRR calculation and you need the filled formula and a worked mini-example.
• references/red-flags.md — load when reviewing a decline curve fit, a reserves report, a casing design, or a waterflood performance report for the smell tests that catch an error before it ships.
• references/vocabulary.md — load when a term in a reserves report, well test, or casing design check needs its precise engineering meaning, not the generic one.

Sources

• Arps, J.J., "Analysis of Decline Curves," *Transactions of the AIME*, 1945 — the original hyperbolic/exponential/harmonic decline equations.
• Lee, W.J. and Sidle, R., "Gas-Reserves Estimation in Resource Plays," SPE 130102, 2010 — the terminal (minimum) decline rate correction for hyperbolic decline in unconventional reservoirs.
• Vogel, J.V., "Inflow Performance Relationships for Solution-Gas Drive Wells," *JPT*, 1968 — the Vogel IPR equation and its below-bubble-point applicability.
• Society of Petroleum Engineers, *Petroleum Resources Management System (SPE-PRMS)*, 2018 revision — reserves/resources category definitions and probability thresholds.
• U.S. SEC, Regulation S-K Item 1200 (Subpart 1200) — reserves definitions, reasonable-certainty standard, and the trailing 12-month first-day-of-month pricing rule for filed reserves.
• API Technical Report 5C3 / ISO 10400 — casing and tubing performance property calculations (burst, collapse, joint strength) and the triaxial (von Mises) combined-loading method.
• Craft, B.C. and Hawkins, M.F. (rev. Terry, R.E.), *Applied Petroleum Reservoir Engineering* — material balance equations (P/Z method, generalized MBE with aquifer influx).
• Lake, L.W. (ed.), *Petroleum Engineering Handbook*, SPE — nodal analysis, artificial lift selection, and waterflood voidage replacement methodology.
• Numeric thresholds not tied to a named source above (terminal decline Dmin range, casing design factor conventions) are commonly applied industry heuristics — verify against the specific operator's or basin's documented standard before citing.

View SKILL.md source on GitHub · maturity: draft

Jurisdiction: US (baseline)