What are the best practices for reservoir simulation?

At-a-Glance: Reservoir simulation best practices ensure physically consistent models, numerical stability, and decision-grade forecasts by rigorously controlling inputs (PVT/SCAL), grids, well models, and numerical settings, then validating via disciplined history-matching and uncertainty workflows. Focus KPIs: forecast accuracy, runtime stability, and actionable optimization insights.

I. Objective Definition and Key KPIs

• I.1 Objective: Build and run robust reservoir simulation models that are physically sound, numerically stable, and predictive for development planning, production optimization, and reserves/uncertainty assessment.
• I.2 Primary KPIs:
  • I.2.1 Forecast accuracy: 12-month-ahead oil/gas/water rates and pressures within ±10% (P50), actuals bracketed by P10–P90.
  • I.2.2 Material balance closure: cumulative mass error =0.5% PV over history period.
  • I.2.3 Numerical performance: =95% timestep acceptance; average Newton iterations =6 per step; CFL compliance; CPU-hours/case minimized for target fidelity.
  • I.2.4 Decision latency: ensemble cycle time to recommendation =2–4 weeks (depends on model size).
  • I.2.5 Economics linkage: NPV uplift from scenario selection vs. baseline; reserves categorization confidence (volumetric vs. dynamic).
• I.3 Secondary KPIs: reproducibility (100% case re-runs identical), data freshness (PVT/SCAL vintage =3 years or justified), well constraint violation rate = 0.

II. Critical Parameters and Target Ranges

III. Step-by-Step Procedure / Workflow / Checklist

III.A Data QC and Model Scoping

• III.A.1 Define objectives: short-term optimization vs. long-term development; select physics (black-oil, compositional, thermal) accordingly.
• III.A.2 QC inputs: structure, faults, facies, petrophysics, PVT, SCAL, well surveys, logs, tests, RFT/PLT, rates/pressures, water chemistry, tracer, seismic.
• III.A.3 Align datum/elevation and units; establish consistent depth and pressure references.
• III.A.4 Delineate boundaries (no-flow, aquifer, lateral seals); choose aquifer model if required.

III.B Physics and Rock/Fluid Characterization

• III.B.1 PVT:
  • III.B.1.a Black-oil: ensure smooth Rs(p), Bo(p), µo(p); Bw/µw salinity-corrected; Bg(p,T) realistic.
  • III.B.1.b Compositional: tune EOS to lab (CME, CVD, DL, MMP/MIS) using minimal parameters; validate Rachford–Rice residual to machine precision.
• III.B.2 SCAL:
  • III.B.2.a Endpoint and Corey exponents from SCAL; correct for wettability and hysteresis where needed.
  • III.B.2.b Capillary pressure via Leverett J-function scaling across rock types.
• III.B.3 Geomechanics (if relevant): include rock compressibility and stress-dependent k/f; couple to simulator or apply tabular functions.

III.C Gridding, Upscaling, and Property Modeling

• III.C.1 Use corner-point grids honoring structure/faults; apply LGR around wells, fractures, and thin high-perm streaks.
• III.C.2 Upscale k/f and SCAL with flow-based methods in key flow directions; verify via single-phase and multiphase benchmarks.
• III.C.3 Enforce transmissibility multipliers across faults/seals and anisotropy (kx, ky, kz) consistent with geology.

III.D Well and Facility Representation

• III.D.1 Build deviated/horizontal trajectories; perforation by layer with mechanical skin and damage/cleanup scenarios.
• III.D.2 Use proper well index (Peaceman) and multi-segment well (MSW) for long horizontals or complex completions.
• III.D.3 Apply correct constraints: BHP, THP, max liquid, gas-lift, ESP/PCP performance curves; surface network coupling where applicable.

III.E Initialization and Diagnostics

• III.E.1 Initialize via hydrostatic columns with capillary effects; validate Sw–So–Sg sum to 1 everywhere.
• III.E.2 Run single-well/sector tests: PI, coning tendency, breakthrough times; compare against analytics.
• III.E.3 Check mass balance closure on short pilots; verify pressure diffusion behavior and aquifer response.

III.F Numerical Controls and Stability

• III.F.1 Adaptive timestepping based on ?S, ?p, and non-linear iteration counts; cap growth factor (e.g., ×1.2–1.5).
• III.F.2 Choose robust linear solver/preconditioner; tighten tolerances for compositional/thermal cases.
• III.F.3 Limit well control switches per step; use smooth ramping of constraints to avoid oscillations.

III.G History Matching (HM)

• III.G.1 Define misfit metrics and weights: rates, BHP/THP, RFT, PLT, WOR/GOR, FWHP; avoid over-weighting noisy data.
• III.G.2 Prioritize physics: match pressure support and contacts first, then rates, finally saturation diagnostics.
• III.G.3 Use sensitivity (design of experiments, adjoint) to prioritize influential parameters; guard against non-uniqueness.
• III.G.4 Employ ensembles (e.g., EnKF/EnRML) for uncertainty-consistent HM when data density supports it.

III.H Forecasting, Scenarios, and Uncertainty

• III.H.1 Run decision-focused scenarios: well placement, lift optimization, WAG cycles, polymer/surfactant, infill timing, facility limits.
• III.H.2 Quantify uncertainty: P10/P50/P90 from ensembles; communicate with tornado/Sobol indices to show drivers.
• III.H.3 Validate near-term forecasts monthly vs. actuals; recalibrate controls as needed.

III.I Governance and Reproducibility

• III.I.1 Version control all inputs (grids, props, schedules); maintain a changelog and case lineage.
• III.I.2 Automate case runs and post-processing; record random seeds for stochastic workflows.
• III.I.3 Peer review at key gates: pre-HM, post-HM, pre-decision; archive sign-offs.

IV. Risk & Mitigation (HSE, Reliability, Redundancy)

• IV.1 Decision risk from model bias: Mitigate via multi-concept models, ensemble HM, and explicit uncertainty communication.
• IV.2 Numerical instability: Control with adaptive timesteps, LGR, proper well indices, and solver tolerance discipline.
• IV.3 Data quality risk: Systematic QC, outlier detection, reconciliation (rates vs. meters vs. test separators).
• IV.4 Operational misalignment: Integrate surface constraints/network; validate wellbore hydraulics to avoid infeasible forecasts.
• IV.5 Computational risk: Use checkpointing, run monitors, and failover strategies; run pilot cases before full ensembles.
• IV.6 HSE (indirect): Avoid recommendations that push wells beyond safe drawdown or sand control limits; include integrity envelopes in constraints.

V. Optimization Levers (Data Analytics, Maintenance Strategy, Debottlenecking)

• V.1 Data analytics: Use adjoint gradients for well placement/control optimization; proxy/surrogate models (kriging, polynomial chaos, ML) for rapid screening; global sensitivity (Sobol) to focus parameters.
• V.2 Numerical debottlenecking: Hybrid grids (unstructured near wells), transmissibility upscaling, domain decomposition, and HPC parallelism to reduce CPU-hours by 2×–10×.
• V.3 Closed-loop reservoir management: Assimilate surveillance (RFT/PLT, tracers, 4D seismic) on cadence; update controls (lift, choke, WAG ratio) to maximize NPV subject to constraints.
• V.4 Model maintenance: Refresh PVT/SCAL when new lab data arrives; recalibrate aquifer strength with long-term pressure trends; implement automated regression tests after any model change.
• V.5 Scenario governance: Standardize naming, input templates, and QA gates; de-duplicate cases; focus on decisions with material impact (NPV, reserves, uptime).

VI. Verification & Monitoring Plan

VI.A What to Measure and How Often

• VI.A.1 Weekly: Runtime KPIs (CPU-hours/case, iterations/step), timestep rejection rate, constraint violations.
• VI.A.2 Monthly: Forecast vs. actuals (rates, BHP/THP, WOR/GOR), misfit trends, material balance closure, aquifer influx diagnostic.
• VI.A.3 Quarterly: Uncertainty envelopes vs. actuals; retune priors if systematic bias; audit SCAL/PVT relevancy.
• VI.A.4 Event-based: After new wells, workovers, EOR pilots; run update cycle within 1–2 weeks.

VI.B Acceptance Criteria

• VI.B.1 P50 forecast within ±10% on 6–12 month horizon; P10–P90 bracket =80% of observed.
• VI.B.2 Mass balance error =0.5% PV; no persistent misfit bias per well.
• VI.B.3 Numerical stability =95% accepted steps; average Newton =6; no chronic time-step cuts at the same controls.

VI.C Governance and Reporting

• VI.C.1 Dashboards for KPIs and misfit plots per well/field.
• VI.C.2 Quarterly peer reviews; model freeze dates aligned with investment gates.
• VI.C.3 Archive reproducible packages (inputs, deck, scripts, logs, seeds).

Key Equations and Formulas

• Pressure diffusion (slightly compressible): \( \phi c_t \frac{\partial p}{\partial t} = \nabla \cdot \left( \frac{k}{\mu} \nabla p \right) + q \)
• Darcy’s law (linear): \( q = - \frac{k A}{\mu B} \frac{\mathrm{d}p}{\mathrm{d}x} \)
• Radial inflow with skin: \( q = \frac{2 \pi k h}{\mu B \left[\ln\left(\frac{r_e}{r_w}\right) + s\right]} (p_e - p_{wf}) \)
• Well Index (Peaceman): \( WI = \frac{2 \pi k h}{\ln(r_e/r_w) + s} \), and \( q = \frac{WI}{\mu B}(p_{cell} - p_{wf}) \)
• Fractional flow (water/oil): \( f_w = \frac{1}{1 + \frac{k_{ro}}{k_{rw}} \frac{\mu_w}{\mu_o}} \)
• Corey-type relperm: \( k_{rw} = k_{rw}^{end} \left( \frac{S_w - S_{wi}}{1 - S_{or} - S_{wi}} \right)^{n_w} \), \( k_{ro} = k_{ro}^{end} \left( \frac{1 - S_w - S_{or}}{1 - S_{or} - S_{wi}} \right)^{n_o} \)
• Capillary pressure and Leverett J: \( P_{c,ow} = P_o - P_w \), \( J(S_w) = \frac{P_c}{\sigma \cos\theta} \sqrt{\frac{k}{\phi}} \)
• Havlena–Odeh material balance (oil reservoirs): \( F = N E_o + m N E_g + W_e \), with \( F = N_p B_o + W_p B_w - G_p B_g \)
• Rachford–Rice (phase split): \( \sum_i \frac{z_i (K_i - 1)}{1 + \beta (K_i - 1)} = 0 \)
• CFL condition (stability): \( \text{CFL} = \frac{v \, \Delta t}{\Delta x} \le 1 \) (stricter near saturation fronts)
• Saturation closure: \( S_w + S_o + S_g = 1 \)
• Compressibility: \( c = -\frac{1}{V} \frac{\mathrm{d}V}{\mathrm{d}p} \)