By Phase
Gas Lift Optimization — Data Requirements
Gas lift is uniquely data-rich and uniquely data-dependent. The performance of every valve, every mandrel, and every injection point is governed by the interaction of reservoir inflow, tubing hydraulics, surface injection pressure, and fluid properties. Unlike ESP or rod pump systems, gas lift optimization requires simultaneous knowledge of the downhole valve configuration, the surface injection pressure envelope, and the flowing pressure gradient in the tubing. Valve data quality is as important as production data quality.
Phase 01 Initiate Scoping the gas lift optimization opportunity
Determines whether a gas lift optimization project is technically viable and commercially justified. Gas lift systems are deceptively complex — the apparent simplicity of the surface installation conceals a subsurface system where valve spacing, operating envelope, reservoir deliverability, and surface injection pressure interact continuously. Data requirements here are high-level: enough to quantify the optimization gap and confirm data availability.
Reservoir & Production — Preliminary
- Field-level and per-well oil, gas, and water production rates — historical trend
- Number of gas lift wells in scope and their operating status
- Reservoir drive mechanism and depletion stage
- Gross production decline profile and estimated deferral volume
- Known high-GOR or high water cut wells — flags for design complexity
Gas Lift System — Preliminary
- Number of gas lift wells vs. natural flow or other lift methods
- Continuous vs. intermittent gas lift split across the asset
- General valve configuration — number of mandrels, approximate valve depths
- Age of existing gas lift designs and last design review date
- Known underperforming or poorly responding wells
Injection Supply & Surface — Preliminary
- Available injection gas volume and pressure at surface (compressor capacity)
- Current total gas injection rate across the asset
- Injection pressure variability — steady vs. fluctuating supply
- Whether injection gas is metered per-well or allocated from a header
- Current optimization workflow — manual, automated, or unmanaged
Phase Deliverables
Data Availability Assessment Optimization Opportunity Statement Gas Lift Well Inventory Injection Capacity Assessment Project Charter
Phase 02 Assess Data audit, design validation & optimization feasibility
Full technical audit of the gas lift system and its supporting data. This phase interrogates the gas lift designs themselves, validates whether wells are operating as designed, identifies valve condition issues, and determines whether SCADA and historian data is sufficient for an optimization model.
Gas lift-specific complexity: A model built without accurate valve depth and Cv data will optimize against a system it doesn’t actually understand. Operating valve depth — which valve is the live operating valve — must be confirmed by wireline survey before model training begins on any well where the design valve depth is uncertain.
Reservoir & Production Data
- Per-well oil, gas, and water production rates — daily and hourly where available
- Flowing tubing head pressure (FTHP) history per well
- Casing head pressure (CHP) history — critical for valve operating point analysis
- Flowing bottomhole pressure (FBHP) — measured or gradient survey derived
- Static bottomhole pressure and reservoir pressure surveys
- Production allocation methodology — test separator vs. multiphase meter vs. virtual
- Well test history — frequency, method, last test date per well
Fluid Properties & PVT
- Oil API gravity and dead oil viscosity by well / producing zone
- Solution GOR, producing GOR history, and bubble point pressure
- Full PVT analysis — Bo, Rs, µo, µg (required for multiphase flow correlation selection)
- Water cut history and water density / salinity (impacts gradient calculations)
- Injection gas composition and specific gravity — affects valve performance curves
- CO2 and H2S content where relevant to corrosion and flow assurance
Gas Lift Design & Valve Data
- As-installed gas lift design per well — mandrel depths (MD & TVD), valve types, port sizes
- Valve manufacturer, model, and seat/stem Cv values for each installed valve
- Test rack opening (TRO) pressures and transfer pressures for each valve
- Design injection pressure and design injection rate per well
- Operating valve depth — which valve is currently live (design vs. wireline-confirmed actual)
- Date of last wireline survey confirming valve condition and operating point
- Continuous vs. intermittent design — cycle timer settings for intermittent wells
- Wellbore trajectory (MD/TVD) and completion schematic — tubing size, packer depth
Injection Supply & Compression Data
- Compressor nameplate capacity and current operating point (suction/discharge pressure, throughput)
- Total available injection gas volume — firm vs. interruptible supply
- Injection header pressure — steady state and fluctuation range
- Per-well injection rate history — metered or calculated
- Per-well injection choke or flow control valve type and position history
- Gas composition variations at injection point
- Compressor downtime history and planned maintenance schedule
Real-Time & SCADA — Audit
- Historian system type (OSIsoft PI, Aspen IP21, Wonderware, etc.) and tag inventory
- Tag scan rates for CHP, FTHP, injection rate, and injection pressure per well
- Data completeness — % uptime per critical tag over last 3 years
- Injection rate meter type — orifice, Coriolis, turbine — and last calibration dates
- Known meter drift events and periods of missing or suspect injection data
- Availability of casing pressure control valve (CPCV) position data in historian
Valve Failure & Intervention History
- Wireline pull and redress records per well — valve condition found on retrieval
- Known valve failures — leaking valves, dummy valves in place, stuck-open ports
- History of wells operating below the design valve (indicator of valve damage)
- Workover and recompletion history — any changes to mandrel spacing or tubing
- Sand, scale, or hydrate events affecting valve operation
- Unplanned injection interruption events and production response records
Phase Deliverables
Data Audit Report Gas Lift Design Register Valve Condition Assessment Data Quality Assessment Gap Analysis & Remediation Plan Wireline Survey Requirement List Model Feasibility Assessment Risk Register
Phase 03 Select Model architecture, optimization approach & data pipeline design
Audit findings determine the right optimization approach and architecture. Gas lift optimization spans multiple problem types — injection rate allocation across a constrained header, individual well setpoint optimization, valve performance degradation detection, and intermittent lift cycle timing — and the architecture must match the asset’s data reality and dominant production loss driver.
Data Pipeline & Integration
- Historian API access and query performance benchmarking
- Real-time data streaming capability — OPC-UA, REST, or MQTT from SCADA
- Data lake or cloud storage target for training data
- ETL requirements — merging injection meter data with production historian
- Latency requirements for injection setpoint recommendation loops
- Integration pathway for writing recommended setpoints back to CPCV controllers
- Cybersecurity and OT segmentation constraints
Training Dataset Definition
- Feature set — CHP, FTHP, injection rate, GOR, water cut, reservoir pressure proxy
- Label definition — production rate, GLR efficiency, or injection sensitivity target
- Train / validation / test split accounting for temporal correlation
- Handling of injection interruption events — exclusion or separate treatment
- Time-windowing strategy for transient response modeling post-injection change
- Per-well vs. pattern-level model approach based on well interference assessment
Gas Lift Physics Inputs
- Multiphase flow correlation selection — Beggs & Brill, Hagedorn-Brown, or field-calibrated
- Valve performance curves (Thornhill-Craver) per installed valve for physics-informed models
- Nodal analysis model inputs: IPR curve, tubing geometry, completion details
- Gas lift performance curve (GLPC) data from historical injection rate step tests
- Compressor suction/discharge performance curve for supply-side constraint modeling
- Documented injection rate changes with corresponding production response
Constraint Data for Optimization
- Total injection gas availability — firm daily volume and pressure envelope
- Per-well minimum and maximum injection rate operating limits
- Surface facility throughput constraints — separator capacity, export line pressure
- Well interference matrix — which wells share a compressor or injection header segment
- Regulatory injection limits per well or field
- Operational constraints — injection restrictions for scale or hydrate prevention
Phase Deliverables
Optimization Model Architecture Decision Data Pipeline Design Feature Engineering Specification GLPC Step Test Plan Technology Stack Selection
Phase 04 Define Data preparation, model development & validation baseline
All data is prepared, models built, and validation conducted against held-out field data. For gas lift, this phase must account for the physics of multiphase flow and valve behavior — a model that cannot distinguish between a well operating below potential due to suboptimal injection versus one with a failed valve will generate unsafe recommendations.
GLPC step test requirement: If historical data lacks sufficient injection rate variation to characterize the gas lift performance curve, a planned step test program must be executed before model training. Step tests — systematically varying injection rate and recording production response — are the single most valuable data collection activity in a gas lift optimization project. Plan for 4–6 weeks of step testing on priority wells.
Cleaned & Labeled Training Dataset
- Cleaned time-series: injection rate, CHP, FTHP, and production rates — outliers removed
- Injection interruption events labeled and excluded from steady-state training windows
- GLPC step test results incorporated — injection rate vs. production response per well
- Operating valve depth confirmed and encoded as model feature per well
- Valve failure events labeled with operating conditions at time of failure
- Synchronized multi-source dataset — injection meter data merged with production historian
- Per-well dimensionless GLR efficiency feature engineered
Validation & Benchmarking Data
- Held-out well test data for production rate prediction validation
- Nodal analysis GLPC output for hybrid model calibration
- Historical injection rate change events with documented production response (held out from training)
- Known valve failure events withheld from training for anomaly detection testing
- Operator log entries correlated with anomaly events
Reservoir & Fluid Updates
- Latest PVT data — any fluid sample or recombination updates since Assess phase
- Updated reservoir pressure from most recent static gradient or BHP surveys
- Revised IPR curves post any stimulation or workover
- Updated GOR and water cut trends — rising GOR or water cut shifts the GLPC optimum
- Updated injection gas composition if supply source has changed
Acceptance Criteria Data
- Baseline KPIs: gas lift efficiency (Mcf injected per bbl produced), deferral rate, injection utilization
- Model performance thresholds agreed with operations — production accuracy, injection sensitivity accuracy
- Constraint compliance validation — model must not recommend rates outside operating envelopes
- Integration test data for SCADA / DCS write-back validation of injection setpoint commands
Phase Deliverables
Cleaned Training Dataset GLPC Library per Well Trained & Validated Optimization Model Model Performance Report Backtesting Results Data Pipeline (Production-ready) Deployment Specification
Phase 05 Execute Deployment, live optimization & continuous learning
The model deploys into the live production environment — consuming real-time injection and production data, generating injection setpoint recommendations per well, and allocating injection gas across the header against live facility constraints. A continuous learning loop captures actioned outcomes, production responses, and valve change events to progressively improve recommendation accuracy. Gas lift system changes that go unlogged will silently degrade model performance.
Live Real-Time Inference Feeds
- Streaming per-well injection rates — <5 min latency from injection meters
- Streaming CHP and FTHP per well
- Live production rates — test-corrected allocation or real-time multiphase meter
- Injection header pressure — real-time supply side constraint
- Compressor operating point — suction pressure, discharge pressure, throughput
- Surface facility real-time constraints — separator level, export pressure, water handling rate
- Injection choke / CPCV position data where available for setpoint tracking
Continuous Learning Data
- Operator override log — when and why injection setpoint recommendations were rejected
- Production response data post-injection change (actioned recommendations with time lag tracking)
- New valve failure events labeled in real time — date, condition found, and pre-failure SCADA signatures
- Monthly well test updates to refresh production allocation
- Periodic GLPC re-characterization data — injection step responses as field conditions evolve
- Model drift monitoring — prediction error trending and injection sensitivity deviation tracking
Gas Lift System Change Tracking
- Wireline valve pull and replacement records — new valve type, Cv, TRO pressure, and installation depth
- Operating valve depth changes — must be updated in model feature set within 48 hours
- Conversion of continuous to intermittent lift (or reverse) — requires model re-parameterization
- Workover outcomes — new completion parameters, tubing changes, packer depth updates
- New well tie-ins or abandonments — updates to header constraint model
- Compressor upgrades or replacements — revised supply curve required
Performance & Value Tracking
- Production uplift attributed to injection optimization vs. pre-deployment baseline
- Gas lift efficiency improvement — Mcf injected per incremental barrel (vs. baseline)
- Compression cost savings from injection gas reduction on over-injected wells
- Valve failure events predicted vs. missed — predictive detection value tracking
- Operator adoption rate and recommendation acceptance by well and engineer
- Model retraining triggers — driven by GOR shift, water cut increase, or reservoir pressure decline
Phase Deliverables
Deployed Gas Lift Optimization Model Live Injection Allocation Engine Operator Dashboard Valve Anomaly Detection Module Performance Monitoring Report Continuous Learning Framework Data Governance Protocol