Abstract Gas Lift Deepening is a proven production enhancement technique that improves well performance by enabling deeper gas injection in selected mature wells. The value realized from the intervention is closely linked to deployment method selection, which influences technical feasibility, execution efficiency, operational risk, and overall economics. In multi-asset environments, a structured deployment selection approach offers opportunities to further improve intervention effectiveness and portfolio scalability. In practice, Gas Lift Deepening interventions are frequently deprioritized within Production Enhancement and Idle Well Reactivation (PE/IWR) portfolios due to economic constraints when executed using conventional coiled tubing (CT) deployment. This study presents a comparative evaluation of Gas Lift Deepening deployment using CT and braided line, based on two field applications in mature offshore wells in Peninsular Malaysia. The evaluation integrates technical feasibility, operational performance, and economic effectiveness to establish a structured, fit-for-purpose approach for deployment method selection. The CT deployment, applied in a deep, highly deviation well, provided superior depth control and circulation capability, enabling successful valve placement and significant production gain. Conversely, the braided line deployment, executed in a moderate depth well with favourable wellbore conditions, enabled rapid intervention with lower operational cost while delivering meaningful production gain. Results demonstrate that CT ensures execution reliability in high-complexity wells, while braided line delivers up to 50–70% cost reduction with competitive production gains in moderate-complexity environments. This study establishes a structured, field-validated deployment framework that enables scalable, fit-for-purpose Gas Lift Deepening strategies, enhancing its viability as a competitive portfolio-level production enhancement solution across diverse well conditions. Critically, the study demonstrates that while coiled tubing and braided line differ in execution capability and cost efficiency, optimal value is achieved through aligning candidate selection, well characteristics, and fit-for-purpose deployment to maximize production gain and portfolio scalability.
Abstract Petrophysical evaluation of heterogeneous, variably silicified carbonate reservoirs is challenged by ambiguous log responses: quartz and chert compress density-neutron separation, while anhydrite and dolomite introduce further compositional ambiguity, and deterministic multi-mineral solvers require extended log suites that are frequently unavailable in legacy or cost-constrained wells. This paper presents an integrated, data-driven workflow that simultaneously predicts eleven targets—total and effective porosity, water saturation, nuclear-magnetic-resonance- calibrated permeability, five mineral volumes (clay, calcite, anhydrite, dolomite, quartz), and two pseudo-logs (photoelectric factor and potassium-thorium spectral gamma ray)—from triple-combo inputs (gamma ray, neutron porosity, bulk density, true resistivity) alone. Thirteen wells from a deepwater pre-salt and post-salt carbonate province in the Kwanza Basin, offshore Angola, were partitioned into eleven training wells and two blind-test wells (anonymized herein as Wells A-N). A shared library of physics-informed engineered features—crossplot-endpoint transforms, depth-lagged differences, rolling- window statistics, and rock-flag indicators—fed eleven target-specific regressors. Five candidate algorithms (Gradient Boosting, Extra Trees, XGBoost, LightGBM, Random Forest) were tuned and selected per target under GroupKFold well-level cross-validation. Blind-well performance reached coefficients of determination (R2) of 0.92 for total porosity, 0.91 for effective porosity, 0.97 for water saturation, 0.80 for logarithmic permeability, 0.85-0.91 for the five mineral volumes, 0.89 for synthetic photoelectric factor, and 0.82 for synthetic spectral gamma ray. Best-in-class algorithms differed by target, confirming that no single learner is optimal across heterogeneous geophysical targets. Once trained, the full eleven- target prediction executes in seconds per well, enabling near-real-time reservoir characterization, legacy-well rejuvenation without re-logging, and reduction of logging tool runs in future campaigns.
Abstract Cement plugs are widely accepted as the primary barrier for plug and abandonment (P&A) operations. However, post placement validation is often limited to tagging and short duration pressure tests conducted immediately after plug setting, providing limited insight into long-term integrity. Cement plug failures driven by in-situ wellbore changes and pore-pressure evolution remain insufficiently evaluated, and existing industry models largely neglect poromechanical coupling within the plug and its impact on abandonment failure. This work presents a novel poromechanical analysis framework that identifies the drivers responsible for cement plug failure resulting in significant financial impact and weeks of downtime for an offshore operator in Malaysia. Furthermore, the analysis method provides a basis for tailoring cement properties and placement strategies to maintain long-term integrity. The evaluated abandonment plug spanned both a cased-hole and an open-hole sections of the well and was successfully subjected to pressure and inflow tests of approximately 1,800 psi and 5,000 psi, respectively. However, a subsequent reduction in equivalent mud weight (EMW) to 8.1 ppg triggered an inflow event, forcing well shut-in operations. Remedial drillout operation revealed alternating intervals with negligible weight on bit (WOB) and hard cement, indicating intermittent structural integrity of the plug. Conventional three-dimensional near-wellbore stress analysis assuming the cement as a non-porous solid failed to explain the observed failure mode. As a result, our research efforts highlighted the need for a poromechanical framework capable of modeling a coupled near-wellbore system comprising an abandonment plug. The proposed method can model the effects of thermal, structural and hydraulic loads and predicts the evolution of stresses and pore pressure within both the abandonment plug and surrounding formations. Conventional three-dimensional near-wellbore stress analysis assuming the cement as a non-porous solid failed to explain the observed failure mode. The proposed method successfully identified that shrinkage coupled with pore pressure induced tension are the reasons for the observed failure of a conventional cement plug during the mud weight reduction, while pressure and inflow tests alone were insufficient to induce such a failure, consistent with well observations. The framework further predicted low failure risk for a remedial mechanically enhanced cement system, accurately reproducing the successful re-abandonment outcome. The proposed method enables comprehensive assessment of long-term plug integrity and provides a quantitative basis for tailoring cement poromechanical properties. Additionally, the framework allows evaluation of fluid-influx risk through porous plug material, extending beyond conventional integrity assessments.
Background: Mixed matrix membranes (MMMs) containing three components consisting of a polymeric continuous phase, a solid inorganic material, and an ionic liquid are widely explored for CO2 removal from natural gas to increase energy content, reduce corrosion, and enable safer utilization. However, most of the previous studies have relied on physically blended or impregnated ionic liquids (ILs), which suffer from leaching and membrane instability, ultimately limiting their separation performance. Moreover, experimental methods alone cannot fully explain gas transport mechanisms or interactions between polymers, fillers, and gases with sorption sites. Methods: This work employs a grafting strategy to covalently support 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide ([EMIM][Tf2N]) onto silica surface (IL-Si), ensuring long-term stability, uniform dispersion, improved compatibility between polymer and filler, and enhanced gas separation performance. Hybrid membranes with filler contents ranging from 5 to 20 wt.% were experimentally fabricated and analysed from an atomistic perspective using molecular dynamics (MD) simulations. Significant Findings: The IL-functionalized filler enhanced interfacial adhesion, as evidenced by increased thermal stability with delayed degradation and improved glass transition temperature (T-g) from 181.5 to 189.46 degrees C, reflecting stronger polymer-filler interactions. At 10 wt.% IL-Si, the membrane achieved a CO2 permeability of 25 Barrer and CO2/CH4 selectivity of 37, representing 246% and 208% improvements over neat polysulfone (PSF). Compared to non-modified silica/PSF, the permeability and selectivity improved by 140% and 40%, respectively. MD simulations, with <10% deviation, confirmed [Tf2N](-) anions enhance CO2 sorption while [EMIM](+) cations strengthen filler dispersion and compatibility. Based on this, future work needs to focus on testing functionalized ILs, scaling up fabrication, assessing long-term stability under harsh conditions, and expanding membrane studies to other relevant gas pairs.
Abstract Accurate reservoir characterization in greenfield hydrocarbon exploration remains challenging because sparse well control, seismic inversion non-uniqueness, and geological heterogeneity introduce significant uncertainty in predicting reservoir properties beyond existing wells. Conventional machine learning workflows commonly employ single-stage models that neglect the hierarchical dependency among petrophysical properties and rarely integrate rock physics calibration and geological constraints into the prediction process. This study presents an integrated machine learning-assisted rock physics workflow that combines petrophysical interpretation, rock physics modelling, seismic calibration, geological constraints, and supervised machine learning within a unified framework for field-scale reservoir characterization. Five exploration wells were used to construct a calibrated rock physics database incorporating acoustic impedance, reflection coefficients, synthetic seismograms, seismic inversion, and interpreted petrophysical properties. A physics-guided Sequential Random Forest workflow was developed to emulate the conventional petrophysical interpretation sequence by sequentially predicting porosity, lithology-related properties, shale volume, sand volume, water saturation, and bulk volume water. The workflow was benchmarked against Extreme Gradient Boosting (XGBoost) using blind-well validation. Cumulative hydrocarbon pore volume analysis established an optimum shale-volume cut-off of 0.65–0.70 v/v, while calibrated AI–Vp/Vs rock physics templates demonstrated that reservoir and non-reservoir intervals should be interpreted probabilistically because of the combined influence of multiple geological factors on elastic properties. Blind-well validation showed that both machine learning approaches produced reliable reservoir-property predictions; however, the Sequential Random Forest workflow achieved comparable predictive performance with lower model-development complexity, reduced hyperparameter optimization, and greater physical interpretability. The proposed workflow extends well-based reservoir characterization into the interwell seismic domain by integrating rock physics, geological constraints, and sequential machine learning within a physically consistent predictive framework. This methodology provides a scalable approach for reducing exploration uncertainty and supporting more reliable hydrocarbon prospect evaluation and well-placement decisions.
