Enhancing Oil Production: A Machine Learning Approach in Efficient Waterflooding

Abstract

Abstract The petroleum industry maintains a keen interest in asset assessment tools. This paper presents a practical case study involving high-level geological and dynamic data analysis to evaluate petroleum asset potential for further investment aimed at optimizing secondary recovery. The economic model, grounded in the balanced waterflood flow approach, determines the optimal injection volumes and the associated number of oil-producing and water-injecting wells. Analyzing production data is complex, relying on numerous diagnostic metrics such as reserve properties analysis, reservoir energy diagnostics and watercut/GOR diagnostics, productivity measures, and economic factors. This analysis facilitates rapid modeling of future performance and forecasts economic outcomes in response to redevelopment investments. Automation has revolutionized modern production analysis, enabling the generation of comprehensive diagnostic metrics with a simple "mouse click"—a process that typically spans months. Newly developed diagnostic metrics improve upon traditional production/injection performance analysis, especially those based on automatically generated numerical 3D micro-models that simulate expected rock/fluid properties. Well interference is assessed through mathematical algorithms for multiwell deconvolution, utilizing extensive bottomhole pressure and surface rate data. This deconvolution, either fully or semi-automated, searches for initial pressure and unit-rate transient responses in tested and adjacent wells, aligning them with actual pressure records and aggregate flow rates. Further advancements include the automated analysis of these diagnostic metrics, supported by AI-based digital tools that offer economic insights for enhancing production. The case study in Western Siberia identifies deposits and wells where not all proven recoverable reserves are being tapped. It advocates for side-tracking from current wells and implementing multi-stage fracking to activate these reserves and sustain pressure. The economic model generated by this study proposes investment scenarios with a profitability index (PI) of 1.4, an attractive prospect considering the reserves’ current maturity. The application of deconvolution in cross-well pressure interference analysis has fine-tuned production and water injection targets, yielding a 6% uplift in field oil production without the need for well interventions. This paper presents a couple of examples of waterflooding efficiency assessment and a ranked list of investment opportunities to unlock field potential. Integrating open-hole data with meticulous well-by-well production analysis, we pinpoint prospective drilling sites. Advanced production analysis notably accelerates the analysis process, thereby diminishing the risk of overlooking enhancement opportunities.