The studied case addresses oil production decline in a field characterized by complex geology and challenging waterflood monitoring. A notable example is the oil rim reservoir, where water mobility significantly exceeds oil mobility due to high relative permeability. Despite low drawdown values at horizontal producers, partial aquifer support, and existing pressure maintenance facilities, oil production rates decline sharply.
The objective of the study was to reconstruct the history, pace, and future trends of formation pressure decline and to propose waterflood optimization strategies. A machine learning-based approach was applied to design these activities and enhance oil recovery.
Conventional production history was supplemented with deconvolved cross-well interference signals derived from long-term pressure records collected by permanent downhole gauges. Modern dynamic data interpretation techniques were used to assess dynamic skin factors and pressure depletion patterns. In addition, reservoir permeability and coning gradients were calculated, and open-hole logs were reinterpreted to identify geological features influencing well performance.
Machine learning models trained on historical data uncovered subtle reservoir behavior and supported decision-making in areas such as waterflood monitoring, stimulation potential, and pressure maintenance. The study made extensive use of AI for well and reservoir analysis, contributing to the development of encroachment based on skin factor dynamics, and calculating optimal drawdowns for horizontal producers.
Cross-well pressure transient analysis confirmed that the pay zone is connected to an aquifer—even during early production stages when water had not yet reached the wells. A large aquifer at the base and edges of the pay zone creates a composite system that can give the illusion of high skin around producers. This effect was carefully accounted for when calculating the true mechanical skin factor, which remained elevated—indicating potential for near-wellbore stimulation.
A decreasing skin factor over time was found to be a reliable indicator of approaching water encroachment. This insight enabled preemptive measures, such as redistributing injection volumes, to delay water breakthrough and reduce the risk of increased operating costs or expensive remediation.
The study introduced an empirical method to calculate optimal drawdown pressure values for each well individually and recommended redirecting injection efforts toward zones with more rapid pressure decline. Most of the proposed activities—including adjustments to injection and production targets, as well as targeted reservoir stimulations—were successfully implemented in the field. The observed production increase aligned closely with the study's forecasts, confirming the effectiveness of the proposed approach.
