Evaluation And Lessons Learned from Water Shut-off Treatments in Kuwait Reservoirs

Over the past five years, 170 water shutoff (WSO) treatments were applied in three major Kuwait Oil Company (KOC) oil-producing assets. Of these, 60% were chemical treatments, and the remaining 40% were mechanical. An extensive examination was conducted to (1) identify where and how these treatments are most effective and (2) develop a road map for future applications. This paper presentsthe major lessons learned from the evaluation of those WSO treatments.

The evaluation incorporated treatment reports, production test data, wireline open-hole and cased-hole logs, workover history, PVT and SCAL data, as well as geo-cellular and dynamic simulation models. The examination included well events (e.g., ESP placements), water injection effects, high-permeability streaks, cement integrity, unintended crossflow, water coning, and natural fractures (Bailey, 2000). A substantial database was developed to systematically organize and analyze the treatment results. Production data before and after the WSO treatments were analyzed, applying a success-criterion to distinguish wells with effective outcomes from those with poor performance (Seright, 2003). Artificial-intelligence/machine-learning methods were also applied to the data (Mohaghegh, 2000). Success and failure drivers were systematically identified and tabulated, supported by insights from full-field geo-cellular and history matched simulation models.

Approximately 50% of the wells showed a favorable response, particularly those with perforations in multiple reservoir sub-zones. The treatments in Asset 1 achieved over 60% success, whereas the treatments in Assets 2 and 3 only had 30-35% success rates. Wells with successful WSO jobs in Asset 1 were dominantly chemical treatments in the crestal areas of the field. Good responses in the Asset 3 occurred when water was isolated in identifiable thin high-permeability layers (Sydansk, 2011). Mechanical methods worked notably better than chemical methods in this asset. Poor responses were attributed to (1) unintended crossflow/uneven injection configurations, (2) short completion intervals, (3) completions in thick permeable layers where the entire interval was water swept, and (4) close proximity of injectors to the waterfront (Willhite, 1998). In Asset 2, good responses were seen (1) with separated perforation sets, (2) large completion intervals, and (3) large standoff between current perforations and current fluid contact. Poor responses were seen (1) when partial WSO in continuous perforation intervals failed to restrict the water movement, and (2) with low standoff between the current perforations and the water contact.

Chemical treatments particularly showed poor performance in short completion intervals. This paper applies logical engineering analyses to understand the results and points towards how these learnings can improve future applications of WSO. Multiple machine-learning models produced debatable success, while basic engineering insights proved more effective (Mohaghegh, 2000). Leveraging years of accumulated field data—rather than relying on unstructured technology deployment—can identify proven success factors, avoid repeating suboptimal practices, and provide actionable guidance for future WSO planning and execution.