Unveiling the Water Cut – Gradient Relationship: Statistical Analysis and Deployment Insights from NFD-NK

Abstract Objectives/Scope This paper presents an exploration of the fundamental relationship between water cut and fluid gradient based on field deployment in Kuwait Oil Company's NFD North Kuwait (NFD-NK) area. The objective was to strengthen predictive capabilities for water cut estimation by uncovering statistical relationships between ESP sensor-derived gradients and water cut levels. The study validates field-scale deployment outcomes and offers recommendations for improving reservoir monitoring strategies using these insights. Methods, Procedures, Process Using a structured data science approach, ESP sensor data—including pressure discharge (Pd), wellhead pressure (WHP), and derived fluid gradient values—was extracted from KOC's Oracle systems. Water cut data from PGOR tests served as validation references. Following the OSEMN framework (Obtain, Scrub, Explore, Model, iNterpret), the data underwent cleaning, missing value treatment, Z-score outlier removal, and normalization. Statistical techniques—including linear regression, clustering, and correlation analysis—were applied to explore the relationship between gradient and water cut. Wells were segmented into low, medium, and high water cut groups. Models assessed the predictive strength (R-squared) of Gradient as a single explanatory variable. Deployment integrated these findings into a real-time system, dynamically updating predictions with new sensor inputs. Model performance was evaluated across over 150 wells operating under varied conditions. Results, Observations, Conclusions Statistical analysis confirmed a strong inverse relationship between fluid gradient and water cut. Wells with lower gradients consistently correlated with higher water cuts, supporting fluid column behavior theories. Regression models achieved median R-squared values around 0.7, validating Gradient as a practical predictor for water cut behavior. Field deployment showed that wells transitioning into lower Gradient zones typically experienced faster water cut escalation, enabling earlier interventions. The move from manual spreadsheet analysis to real-time modeling accelerated evaluation time by more than 5,400 times, reducing analysis from over two months to less than two hours. Operationally, early detection of water cut escalation helped prevent approximately 1.29 million barrels of potential oil loss, protecting production value near USD 77.2 million. Lessons from deployment emphasized integrating anomaly detection layers to flag sudden shifts in gradient trends, which often preceded PGOR-confirmed water cut increases by several weeks. Sensor variability affected a small portion of wells, highlighting the importance of periodic recalibration and validation against well test results to sustain prediction quality. Novel/Additive Information This paper provides one of the first field-wide validations of the water cut–gradient relationship using ESP sensor analytics. By combining statistical analysis with continuous deployment, the study advances water management strategies in mature producing fields. Recommendations include using gradient monitoring as an early warning tool, developing hybrid models that integrate gradient and multivariable predictors, and applying cluster-specific strategies to improve prediction reliability across different well profiles.