AI Powers Hydrate Detection, Prevention in North Kuwait Jurassic Gas Field

_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 222206, “Enhancing Production Efficiency: AI-Powered Hydrate Detection and Prevention in North Kuwait Jurassic Gas Field,” by Abrar Hajjeyah, SPE, Alaa Al-Saleh, and Shaikha Al-Qahtani, SPE, Kuwait Oil Company, et al. The paper has not been peer reviewed. _ The Kuwait Oil Company and SLB’s joint Kuwait Integrated Digital Field project for the North Kuwait Jurassic Gas asset is a key aspect of Kuwait’s gas production. The platform presented here streamlines planning and execution with advanced workflows and automation. However, the formation of hydrates in flowlines and separators resulted in substantial deferments of gas and oil production. To combat this issue, a comprehensive hybrid approach has been adopted, combining machine learning (ML) with a physics-based risk-prediction model. Introduction The North Kuwait Jurassic Gas field is a high-pressure, high-temperature gas-condensate field that has been in production since 2010, with over 160 active wells and six gas-processing facilities and plans to more than double the producing wells during the next 3 years. The field has consistently faced challenges with hydrate formation. To overcome these challenges, a hybrid solution was developed that integrates both ML and physics-based modeling. The ML model was designed to detect hydrate-formation events by analyzing high-frequency sensor data, while the physics-based approach aimed to predict and prevent these events by understanding the conditions leading to hydrate formation. Both models were integrated into a single visualization platform, providing a comprehensive tool for hydrate management (Fig. 1). This platform allows for the validation of predicted hydrate events by comparing them with actual events detected by the ML model and reported by the production operations team. By combining predictive capabilities with real-time detection, the solution not only minimizes the risk of hydrate-related incidents but also optimizes field performance by enabling proactive management and reducing unplanned downtime. This synopsis concentrates on the methodology of the blended approaches rather than the system architecture. Physics-Based Prediction Approach Hydrate-formation temperature may be predicted using a Gibbs free-energy-minimization approach with compositional fluid modeling. The compositional fluid modeling package used in this study is a proven compositional fluid modeling software validated through many studies in hydrate prediction and mitigation. Additionally, compositional fluid modeling also accounts for the inhibition of hydrate-formation temperature as a function of methanol concentration. To determine if a specific flow path is at risk for hydrate formation, thermohydraulic simulation of the system is necessary to determine the pressure and temperature profile. The resulting flow profile then may be compared with the hydrate-formation curve to calculate the maximum hydrate subcooling delta temperature (MHST). A positive value of MHST indicates that hydrates are likely to form. The MHST often occurs near the outlet of the flowline where temperature is typically the lowest; however, this is not always the case, particularly for lines that reach ambient temperature well ahead of the delivery point.