This paper describes the achievements of reservoir and well management during the first two years of production from Thunder Horse in the Deepwater Gulf of Mexico (DW GoM). Thunder Horse is a subsea development of several large and complex reservoirs. It came on production in June 2008 and had ramped up to a field rate in excess of 250,000 boe/day from nine producing wells by December 2009. Successful ramp-up and high first year efficiency were achieved using advanced technology and multidisciplinary interaction. Lessons were learned in sand management and flow induced vibration and well management problems were mitigated.Thunder Horse has a flexible development concept, particularly with regard to water injection. This reflects its high initial subsurface uncertainty, especially regarding compartmentalization and the degree of aquifer support. The value of the flexible approach has been demonstrated. However, it makes strong demands on rapid learning. This paper describes how this learning has been achieved and how it is being used to drive the flexible development.Surveillance data and analysis were the keys to rapid definition of the subsurface risks. Data from permanent pressure gauges were analyzed using advanced techniques. Reservoir simulation models were calibrated to the full set of real time data. Temperature and oil composition data provided additional controls. Results demonstrated that faults in Thunder Horse, though sealing in some places, do not in general cause compartmentalization. They can even provide vertical connectivity to aquifer in deeper sands, allowing a high degree of energy support. At the same time it creates complex paths for water movement. This changes the role of water injection in large areas of Thunder Horse from early pressure support to later optimization of an aquifer-dominated sweep.

This paper describes the achievements of reservoir and well management during the first two years of production from Thunder Horse in the Deepwater Gulf of Mexico (DW GoM). Thunder Horse is a subsea development of several large and complex reservoirs. It came on production in June 2008 and had ramped up to a field rate in excess of 250,000 boe/day from nine producing wells by December 2009. Successful ramp-up and high first year efficiency were achieved using advanced technology and multidisciplinary interaction. Lessons were learned in sand management and flow induced vibration and well management problems were mitigated.Thunder Horse has a flexible development concept, particularly with regard to water injection. This reflects its high initial subsurface uncertainty, especially regarding compartmentalization and the degree of aquifer support. The value of the flexible approach has been demonstrated. However, it makes strong demands on rapid learning. This paper describes how this learning has been achieved and how it is being used to drive the flexible development.Surveillance data and analysis were the keys to rapid definition of the subsurface risks. Data from permanent pressure gauges were analyzed using advanced techniques. Reservoir simulation models were calibrated to the full set of real time data. Temperature and oil composition data provided additional controls. Results demonstrated that faults in Thunder Horse, though sealing in some places, do not in general cause compartmentalization. They can even provide vertical connectivity to aquifer in deeper sands, allowing a high degree of energy support. At the same time it creates complex paths for water movement. This changes the role of water injection in large areas of Thunder Horse from early pressure support to later optimization of an aquifer-dominated sweep.

This paper introduces a novel technique that allows real-time injection monitoring with distributed fiber optics using physics-informed machine learning methods and presents results from Clair Ridge asset where a cloud-based, real-time application is deployed. Clair Ridge is a structural high comprising of naturally fractured Devonian to Carboniferous continental sandstones, with a significantly naturally fractured ridge area. The fractured nature of the reservoir lends itself to permanent deployment of Distributed Fiber Optic Sensing (DFOS) to enable real-time injection monitoring to maximise recovery from the field. In addition to their default limitations, such as providing a snapshot measurement and disturbing the natural well flow with up and down flowing passes, wireline-conveyed production logs (PL) are also unable to provide a high-resolution profile of the water injection along the reservoir due to the completion type. DFOS offers unique surveillance capability when permanently installed along the reservoir interface and continuously providing injection profiles with full visibility along the reservoir section without the need for an intervention. The real-time injection monitoring application uses both distributed acoustic and temperature sensing (DAS & DTS) and is based on physics-informed machine learning models. It is now running and available to all asset users on the cloud. So far, the application has generated high-resolution injection profiles over a dozen multi-rate injection periods automatically and the results are cross-checked against the profiles from the warmback analyses that were also generated automatically as part of the same application. The real-time monitoring insights have been effectively applied to provide significant business value using the capability for start-up optimization to manage and improve injection conformance, monitor fractured formations and caprock monitoring.