State and Parameter Identification Applied to Dual Gradient Drilling with Water Based Mud
Abstract
There is a continuous search for new reserves in the oil and gas industry. Most of thelarger fields that are accessible with conventional drilling technology have been drilled.The remaining fields typically contain less oil and gas, are harder to drill and locatedat significant depths. The costs of drilling wells have increased while the oil price hasdropped. It is therefore a strong demand for drilling technologies able to drill where conventionaldrilling cannot be used, while still being cost and time efficient. In addition,after the blowout in the Gulf of Mexico, the industry has been challenged to develop newsolutions improving safety. As a response to increased demands, solutions with a higherdegree of automation, improving pressure control have been developed, and are referred toas Managed Pressure Drilling. A sub technology in this category is Dual Gradient Drilling.Statoil ASA uses this technology at one of their offshore installations, and after gainingoperational experience, field data has become available. This field data gives the opportunityto validate mathematical models and estimate unknown parameters. Once a verifiedmodel has been established, it can be used to experiment with controller design and tuning.This will ease controller tuning offshore, which in turn saves valuable rig time.In this thesis, a fit-for-purpose mathematical model for a Dual Gradient Drilling systemwith partially filled riser is derived. Using the available field data, unknown parametersrelated to the mud circulation part of the system is identified. The model is also augmentedto account for the presence of the U-tubing effect. A steady state friction model was foundsufficient to describe the frictional losses, and a subsea pump model was obtained by optimization.The presence of U-tubing was found when ramping down the booster pump.Due to the lack of measurements and in-depth system knowledge, a complicated modelwas discarded in favor of adding simplified dynamics to the booster pump. The simplemodel successfully compensated for U-tubing. Through simulation the derived model,with identified parameters, was found to be able to reproduce the field data in a satisfactorymanner with only small deviations during U-tubing.