A Quantitative Drilling Fluids Advisor

Abstract The management of drilling fluid properties has, in most cases and continues to be a highly manual process. A combination of infrequent measurements of insufficient accuracy, fluids data not being stored as a time series, limited understanding of the interactions between critical components and low levels of mixing system automation has made the creation of a system to advise on fluid treatment a considerable challenge. Our solution is to create a fluids advisor - a system of inter-connected algorithms, which work together to give quantitative recommendations of fluid treatment and management. We achieve this by combining field data collection and analysis with an in-depth physicochemical understanding to enable automated treatment advice. This advisory system was tested successfully during the drilling of two oil-based and one water-based sections in the UK North Sea. The advisor's self-learning framework creates a workflow that encompasses the automated detection of critical fluid treatment decisions points, calculates the exact quantities of treatments required to achieve an objective and can autonomously decide if a treatment action should be taken. In addition, the advisor automatically detects the presence of contaminants in the fluid and any operation on the rig that affects the fluid system (Fluid States). This is achieved using a hybrid data-physics model created using sequential design of experiment (DoE), combined with extensive experimental work on fluid formulation and incorporation of contaminants and low-gravity solids. This allows the determination of fluid properties resulting from any treatment made to the fluid system and to prescribe quantitative treatments to fluid engineers. A self-learning data framework enables treatment recommendations from the model to be modified during and after each operation to improve advice and predictions for future operations. Advice provided to the offshore engineers consisted of qualitative advice on different treatment and operational decision points and quantitative recommendations on what treatments were required to achieve a particular objective. 100% of qualitative recommendations were accepted by the offshore engineers with 75% being implemented immediately. Changes to properties due to treatments were predicted accurately with a MAPE of ∼5% across an API fluids check, indicating the excellent agreement between measurements and prediction. The rig's fluids engineers used the advisor (via remote operations) to optimise treatments to the fluid system. Overall, use of the advisor used 15.4% less chemical additives than offset sections. Our approach combines both physics and data with innovative work on clay exfoliation to enable truly accurate, quantitative fluid treatment recommendations. For the first time, the process of recommending, calculating and implementing fluid treatments is automated fully.