System Dynamics and Control of Deep Drilling Systems

获取地下深处的资源仍然是许多行业面临的挑战，包括石油和天然气开采以及地热能源的利用。对于油气资源来说，容易开采的储层已经被开发出来，而新的目标则处于更具挑战性的环境和地层中。从历史上看，进入储层的井具有简单的垂直或切线轨迹，但现代井越来越多地具有复杂的三维轨迹，具有高弯曲度，或弯曲的井道。为了提高钻井效率，减少钻柱部件的故障，并通过减少钻井次数来最大限度地减少钻井过程对环境的影响，在钻井过程中控制钻柱的动态至关重要。

Accessing resources deep underground remains a challenge across a variety of industries, including oil and gas extraction and harnessing geothermal energy. For oil and gas resources, easy to access reservoirs are already being exploited and newer targets are in ever more challenging environments and formations. Historically, wells drilled to access reservoirs had simple trajectories either vertical or tangent but modern wells increasingly have complex three-dimensional trajectories with high tortuosity, or snaking of the wellpath. To improve drilling efficiency, reduce failures of drillstring components and to minimize the impact of the drilling process to the environment by reducing drilling times, it is paramount to control the dynamics of the drillstring during the well drilling process. The drilling process entails transmission of torque and axial force from the drilling rig at the surface to the drillbit along a thin (7-15cm in diameter), kilometers long drillstring that lies inside a snaking and tortuous wellpath. Downhole sensing is typically limited to near bit sensors, but low bandwidth (~10 bits / second) and high latency (up to 30 seconds) means that traditional closed loop feedback control is highly inefficient for control of bit-rock interaction and drillstring dynamics. To achieve fully closed loop, automated control of the drilling process, reliable real-time, physics-based models of the system dynamics and effective online parameter fitting are necessary. Models exist that seek to quantify the effects of borehole inclination and tortuosity on the static behavior of the drillstring; however, the dynamic behavior of the system is only now being fully measured, understood, and quantified. Drillstring modelling has stretched back sixty years, and there have been significant efforts to create comprehensive drillstring models, but these models are computationally complex and are unsuited to real-time optimization or control. The proposed research program seeks to develop a series of feedforward or model predictive control strategies by developing a set of reduced order models of the dynamic behavior of the drillstring system, validating them with high quality calibrated, precise and continuous data recorded in the laboratory and the field, and implementing online fitting of model parameters through machine learning techniques. The development of these accurate, validated and computationally efficient models of the drilling system will improve control systems and equipment design and will drive an overall increase in the efficiency of the drilling process. The control systems developed will further increase the efficiency and safety of drilling operations and reduce the carbon footprint of operations by mitigating the adverse effects of coupled vibrations, improving drilling performance and improving wellbore quality.