Research And Implementation Of Intelligent Decision Algorithm For Sliding Directional Drilling

In the process of conventional sliding directional drilling,due to the non-rotation of the drill string,the friction between the drill string and the borehole wall tends to generate a large resistance,which causes "back pressure".Considering production safety,production costs and other factors,Torque Rocking Technology is commonly used in practical production to convert static friction into dynamic friction by applying forward and reverse torque to the drill string,which can reduce the sliding frictional resistance of the drill string,and ease "back pressure".At present,due to the lack of related theories such as dynamics,it is necessary to manually adjust the value of torque to ensure the mechanical speed while maintaining the stability of the tool face,resulting in large individual differences,slow response speed of tool face,and low degree of automation.Deep learning is a data-driven autonomous learning method in machine learning.Because it can linearize complex problems and solve the problems such as high data dimension,non-linear separability,and uncertain information delay,it has become a new research hotspot for rapid sliding directional drilling technology in recent years.In view of the above situation,based on the development project commissioned by Drilling Engineering Technology Research Institute,this paper combined with relevant knowledge of deep learning field to propose a torsion system based automatic control method of sliding directional drilling.It mainly includes three aspects: the processing and analysis of sliding directional drilling data,the research of sliding directional drilling intelligent decision-making algorithm,and the design of sliding directional drilling intelligent decision-making platform.According to the research contents,the paper first took the job data as the starting point to complete the establishment of the data set,and discussed the relationship between the parameters using hierarchical clustering.And it laid a good data foundation for the research of intelligent decision-making algorithm.Furthermore,through analyzing the methods of sliding directional drilling control,this paper established an intelligent decision-making model for sliding directional drilling based on Long Short-Term Memory(LSTM),which realized the calculation of torque values and effectively avoided the limit of physical models.Finally,combined with the establishment of the MySQL database,this paper completed the development of the sliding directional drilling intelligent decision-making platform,which realized the real-time collection,storage and monitoring of job data and adjusted control parameters automatically.In order to evaluate the feasibility and effectiveness of the method,the project team conducted a large-scale application test in Weiyuan,Moxi,Shangluo and other blocks in Sichuan.The experimental results show that,under the premise that the torsion system works normally,the sliding directional drilling intelligent decision-making platform can automatically control the tool face to rotate toward the target area and keep it stable,basically achieving the expected effect,and it is expected to realize automatic control of sliding directional drilling and reduce individual differences.