Research On Drilling Parameter Optimization Method Based On Machine Learning

With the development of oil and gas exploration and development in China to complex oil and gas fields such as deep,deep water and low permeability,the impact of complex formations with large hardness,strong abrasiveness and alternating soft and drilling efficiency and safety has become increasingly significant,which has increased the demand for drilling parameter optimization and acceleration technology.The existing drilling parameter optimization and acceleration methods based on expert experience and physical model of drilling speed have a single goal,efficiency and accuracy need to be improved,and it is difficult to cope with the real-time changes in drilling conditions and formation environment.In recent years,the intelligent development of oil and gas engineering has developed rapidly,providing new and feasible ideas for solving the complex problems of non-linearity,multi-parameter and volatility.This study fully considers the impact of drill bit wear and fault on drilling efficiency,and on the basis of the previous study with mechanical drilling speed as a single optimization goal,integrates mechanical drilling speed and mechanical specific energy,and constructs a comprehensive evaluation index of drill bit performance that considers both mechanical drilling speed and drill bit wear.The optimal stochastic forest algorithm establishes an intelligent prediction model of the comprehensive evaluation index of drill bit performance,analyzes the response relationship between drilling parameters(drilling pressure,speed and displacement,etc.)and the comprehensive evaluation index,and realizes the optimization of drilling parameters with the comprehensive evaluation index of drill bit performance as a single goal through special mathematical planning methods.In addition,in this study,hydraulic parameters are introduced to optimize the mechanical specific energy of the drill bit,strengthen the response intensity of relevant parameters such as displacement in the process of drill speed optimization,and use the non-dominant sorting genetic algorithm and particle swarm algorithm to construct an automatic optimization method of drilling parameters with mechanical drilling speed and mechanical specific energy as the multi-objective.Simulation experiments of on-site drilling data show that the above two optimization methods can significantly increase the mechanical drilling speed while reducing drill bit wear,and the drilling parameter optimization results based on the comprehensive evaluation index of drill bit performance are concentrated in the Pareto optimization solution of the multi-objective optimization method,and the time required is shorter.In short,the above two drilling parameter optimization methods can meet the on-site needs of drilling parameter optimization,and can provide theoretical basis and technical support for the automatic optimization and acceleration of complex oil and gas drilling parameters.