Intelligent Modeling Of Formation Drillability Field And Drilling Rate Of Penetration Optimization In Complex Conditions

Ensuring the safety of resources and energy is one of the keys to realizing the sustainable development of national economy and an essential part of national security.With the increasing depletion of shallow mineral resources and the exploration of a large number of deep resources,deep geological exploration becomes inevitable.However,in the process of deep geological drilling,there exists a complex geological and mechanical environment which resulting in low accuracy of formation drillability model and low efficiency of drilling process.This paper focuses on the research of formation drillability modeling,three-dimensional formation drillability field modeling,high-accuracy prediction of drilling rate of penetration(ROP),drilling ROP optimization and other related issues.The main research results and innovations are as follows.(1)Data-driven intelligent modeling methods are proposed to predict the formation drillability.Based on the analysis of the correlations between logging parameters and formation drillability,two kinds of data-driven intelligent modeling methods for formation drillability are proposed.Firstly,a data-driven fusion modeling method for formation drillability is proposed.Pearson correlation analysis method is used to determine the input parameters which have a strong correlation with formation drillability,so as to reduce the coupling between parameters.Combining the extreme learning machine algorithm with the improved adaptive ensemble learning algorithm,a new fusion modeling method is proposed and applied to the calculation of formation drillability.Simulation and comparison experiments verify the advantages of the proposed method.Based on the previous fusion modeling method,a recursive least square algorithm is introduced and the method of on-line calculation of formation drillability is proposed.It is easy for the proposed method to capture and predict the abrupt change of formation drillability in the complex geological drilling process.(2)A spatial modeling method of three-dimentioanl(3D)formation drillability field based on geostatistics and machine learning is proposed.Traditional spatial interpolation methods are hard to calculate regional formation drillability data changes caused by formation mutation.Based on this,a 3D spatial model structure of formation drillability field is proposed.The mutual information analysis method is used to determine the spatial correlation between three-dimensional coordinates and formation drillability,and the conclusion that increasing the number of drilling wells can improve the spatial correlation is obtained through comparative experiments.Moreover,four methods of geostatistics and machine learning are introduced for comparative analysis.The application results of an example verify the effectiveness and superiority of the machine learning method in spatial calculation of formation drillability field.The model realizes 3D spatial description of formation drillability information.(3)To achieve high-accuracy prediction of drilling ROP,intelligent prediction methods of drilling ROP based on layered and mixed models are proposed.By further analyzing the direct or indirect relationship between various drilling parameters and drilling ROP,two intelligent high-accuracy prediction methods of ROP based on layered and hybrid model and an improved bat algorithm are proposed.Firstly,a new double-layer intelligent drilling ROP prediction method is proposed.The piecewise cubic Elmite interpolation and mutual information analysis are introduced to match drilling data and reduce the coupling between drilling parameters,respectively.The first layer formation drillability sub-model is established by using the fusion modeling method proposed previously,and the second layer ROP sub-model is established by using the improved particle swarm optimization with radial basis function neural network.The application results show that multi-level and multi-model collaborative description is a new way to realize drilling ROP modeling.How to optimize the hyper-parameters of machine learning model is the key to predict drilling ROP accurately.An improved bat algorithm based on iterative local search and random inertia weight is proposed in the modeling and optimization process.Three types of experiments(traditional test set,standard test set of IEEE CEC2005 and real-world problems)verify the efficiency of the proposed algorithm.A hybrid support vector regression-based drilling ROP prediction method is proposed,and the hyper-parameters of the support vector regression model are optimized by the improved bat algorithm.Wavelet filtering and mutual information analysis are introduced to reduce data noise and coupling between drilling parameters.Compared with eleven well-known ROP prediction algorithms,the experimental results show that the proposed method achieves high-accuracy ROP prediction.(4)A hybrid bat algorithm is proposed to solve the problem of drilling ROP optimization with non-convex model and non-linear constraints.Traditional deterministic or heuristic optimization algorithms are easy to fall into local optimum when facing non-convex and non-linear drilling rate optimization problems.Based on the improved bat algorithm proposed previously,a hybrid bat algorithm is proposed by introducing directional echo localization and improving local random step strategy,which can further improve the global search ability and the stability of optimization solution.The results are compared with ten well-known optimization algorithms in the test set function experiment of the standard IEEE CEC2005.The effectiveness of the proposed algorithm is validated;a new framework of drilling rate optimization based on hybrid bat algorithm is proposed.The above algorithm is applied to drilling ROP optimization,taking into account various equality or non-linear inequality constraints such as formation condition change,bit wear,and so on.The experimental results are compared with four well-known drilling rate optimization algorithms.The advanced nature of the proposed method is presented to provide recommended values of drilling operational parameters for ensuring the efficient operation of the drilling process.By studying the intelligent modeling and drilling ROP prediction and optimization of complex formation drillability field,new theoretical technologies and solutions are provided for the modeling and optimization of complex geological drilling process,and the technological commanding point of modeling and optimization of drilling process is seized,which will lay an important foundation for improving the technology level of drilling process control in China and realizing safe and efficient intelligent control in complex geological drilling process.