Intelligent Modeling Of Wellbore-Stability Characteristics Parameters In Complex Geological Conditions

Ensuring the safety of the supply of resources and energy is vital to economic and social development,it is also an important part of national security because there is still a great demand for resources and energy.Deep geological exploration is inevitable consider the intricate nternational relations.However,there is a complex geological environment including sensitive formation and complex geomechanical pressure system,which would results in wellbore instability,and makes the drilling cost too high.Achieving safe and efficient drilling engineering requires establishing accurate geological environment model to describe the underground geological environment intuitively,and then analysing wellbore stability.The characteristic parameters after analysis mainly include formation lithology,pore pressure,and rock mechanics parameters.Formation lithology describes the physical and chemical properties of rocks,some water-sensitive formations and water-soluble formations are more prone to collapse compared with other lithology,and abnormal pore pressure is easily associated with well instability.Rock mechanics parameters are also key parameters for wellbore stability analysis,and the greater formation strength would make a safer drilling process.Based on the above analysis,this thesis focuses on the research of intelligent modeling of wellbore-stability characteristics parameter in complex conditions.The main research results of the thesis are as follows:(1)A Lithology identification ensemble method is proposed considering the characteristic of data distribution.Lithology identification is significant for wellbore stability analysis.Different thickness of rock leads to obvious imbalance characteristics.In addition,the similar geological characteristics of formation leads to data-overlapping,which are easily confused during lithology identification.Therefore,in view of the unbalanced characteristics and the overlapping of logging data,a Lithology identification ensemble method considering the characteristic of data distribution is proposed,the method utilizes relatively small number of base classifiers,splits a multi-class problem into simpler binary class problems,and then a minority class would be integrated into other classes,to relieve data imbalance and data overlapping at the same time.It can greatly improve the lithology identification accuracy and efficiency.(2)A hybrid depth series analysis-based real time pore pressure estimation method for complex geol ogical drilling processAccurate real-time estimation of pore pressure is essential for the geomechanical analysis of wellbore stability.Aiming at the problem of data noise,strong nonlinearity and coupling between drilling parameters,a data-driven pore pressure estimation method is developed on the basis of depth series analysis.First,concerning the existence of outliers and noises,an outlier detection and wavelet filtering algorithm are introduced to obtain reliable model parameters.Additionally,aiming at the problem of data coupling and strong nonlinearity,the intermediate parameter is calculated on the basis of mechanism model analysis.Then Pearson correlation-analysis is employed to determine strongly correlated attributes with pore pressure in the data preprocessing stage.Afterward,an online principal component analysis similarity method is proposed for depth series segmentation,considering the varying drilling depth.Finally,a real-time data-driven pore pressure estimation model that integrates conventional empirical methods is established on the basis of depth segmentation,and a model switching strategy is further developed and will be activated when performance deteriorates.A real case study is conducted using actual data from a drilling site in Utah.The results indicate the e ectiveness of the proposed method compared with eight well-known conventional methods both in online and offlinene condition.It can provide timely pore pressure information for wellbore stability analysis.(3)A hybrid spatial model based on identified pressure conditions for 3D pore pressure estimationAccurate spatial modeling of pore pressure helps to safe drilling plans,performance analysis,and efficient reservoir modeling.Considering the varying drilling depth and multiple pore pressure modes of eight study wells,the strategy of dividing similar pressure subspaces before establishing a three-dimensional pore pressure model is determined,and a hybrid spatial modeling method is developed based on identified pressure conditions to estimate the pore pressure of an area.First,an appropriate number of pressure conditions is determined adaptively based on quantization error modeling,and similar pressure conditions are identified based on a fuzzy c-means method.Additionally,a kriging interpolation algorithm is introduced to produce profiles of different pressure subspaces.Afterward,random forest submodels are built separately according to the pressure profiles.In the last part,all of these submodels coalesced into one whole model.The results indicate that the proposed method can better capture the local variations in pore pressure,and can effectively improve the modeling accuracy comparing with other five well-known methods.The proposed method can realizes 3D spatial description of pore pressure and can provide extensive and continuous regional pore pressure information for wellbore stability analysis.(4)Semi-supervised support vector regression based on data similarity and its application to rock-mechanics parameters estimationRock-mechanics parameters are critical indexes in wellbore analysis and resource exploration.However,the number of these parameters is always insufficient and discontinuous because only a few cores are measured and labeled in laboratory.To improve prediction performance of these parameters,a novel semi-supervised support vector machine method that takes into account data similarity is devised by leveraging unlabeled data,which is obtained continuously in drilling process.Aiming at the existence of outliers,an outliers deletion algorithm is developed for a better similarity comparison.Then the training datasets are selected by measuring the similarity between labeled data and tested data to improve prediction performance.Finally,the labeling confidence evaluation was employed to choose appropriate labeled data to train the model.Eight classical supervised regression methods and semi-supervised regression methods were employed as performance evaluation benchmarks for our semi-supervised method.The application results verify the superiority of the proposed method.It provides basis for wellbore stability analysis.By analyze the characteristic parameters that affect the wellbore stability and studying the intelligent modeling method of wellbore-stability characteristics parameters in complex conditions to ensure that the borehole does not collapse,break,or shrink is the basis of drilling process control and trajectory optimization design,which will lay an important foundation for achiving safe and efficient intelligent control of complex geological drilling process.