摘要： 储层可压性评价是提高非常规油气压裂均衡改造效果的先决条件之一.目前储层可压性评价主要依赖测井数据理论解释岩石力学参数,应用效果不均衡.本文利用钻头破岩数据直接反映岩石力学参数的特点,以钻录井和测井数据驱动聚类储层可压性,建立了基于SOM无监督聚类算法的储层可压性聚类模型,手肘法确定最优聚类数,形成了压裂布缝位置参数优化方法.针对塔里木盆地巨厚储层典型直井,开展了三簇射孔布缝位置优选设计.结果表明,钻井钻时、dc指数、钻压、扭矩和测井地层电阻率、声波时差和中子等参数与储层可压性显著相关,可作为特征参数;所建立的模型可有效区分储层可压性沿井筒轴向的差异性,优选同类别储层可压性井段布置裂缝,有望提高均衡压裂改造效果.

关键词: 机器学习, 无监督学习, 水力压裂, 优化设计

Abstract: Reservoir fracability evaluation is one of the prerequisites to improve the effect of balanced fracturing of uncon-ventional oil and gas fields.At present,reservoir fracability evaluation mainly depends on logging data theory to explain rock mechanics parameters,and the application effect on fracturing is uneven.In this paper,the characteristics of rock mechanical parameters are directly reflected by the bit rock breaking data and the reservoir fracability is clustered by drilling and logging data.We established a reservoir fracability clustering model based on a self-organizing map(SOM)unsupervised clustering algorithm.The elbow method is used to determine the optimal clustering number,and the parameter optimization method of fracture placement is formed.The optimal design of three-cluster perforation placement is carried out for typical vertical wells in the Tarim Basin with large thickness reservoirs.The results show that the drilling time,dc-exponent,weight on bit,torque,true formation resistivity,acoustic and neutron data are significantly correlated with reservoir fracability and can be used as character-istic parameters.The established model can effectively distinguish the difference of reservoir fracability along the wellbore axis,and select the fractures in the fracturable well section of the same type of reservoir,which is expected to improve the effect of balanced fracturing.

Key words: machine learning, unsupervised learning, hydraulic fracturing, optimization design