自适应粒子群反演算法在阵列侧向测井严重失真井段的应用  

Application of adaptive particle swarm optimization inversion algorithm in array lateral logging with serious distortion

作　　者：倪小威 徐思慧 侯少勇 赵庆广 张玮[1] 王彦秋[1] 熊昶 刘迪仁[2] NI Xiaowei;XU Sihui;HOU Shaoyong;ZHAO Qingguang;ZHANGWei;WANG Yanqiu;XIONG Chang;LIU Diren(Tarim Oilfield Company,PetroChina,Korla 841000,China;Geophysics and Petroleum Resources Institute of Yangtze University,Wuhan 434000,China)

机构地区：[1]中国石油塔里木油田分公司,新疆库尔勒841000 [2]长江大学地球物理与石油资源学院,湖北武汉434000

出　　处：《复杂油气藏》2022年第3期44-50,共7页Complex Hydrocarbon Reservoirs

基　　金：国家重点研发计划项目“地下及井中地球物理勘探技术与装备”(2018YFC0603300)。

摘　　要：阵列侧向测井往往因为环境因素的影响而不能准确反映储层真实电阻率信息,需要进行反演处理。在经典粒子群算法的基础上引入混沌序列优化粒子群算法初始种群生成方式,加强初始种群多样性,同时对惯性权重等参数进行自适应改进,平衡其反演过程中的局部收敛和全局寻优能力。在反演之前预先建立阵列侧向测井正演响应数据库,在反演过程中直接插值调用,三次样条插值与严格正演计算相对误差在5%以内,可在保证反演精度前提下加快反演速度。实际资料处理表明,在阵列侧向测井资料严重失真井段,基于自适应粒子群算法的反演方法可有效恢复泥浆侵入剖面信息,反演地层真实电阻率与随钻电阻率基本一致,同时基于反演地层电阻率相较视电阻率、经典粒子群算法反演地层电阻率计算得到的含水饱和度降低了53.4%,结果更加符合MDT取样结论。Because the frequently environmental influences prevent the real resistivity information of the reservoir from being accurately reflected,the array logging requires inversion processing.Based on the classic particle swarm optimization,the chaotic sequence is introduced to improve the particle swarm optimization algorithm's initial population generation technique and increase the diversity of the initial population.The inversion process's ability to balance local convergence and global optimization is done at the same time by adaptively improving parameters such as inertia weight.Before the inversion,the array lateral logging forward response database is established and the interpolation is then called immediately during the inversion procedure.Under the premise that the inversion accuracy is guaranteed,the relative error between cubic spline interpolation and strict forward calculation is less than 5%,which can accelerate the inversion speed.The actual data processing demonstrates that the adaptive particle swarm algorithm-based inversion method can successfully restore the mud invasion profile information in the well section with serious distortion of the array lateral logging data.The actual resistivity of the inversion formation is basically consistent with the resistivity while drilling.And then the water saturation calculated by the inversion formation resistivity is 53.4%lower than that calculated by apparent resistivity and the inversion of the formation resistivity by the classical particle swarm algorithm,which is more in line with the MDT sampling conclusion.

关 键 词：粒子群算法 阵列侧向测井 严重失真 自适应惯性权重 

分 类 号：P631[天文地球—地质矿产勘探]