一种基于PSO-ELM的低渗透砂岩水淹层测井识别方法  

A logging identification method of water-flooded layer of low permeability sandstone based on PSO-ELM

作　　者：杨波[1] 黄长兵 何岩 李垚银 李路路[1] YANG Bo;HUANG Changbing;HE Yan;LI Yaoyin;LI Lulu(Xinjiang Oilfield Company,PetroChina,Karamay 834000,China;Key Laboratory of Solid Waste Treatment and Resource Recovery of Ministry of Education,Southwest University of Science and Technology,Mianyang 621010,China)

机构地区：[1]中国石油新疆油田公司陆梁油田作业区,新疆克拉玛依834000 [2]西南科技大学固体废物处理与资源化教育部重点实验室,四川绵阳621010

出　　处：《断块油气田》2024年第4期645-651,共7页Fault-Block Oil & Gas Field

基　　金：国家科技重大专项“准噶尔盆地致密油开发示范工程”(2017ZX05070)。

摘　　要：水淹层测井识别对油田开发方案部署及提高采收率有着重要意义。新疆陆梁油田作业区某区块油层水淹类型主要为污水水淹,测井响应特征复杂多变,传统识别图版方法难以对水淹层有效识别。文中基于测井、地质、试油等资料,在水淹层测井响应特征分析基础上,提出了一种利用改进粒子群优化算法(Particle Swarm Optimization,PSO)及极限学习机(Extreme Learning Machine,ELM)的水淹层识别方法。首先,利用相关系数优选6个主控因素:RD,RS,GR,SP,DEN,AC。其次,采用改进粒子群算法对极限学习机模型进行参数寻优;最后,利用优化后的模型对研究区水淹层进行预测。结果表明,利用PSO-ELM模型识别水淹层,识别符合率达到91.7%,应用效果优于ELM模型及传统识别图版,为水淹层测井识别提供了新思路。Logging identification of water-flooded layer is of great significance to the deployment of oilfield development scheme and the enhancement of oil recovery.The flooding type of reservoir in a block of Luliang Oilfield in Xinjiang is mainly sewage flooding,and the logging response characteristics are complex and varied,it is difficult to identify the water-flooded layer effectively by traditional chart identification method.Based on the data of logging,geology,oil testing,and on the basis of logging response characteristics analysis of water-flooded layer,this paper proposed a new water-flooded layer identification method using improved Particle Swarm Optimization(PSO)algorithm and Extreme Learning Machine(ELM).First,correlation coefficient was used to select six main controlling factors,RD,RS,GR,SP,DEN and AC.Secondly,the improved Particle Swarm Optimization was used to optimize the parameters of the ELM model.Finally,the optimized model was used to predict the water-flooded layer in the study area.The results show that the identification coincidence rate of water-flooded layer can reach 91.7%by using PSO-ELM,which is better than the application effect of ELM and traditional identification chart,providing a new idea for logging identification of water-flooded layer.

关 键 词：相关系数 粒子群优化算法 极限学习机 水淹层识别 

分 类 号：TE319[石油与天然气工程—油气田开发工程]