基于PSO-BP的岩性识别方法研究
Research on Lithology Identification Method Based on PSO-BP
高雅田 1杨俊国1
作者信息
- 1. 东北石油大学计算机与信息技术学院 大庆 163318
- 折叠
摘要
近些年来,数据分析、深度学习技术取得了长足的发展,并为社会带来了可观的收益.故利用深度学习手段进行岩性识别也成为了一个研究热点.岩性识别是录井解释的核心业务,准确而有效地预测储层性质对石油勘探工作有着重大意义.为解决传统岩性识别方法成本高、耗时长等缺点.论文利用松辽盆地中若干井的测井数据进行模型研究,提出了一种基于PSO-BP的岩性识别方法.通过对测井源数据进行数据预处理、构建网络识别模型、优化岩性识别模型、评价模型输出结果等步骤,实现基于PSO-BP岩性识别方法.经过反复试验,结果表明采用PSO-BP的岩性识别方法对岩性进行识别的平均准确率可达92.2%,为储层预测工作提供了可靠的支撑.
Abstract
In recent years,data analysis and deep learning technology have made great progress and brought considerable ben-efits to the society.Therefore,the use of deep learning method for lithology identification has become a research hotspot.Lithology identification is the core business of logging interpretation,accurate and effective prediction of reservoir properties is of great signifi-cance to petroleum exploration.However,the traditional lithology identification scheme has some disadvantages,such as high cost,long time and so on.Therefore,this paper uses the logging data of some wells in Songliao basin to study the model,after comparing the lithology identification results of different algorithms,a lithology identification method based on PSO-BP is proposed.Through data preprocessing of logging source data,construction of network identification model,optimization of lithology identification mod-el and evaluation of model output,the lithology identification method based on PSO-BP is realized.After repeated tests,the results show that the average accuracy of lithology identification using PSO-BP method can reach 92.2%,which provides a reliable support for reservoir prediction.
关键词
BP神经网络/粒子群优化算法/岩性识别/数据预处理/KNN/支持向量机Key words
BP neural network/PSO/lithology identification/data preparation/KNN/SVM引用本文复制引用
基金项目
东北石油大学校培育基金项目(PY120225)
出版年
2024
国家科技期刊平台
NSTL
万方数据