摘要
录井解释过程中,由于数据特征维度较高和数据集成化能力不足等,需要人工校正录井综合图的岩性剖面。提出一种基于数据挖掘的录井剖面归位解释处理方法,对选取的录井解释数据进行数据清洗,选取有效影响因子,进行基于主成分分析的特征降维;分析录井解释数据特点,选择神经网络算法作为模式挖掘模型;对模式挖掘模型进行表达与解释,将实验获得的最优特征识别模式运用于实际数据。结果表明,采用模式挖掘模型在未知区块的平均识别准确率接近于92%,模型泛化能力相对稳定,对部分常用的岩性的识别准确率接近于95%,与多次人工校正后的归位结果相近。
In the process of logging interpretation,due to the high dimension of data characteristics and the lack of data integration ability,it is necessary to manually correct the lithologic section of the comprehensive logging map.This paper puts forward a method of logging profile homing and interpretation based on data mining.It cleans the selected logging interpretation data,selects effective influencing factors,and carries out feature dimension reduction based on principal component analysis;analyzes the characteristics of logging interpretation data,selects neural network algorithm as the pattern mining model;expresses and interprets the pattern mining model,and obtains the best result from the experiment.Feature recognition pattern is applied to actual data.The results show that the average recognition accuracy of the pattern mining model in the unknown block is close to 92%,the generalization ability of the model is relatively stable,and the recognition accuracy of some commonly used lithology is close to 95%,which is close to the result of multiple manual correction.
作者
李春生
彭嘉伟
胡亚楠
张可佳
刘涛
崔翔宇
曹琦
LI Chunsheng;PENG Jiawei;HU Yanan;ZHANG Kejia;LIU Tao;CUI Xiangyu;CAO Qi(School of Computer and Information Technology,Northeast Petroleum University,Daqing,Heilongjiang 163318,China)
出处
《东北石油大学学报》
CAS
北大核心
2020年第4期99-104,112,I0008,共8页
Journal of Northeast Petroleum University
基金
国家自然科学基金项目(51774090)
黑龙江省自然科学基金项目(F2015020)
黑龙江省教育厅科研专项引导性创新基金项目(2017YDL-12)
黑龙江省教育规划重大项目(GJ20170006)。
关键词
数据挖掘
神经网络
岩性识别
岩心归位
录井解释
data mining
neural networks
lithology identification
core homing
logging interpretation
