摘要
井漏是钻井中常见的复杂情况,影响因素众多且漏失机理复杂。传统方法识别井漏的准确率不高且时效性较差,而人工智能技术能够很好地解决多参数、非线性的复杂问题。为提高井漏风险诊断准确率和效率,提出了一种利用人工智能模型快速诊断井漏的方法。为减少输入维度提高模型的计算效率,利用相关性分析和经验知识优选出总池体积、进出口流量差和立管压力等7种井漏表征参数,基于随机森林、支持向量机、BP神经网络和逻辑回归4种机器学习算法分别建立井漏智能诊断模型并对模型进行优化。研究结果表明,随机森林模型的表现效果最好,其在测试集上对井漏识别的准确率达到98%。此外相对重要性分析表明,总池体积、立管压力、进出口流量差、钻井液密度和大钩载荷是准确诊断井漏风险的主控参数。研究结果对高效准确识别井漏风险,提高钻井效率具有重要意义。
Lost circulation is a common complexity during drilling,and it has many influencing factors and complicated mechanisms.The traditional method to identify lost circulation is not accurate or timely,while the artificial intelligence technology can solve the multi-parameter and nonlinear complex problems.In order to improve the accuracy and efficiency of diagnosing the lost circulation,a method by using artificial intelligence model to fast diagnose the lost circulation was proposed.In order to reduce the input dimension and improve the model calculation efficiency,7 parameters to characterize lost circulation were selected by using correlation analysis and empirical knowledge,such as total pool volume,flow difference between inlet and outlet and riser pressure.Based on four machine learning algorithms,namely random forest,support vector machine,BP neural network and logistic regression,the intelligent diagnosis model for lost circulation was established and optimized.The results showed that the random forest model had the best performance,with a high accuracy rate of identifying lost circulation up to 98%on the test set.In addition,the relative importance analysis showed that the total pool volume,riser pressure,flow difference between inlet and outlet,drilling fluid density and hook load are the main controlling parameters for accurately diagnosing the risk of lost circulation.The research results are significant for fast and accurately identify the risk of lost circulation and improve the drilling efficiency.
作者
陈凯枫
杨学文
宋先知
陈冬
张伟
韩亮
邢星
Chen Kaifeng;Yang Xuewen;Song Xianzhi;Chen Dong;Zhang Wei;Han Liang;Xing Xing(PetroChina Tarim Oilfield Company;China University of Petroleum(Beijing))
出处
《石油机械》
北大核心
2022年第11期16-22,共7页
China Petroleum Machinery
基金
国家科技重大专项“塔里木盆地重点风险探井试油(含储层改造)工程配套技术攻关”(2016ZX05051-3)。
关键词
井漏
智能诊断
工程录井数据
机器学习
相关性分析
随机森林
lost circulation
intelligent diagnosis
engineering logging data
machine learning
correlation analysis
random forest
作者简介
第一作者:陈凯枫,助理工程师,生于1995年,2017年毕业于中国石油大学(北京)石油工程专业,现从事钻井工作。地址:(841000)新疆库尔勒市。E-mail:chenkf-tlm@petrochina.com;通信作者:宋先知,E-mail:songxz@cup.edu.cn。
