摘要
钻井工程监测参数与井下复杂事故之间存在相关性,故基于SMOTE欠采样的随机森林方法建立钻井工程风险评估监测模型,以此起到规避风险的目的。首先,需要将随钻数据进行筛选与挖掘,根据调研结果确定特征参数,做好建模前的预备工作;其次,利用随机森林算法建立模型;最后,在考虑监测参数与井下复杂事故联系和交叉验证的基础上调整模型参数,计算卡钻事故的风险概率。实验结果表明,基于SMOTE欠采样的随机森林钻井工程风险评估方法预测值与现场实际结果基本吻合。
Due to a correlation between the monitoring parameters of drilling engineering and complex downhole accidents,a monitoring model for drilling engineering risk assessment should be established based on the SMOTE under sampling random forest method to avoid risks.First of all,it is necessary to screen and analyze the data while drilling,determine the characteristic parameters according to the survey results,and make preparations before modeling.Secondly,the random forest algorithm is used to build the model.Finally,on the basis of considering the correlation between monitoring parameters and complex downhole accidents and cross validation,the model parameters are adjusted to calculate the risk probability of sticking accidents.The experimental results show that the predicted value is basically consistent with the actual field results.
作者
易思琦
魏凯
YI Siqi;WEI Kai(School of Petroleum Engineering,Yangtze University,Wuhan 430100,Hubei,China)
出处
《石油地质与工程》
CAS
2023年第4期100-103,共4页
Petroleum Geology and Engineering
关键词
随机森林
钻井工程
风险评估
卡钻预测
random forest
drilling engineering
risk assessment
sticking predictio
