基于井下参数的SCNGO-SVM卡钻预警方法研究  

Research on SCNGO-SVM Sticking Warning Method Based on Downhole Parameters

作　　者：张涛 夏鹏[1,2,3] 李军[4] 王彪 詹家豪[4] Zhang Tao;Xia Peng;Li Jun;Wang Biao;Zhan Jiahao(Beijing Information Science&Technology University;Beijing Key Laboratory of High Dynamic Navigation Technology,Beijing Information Science&Technology University;Key Laboratory of Modern Measurement&Control Technology,Ministry of Education;College of Petroleum Engineering,China University of Petroleum(Beijing))

机构地区：[1]北京信息科技大学自动化学院 [2]北京信息科技大学高动态导航技术北京市重点实验室 [3]现代测控技术教育部重点实验室 [4]中国石油大学(北京)石油工程学院

出　　处：《石油机械》2025年第1期20-27,36,共9页China Petroleum Machinery

基　　金：国家自然科学基金重大科研仪器项目“钻井复杂工况井下实时智能识别系统研制”(52227804);国家自然科学基金面上项目“底部钻具高频扭转振动响应机理及识别方法研究”(52274003);国家自然科学基金青年科学基金项目“干热岩储层双重介质射孔簇内复杂多裂缝起裂及扩展机理研究”(52104001)。

摘　　要：针对卡钻风险预测的问题,提出了一种融合正余弦和折射反向学习的北方苍鹰优化算法(SCNGO)和支持向量机(SVM)的卡钻预警模型。针对北方苍鹰优化算法(NGO)容易陷入局部最优以及初始解的分布具有随机性和非均匀性的特性,引入折射反向学习策略初始化北方苍鹰算法个体、正余弦策略替换原始苍鹰算法的勘察阶段的位置更新公式和正余弦策略的步长搜索因子进行改进,将SCNGO用于SVM寻参,并将模型SCNGO-SVM应用于卡钻预警。研究结果表明:SCNGO在收敛速度、寻优精度等方面明显优于NGO、WOA(鲸鱼优化算法)及SSA(麻雀优化算法);该卡钻预警模型对于卡钻的预测准确率高达97.33%,相较于WOA-SVM、NGO-SVM、SSA-SVM卡钻预警模型,在预测准确率和运算速度上均有较大的提升。该模型为卡钻的预测及其工程应用提供了理论指导。As to sticking risk prediction,a SCNGO-SVM sticking warning model was proposed.It integrates the sine cosine and refracted opposition-based learning Northern Goshawk Optimization(SCNGO)with support vector machine(SVM).Considering that the Northern Goshawk Optimization(NGO)may fall into local optima and yield initial solutions in random and non-uniform distribution,the refracted opposition-based learning strategy was introduced to initialize the individuals of NGO.Meanwhile,the sine cosine strategy was introduced to replace the position update formula of NGO in the survey phase,and the step size search factor of the sine cosine strategy was used to improve the NGO,thus forming SCNGO.The SCNGO was used for parameter search of SVM.Finally,the SCNGO-SVM model was used to conduct sticking warning.The research results show that the SCNGO is obviously better than the NGO,whale optimization algorithm(WOA)and sparrow search algorithm(SSA)in terms of convergence rate and optimization accuracy.The SCNGO-SVM sticking warning model achieves an accuracy of 97.3333%,and is significantly higher in prediction accuracy and operating speed than WOA-SVM,NGO-SVM and SSA-SVM models.The SCNGO-SVM method provides a theoretical guidance for sticking prediction and its engineering applications.

关 键 词：卡钻预警模型 北方苍鹰优化算法 性能测试 折射反向学习策略 正余弦策略 

分 类 号：TE28[石油与天然气工程—油气井工程]