摘要
天然气水合物是高压低温条件下形成的类冰状结晶的烃类能源,燃值高、储量大,极具开发潜力。基于测井曲线可以识别地层水合物,由于传统的水合物地层识别方法依赖专家判断,效率不高、准确率不高,本文引入多种机器学习算法进行水合物层段识别,根据评价指标准确率、F_(1)分数、精确度与召回率等参数对预测结果进行评估,优选最佳算法。另外,通过将测井参数数目进行优化组合测试,找到不同测井参数对测试结果影响程度,优选最佳参数组合。测试结果表明集成学习算法有较好的评价效果,对未来实现智能化识别水合物有重要意义。
Natural gas hydrate is an ice-like crystalline hydrocarbon energy formed under high-pressure and low-temperature conditions,with high flammability value and large reserves,which has great potential for development.Formation hydrate can be identified based on logging curves.Since traditional hydrate formation identification methods rely on expert judgment with low efficiency and accuracy,this paper introduces various machine learning algorithms for hydrate formation segment identification,evaluates the prediction results based on parameters such as evaluation index accuracy,F_(1) score,precision and recall,and preferably selects the best algorithm.In addition,the optimal combination of the number of logging parameters is tested to find the degree of influence of different logging parameters on the test results,and the best combination of parameters is selected.The test results show that the integrated learning algorithm has a good evaluation effect,which is important for the future realization of intelligent hydrate identification.
作者
叶智慧
宁禹强
张敏
李晓蓉
YE Zhihui;NING Yuqiang;ZHANG Min;LI Xiaorong(College of Safety and Ocean Engineering,China University of Petroleum-Beijing,Beijing 102249,China;Daqing Oilfield Production Technology Institute,Daqing 163453,China;College of Petroleum Engineering,China University of Petroleum-Beijing,Beijing 102249,China)
出处
《海洋技术学报》
2021年第5期51-61,共11页
Journal of Ocean Technology
基金
中国科学院战略性先导科技专项(A类)资助项目(XDA14040402)
中国石油大学(北京)校级基金项目(2462020YXZZ051)。
关键词
天然气水合物
机器学习
地层识别
测井曲线
参数优选
natural gas hydrate
machine learning
formation recognition
logging curve
parameter optimization