基于机器学习的钻井工况识别技术现状及发展

Recent developments and future trends of drilling status recognitiontechnology based on machine learning

配备传感器的现代钻井设备带来了持续不断的实时钻井数据,通过监测这些钻井数据可以对钻井工况进行及时有效的判断,进而提高钻井效率,降低钻井成本和钻井事故率。由于钻井的复杂性和不可预知的作业条件,现有的通过数据采集系统执行的钻井工况识别系统容易出现较高的误报率。为了解决高误报率问题,实现从高维钻井数据中得到高精度、高效率的钻井工况识别结果,基于机器学习算法的识别模型被开发,并在应用中表现出了显著的有效性和稳定性。文章简述了机器学习的发展历程和项目流程,介绍了钻井系统参数,描述了支持向量机、BP神经网络、随机森林和深度学习等机器学习分类算法在钻井工况识别技术中的应用现状,对比研究了七个机器学习工况识别模型的框架、超参数、特征参数以及识别性能,并探讨了基于机器学习算法的钻井工况识别技术发展趋势,为实现钻井设备的自动化和钻井工程的智能化提供一些新的思路。

Modern drilling equipment equipped with sensors brings continuous real-time drilling data.By monitoring these drilling data,drilling status can be judged timely and effectively,thereby improving drilling efficiency,reducing drilling costs and drilling accident rates.Due to the complexity of drilling and unpredictable operating status,the existing drilling status recognition system through data acquisition system is prone to high false alarm rates.In order to solve the problem of high false alarm rates and achieve high-precision and high-efficiency drilling status recognition results from high-dimensional drilling data,recognition models based on machine learning algorithms have been developed and have shown remarkable effectiveness and stability in application.In this paper,the development process and project process of machine learning was briefly described,the parameters of the drilling system were introduced,and the application status of machine learning classification algorithms such as Support Vector Machine,BP Neural Network,Random Forest and Deep Learning in drilling status recognition technology was described.The framework,hyperparameters,characteristic parameters and recognition performance of seven machine learning status recognition models were compared and studied,and the development trend of drilling status recognition technology based on machine learning algorithm was discussed,which can provide some new ideas for realizing the automation of drilling equipment and the intelligence of drilling engineering.