大洋钻探过程钻速在线区间预测方法——以微型钻探船室内模拟实验为例

An online interval prediction method for rate of penetration （ROP） during the ocean drilling： A case study on the indoor simulation experiments by micro drilling ship

摘要:

大洋钻探是从事海洋能源资源勘探开发和地壳构造演化研究的主要手段，常面临复杂海况扰动大、海底地层不确定性强等问题。本文提出一种大洋钻探过程钻速在线区间预测方法，以微型钻探船室内模拟实验为例开展方法的验证工作，为工程化应用奠定重要基础。首先，运用数据重采样、数据时深匹配、数据滤波等方法对多源大洋钻探数据进行预处理。其次，引入极限学习机、粒子群优化等方法建立钻速点预测模型。再者，利用非参数估计方法构造置信区间，建立钻速区间预测模型，并开展钻速区间预测。最后，通过滑动窗口方法在线更新区间预测模型参数，实现模型的在线学习和优化。微型钻探船室内模拟实验的对比结果验证了所提方法具有很强的钻速预测能力和鲁棒性，为大洋钻探过程钻速优化控制提供了新的工程解决方案。

Abstract:

Ocean drilling is a major means for the exploration and development of marine energy resources and the study of crustal tectonic evolution， which is often faced with problems such as large disturbances in complex sea state and strong uncertainty in seafloor formation. This paper proposes an online interval prediction method for ROP during the ocean drilling process， and carry out the validation of the method with the indoor simulation experiment of micro drilling ship as an example， so as to lay an important foundation for the engineering application. Firstly， methods such as data resampling， data time-depth matching and data filtering are applied to pre-process the multi-source ocean drilling data. Secondly， methods such as Extreme Learning Machine （ELM） and Particle Swarm Optimization （PSO） are applied to establish a point prediction model of ROP. Furthermore， nonparametric estimation method is utilized to construct confidence intervals， establish ROP interval prediction model， and carry out ROP interval prediction. Finally， the interval prediction model parameters are updated online by moving window to realize online learning and optimization of the model. The comparative results of the indoor simulation experiments on the micro drilling ship verify that the proposed method has strong ROP prediction capability and robustness， and can provide a new engineering solution for the optimization and control of ROP during the ocean drilling process.