An Online Modeling Method for Formation Drillability Based on OS-Nadaboost-ELM Algorithm in Deep Drilling Process *

作者：Chao Gan; Weihua Cao; Min Wu; Xin Chen; Yule Hu; Guojun Wen; Hui Gao; Fulong Ning; Huafeng Ding

来源：IFAC-PapersOnLine, 2017, 50(1): 12886-12891.

• Formation Drillability
• Deep Drilling
• Online Learning
• Extreme Learning Machine
• Nadaboost Algorithm

To achieve safety, high quality, and efficiency in deep drilling, it is necessary to get formation drillability around the borehole during drilling-trajectory planning and intelligent drilling control. Since the drilling data have the characteristics of low value density and noise in the process of deep drilling, it is difficult to model formation drillability in deep drilling. In this paper, a new online modeling method for formation drillability based on online sequential nadaboost extreme learning machine (OS-Nadaboost-ELM) algorithm has been proposed. Firstly, the well logging parameters are chosen as the inputs of the model, whose output is formation drillability. Then, several ELM models are established and the outputs of these models are as weak learners. Then the weak learners are combined by Nadaboost algorithm in order to get a strong learner. Finally, the recursive least squares algorithm is used to adjust the model. The numerical test results show that, in both prediction accuracy and training efficiency aspects, the proposed method is better than other prediction methods such as multiple regression, gray method, back-propagation neural networks, Nadaboost extreme learning machine and online sequential extreme learning machine. Thus the prediction model serves as the online geological model to develop intelligent drilling systems.

• 出版日期2017-7