«Automation and Informatization of the fuel and energy complex»

Application of natural language processing algorithms for automatic classification of drilling operations

UDC: 681.5:622.24
DOI: 10.33285/2782-604X-2022-8(589)-35-42

Authors:

¹ Gazprom EP International B.V, St. Petersburg, Russia
² National University of Oil and Gas "Gubkin University", Moscow, Russia

Keywords: well construction, mathematical model, correlation coefficient, recurrent neural networks, drilling operations, automation

Annotation:

Every operation fulfilled during the process of well construction, is classified in accordance with a standard admitted within a company-operator. Determination of drilling operations categories is required, in particular, to reveal drawbacks in the initial working plan and their correction, while constructing subsequent wells, which is eventually required for proper distribution of the company budget. However, the human factor plays a significant role. In many cases, specialists classify one and the same operation differently. The article presents a mathematical model built on recurrent neural networks which automatically assigns operations to one or other category, based on their free textual description. This approach allows complete elimination of opinions subjectivity as well as considerable time reduction for daily drilling reports preparation for supervisors in the fields. Applicability of the model is assessed with statistical parameters including coefficients of Matthews correlation (0,88), Cohen’s kappa (0,88) and ROC-AUC (0,95) that demonstrate high accuracy.

Bibliography:

1. Svedeniya o stroitel'stve skvazhin na neft' i gaz (forma N 1-TEK (bur)). – 2021. – URL: http://www.consultant.ru/document/cons_doc_LAW_52009/fb5ed26b139141a9951ed91e25e899042ed9c03f/
2. Automated system for drilling operations classification using statistical features / B. Esmael, A. Arnaout, R.K. Fruhwirth, G. Thonhauser // 11 Int. Conf. on Hybrid Intelligent Systems, HIS 2011, Melacca, Malaysia, Dec. 5–8. – 2011. – P. 196–199. – DOI: 10.1109/HIS.2011.6122104
3. Intelligent Real-time Drilling Operations Classification Using Trend Analysis of Drilling Rig Sensors Data / A. Arnaout, R.K. Fruhwirth, B. Esmael, G. Thonhauser // SPE Kuwait Int. Petroleum Conf. and Exhibition, Kuwait City, Kuwait, Dec. 10–12. – 2012. – DOI: 10.2118/163302-MS
4. Bengio Y., Simard P.Y., Frasconi P. Learning long-term dependencies with gradient descent is difficult // IEEE Transactions on Neural Networks. – 1994. – Vol. 5, Issue 2. – P. 157–166. – DOI: 10.1109/72.279181
5. Hochreiter J. Untersuchungen zu dynamischen neuronalen Netzen. – Munchen, 1991. – 74 p.
6. Hochreiter S., Schmidhuber J. Long Short-Term Memory // Neural Computation. – 1997. – Vol. 9, Issue 8. – P. 1735–1780. – DOI: 10.1162/neco.1997.9.8.1735
7. Olah C. Understanding LSTM Networks. – URL: https://colah.github.Io/posts/2015-08-Understanding-LSTMs
8. Embeddings: Motivation From Collaborative Filtering. – URL: https://developers.google.com/machine-learning/crash-course/embeddings/motivation-from-collaborative-filtering
9. Howard J., Gugger S. NLP Deep Dive: RNNs // Deep Learning for Coders with fastai & PyTorch / Ed. by Howard J., Gugger S. – Sebastopol, California: O'Reilly Media, Inc., 2020. – Chapter 10.
10. Matthews correlation coefficient // Scikit-learn. Machine Learning in Python. – URL: https://scikit-learn.org/stable/modules/model_evaluation.html#matthews-corrcoef
11. Cohen’s kappa score // Scikit-learn. Machine Learning in Python. – URL: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.cohen_kappa_score.html#sklearn.metrics.cohen_kappa_score
12. Precision, recall and F-measures // Scikit-learn. Machine Learning in Python. – URL: https://scikit-learn.org/stable/modules/model_evaluation