Early kick detection using data-driven Bayesian network: model development and experimental testing

Dinh, Nhat Minh (2020) Early kick detection using data-driven Bayesian network: model development and experimental testing. Masters thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (1MB)

Abstract

Safety is one of the keys to success for offshore oil and gas development projects. Drilling operation becomes more and more complicated when moving to deeper water and harsher conditions, which requires a higher level of safety. Kick is an event that happens when the hydrostatic pressure of drilling mud is lesser than the formation pressure which let the formation fluid enter the wellbore. Detection of a kick event as soon as it happens will spare drillers more time to make decision and take necessary actions. An uncontrolled or unaware kick event may lead to a well blowout which causes major damage to infrastructure, could kill people and costs lots of money. The conventional method of kick detection entails monitoring surface parameters such as stand-pipe pressure, mud pit volume, changing in flow rate and other drilling parameters can lead to the delay in detection. Some recent studies have successfully proved the ability to employ downhole parameters to realize kick. The new methods show the robust results in detection and the improvement in detection time. Besides, data-driven Bayesian network (BN) has shown to solve the problem in historical data, which is usually available, unlike expensive, and insufficient, expert knowledge. This work presents the application of data-driven BN using downhole parameters to early kick detection. The work includes three main parts: 1) creating and testing a data-based Bayesian network based on historical experiment data and synthetic data; 2) designing drilling sample, setting up and conducting experiments with the new large drilling simulator (LDS); 3) validating the data-based Bayesian network with the data from the LDS experiment. Upon the success of this work, the developed BN model will serve an efficient and effective way to detect kick early, which will enable appropriate corrective actions. The new setup of experiment with LDS can be used to conduct further experimentation to simulate more complicated kick scenarios during drilling operation.

Item Type: Thesis (Masters)
URI: http://research.library.mun.ca/id/eprint/14449
Item ID: 14449
Additional Information: Includes bibliographical references (pages 70-75).
Keywords: Kick detection, Blowout prevention, Data-driven model, Bayesian model, Process safety, Risk engineering
Department(s): Engineering and Applied Science, Faculty of
Date: May 2020
Date Type: Submission
Library of Congress Subject Heading: Underwater drilling--Simulation methods; Bayesian statistical decision theory.

Actions (login required)

View Item View Item

Downloads

Downloads per month over the past year

View more statistics