Drilling well risk prediction method based on Markov chain and Bayesian network

Abstract

The invention discloses a drilling well risk prediction method based on a Markov chain and a Bayesian network, which employs the Markov chain and Bayesian network to perform comparatively comprehensive analysis and prediction on drilling risks from vertical and horizontal aspects, and overcomes the deficiency of treating lack of indicators by the Markov chain. The Markov chain is a vertical prediction method to be used for detecting variable probability distribution determined by samples in future time. The Bayesian network is a horizontal prediction method and displays an index mutual influence relation. The combination of horizontal prediction and vertical prediction methods can solve the problem of lack of non-underlying index data of a multilayer index system to realize risk prediction of macroscopical meaning. The backstepping function of the Bayesian network also provides a basis for risk control.

Description

technical field [0001] The invention specifically relates to a drilling risk prediction method based on Markov chain and Bayesian network. Background technique [0002] Drilling operation risk prediction refers to the use of certain methods to predict the risks existing in drilling operations based on previous drilling operation data, so as to achieve the purpose of prevention and control. In recent years, experts in the field of petroleum and mathematics have carried out in-depth research on the drilling risk prediction problem by using sequential Gaussian co-simulation, fuzzy expert system, case reasoning technology, support vector machine and ARIMA model and other research methods. However, there are still very few articles on the application of Bayesian networks in oil drilling, and the research is still at a relatively new stage. [0003] Drilling operation has a very complicated process, and there are many uncertain factors in the process. Therefore, it is very impor...

AI Technical Summary

Problems solved by technology

Maria established a maritime safety management model based on Bayesian networks; Ozan proposed an evolutionary Monte Carlo method to train Bayesian networks for time series forecasting [18]. These two articles use Bayesian networks for risk prediction. The disadvantage is that Bayesian networks are based on probability forecasting method, the period forecast result is a probability distribution

Images