Scholarlink

<p>The conventional double-constrained inversion model of horizontal in situ stresses, which considers the constraints from the stress polygon and borehole failure data associated with borehole collapse (breakout) and tensile fracture, fails to adequately capture the impact of geomechanical parameter uncertainty on in situ stress. Therefore, this study considers the uncertainty of geomechanical parameters and Bayesian theory, and reconstructs the conventional double-constrained inversion model into a new Bayesian inversion model (i.e., a quantitative uncertainty probability model) of horizontal in situ stresses for an arbitrarily inclined borehole, which is naturally applicable to vertical wells. By introducing the delayed rejection adaptive Markov chain Monte Carlo (DRAM MCMC) algorithm, the model effectively integrates historical prior distribution information of geomechanical parameters with current observed borehole failure data, thereby further alleviating irregularities in the posterior marginal distribution of the traditional Metropolis–Hastings (MH) algorithm. This study demonstrates that the uncertainty characteristic of Maximum horizontal in situ stress tends to be normally distributed under a non-informative prior. When the minimum horizontal in situ stress obeys the normal prior distribution and is updated by the Bayesian method, the predicted Maximum horizontal in situ stress has lower uncertainty than the case of a uniform prior distribution. However, the prior type of the minimum horizontal in situ stress gradually exerts a weakening influence on the prediction of the maximum horizontal in situ stress with increasing observation error. Furthermore, the proposed Bayesian inversion model shows strong stability when dealing with uncertainties in multiple geomechanical parameters. It effectively converts the uncertainty characteristics of the maximum horizontal in situ stress from an independent joint distribution to a correlated one. This change reduces the risk of wellbore instability by up to 40%, significantly enhancing stress prediction reliability for drilling design and reservoir management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Probabilistic Inversion of Horizontal In Situ Stresses Combining Bayesian Theory and Borehole Failure Data to Quantify the Uncertainty of Wellbore Instability Risk

  • 期刊论文

  • 发表日期：

    2025年9月12日

查看全文
  • Jiajia Gao,
  • Fuzhi Chen,
  • Peiyu Zhang,
  • Qian Wang,
  • Yanxian Wu,
  • Gengchen Bian,
  • Weidong Yang

摘要

The conventional double-constrained inversion model of horizontal in situ stresses, which considers the constraints from the stress polygon and borehole failure data associated with borehole collapse (breakout) and tensile fracture, fails to adequately capture the impact of geomechanical parameter uncertainty on in situ stress. Therefore, this study considers the uncertainty of geomechanical parameters and Bayesian theory, and reconstructs the conventional double-constrained inversion model into a new Bayesian inversion model (i.e., a quantitative uncertainty probability model) of horizontal in situ stresses for an arbitrarily inclined borehole, which is naturally applicable to vertical wells. By introducing the delayed rejection adaptive Markov chain Monte Carlo (DRAM MCMC) algorithm, the model effectively integrates historical prior distribution information of geomechanical parameters with current observed borehole failure data, thereby further alleviating irregularities in the posterior marginal distribution of the traditional Metropolis–Hastings (MH) algorithm. This study demonstrates that the uncertainty characteristic of Maximum horizontal in situ stress tends to be normally distributed under a non-informative prior. When the minimum horizontal in situ stress obeys the normal prior distribution and is updated by the Bayesian method, the predicted Maximum horizontal in situ stress has lower uncertainty than the case of a uniform prior distribution. However, the prior type of the minimum horizontal in situ stress gradually exerts a weakening influence on the prediction of the maximum horizontal in situ stress with increasing observation error. Furthermore, the proposed Bayesian inversion model shows strong stability when dealing with uncertainties in multiple geomechanical parameters. It effectively converts the uncertainty characteristics of the maximum horizontal in situ stress from an independent joint distribution to a correlated one. This change reduces the risk of wellbore instability by up to 40%, significantly enhancing stress prediction reliability for drilling design and reservoir management.