基于贝叶斯网络的钻进过程井漏井涌事故预警

Lost circulation and kick accidents warning based on Bayesian network for the drilling process

摘要:

近年来，地质钻探在钻探装备和技术领域取得了长足发展，但在钻进过程事故预警方面的研究仍较欠缺。为保证钻进过程安全高效，降低事故造成的损失，本文提出了一种钻进过程井漏、井涌事故预警方法。首先，通过事故特性分析，选取表征事故特性的钻进参数。其次，考虑事故发生时钻进参数变化的不确定性，基于贝叶斯网络建立井漏、井涌事故预警模型。再者，为从含噪声的实际钻进数据中有效提取钻进参数的变化趋势，综合利用归一化、滑动平均和最小二乘线性拟合方法进行节点状态判断。最后，利用实际钻进数据对井漏、井涌事故预警模型的有效性进行验证，并对不同趋势判断界限和滑动窗口对报警性能的影响进行探讨。实验结果表明，该预警模型可对井漏、井涌事故进行有效预警，并且合适的趋势判断界限和滑动窗口可降低报警延迟，减少误报和漏报现象。

Abstract:

In recent years, geological drilling has made great progress in the field of drilling equipment and technology, but research on drilling accidents warning is still insufficient. To ensure safety and efficiency of the drilling process and reduce the losses caused by accidents, a lost circulation and kick accidents warning method for the drilling process is proposed in this paper. First, the drilling parameters that characterize accidents are selected through the analysis of accidents. Second, considering the uncertainty of the changes of drilling parameters when an accident occurs, a lost circulation and kick accidents warning model is established based on Bayesian network. Third, to effectively extract the trend of drilling parameters from actual drilling data with noises, the normalization, moving average and least square linear fitting methods are jointly used to judge the node status. Finally, the actual drilling data is used to verify the lost circulation and kick accidents warning model. The influence of different trend judgment boundaries and moving windows on the alarm performance is discussed. The experimental results show that the warning model effectively provides warning of the lost circulation and kick accidents. And a proper trend judgment boundary and moving window can reduce alarm delay, false alarm and missed alarm.