基于马尔科夫链和贝叶斯网络的钻井风险预测

钻井作业是高风险高投资的过程，这个过程中存在许多可能导致重大钻井事故的不确定因素，对此类不确定性因素进行预测进而达到预警或控制的目的，提前做好风险预防或降低风险损失具有较大的经济意义。通过研究钻井风险预测、马尔科夫链和贝叶斯网络方法，根据现场采用的指标体系，提出融合马尔科夫链和贝叶斯网络的钻井风险预测新方法。该方法可从纵、横两方面预测钻井事故的风险、弥补单独用马尔科夫链处理上层指标数据欠缺的不足；并可为诊断、监测和控制风险提供理论依据。实例研究表明，该方法是正确和可行的，用马尔科夫链进行纵向预测与实际的吻合度为82%，而贝叶斯网络仅为46%，融合后的方法优于现有方法。

Drilling risk prediction based on Markov chain and Bayesian network

Drilling operation is a risky and costly process, during which many uncertainties may cause a serious accident. In order to prevent or mitigate the risks and thereby avoid economic loss, it is necessary to predict these uncertainties. In this paper, the existing drilling risk prediction methods (e.g. Markova chain and Bayesian network) were reviewed, and then a new drilling risk prediction method was proposed by integrating the Markova chain and Bayesian network based on the index system adopted on site. This new method can be used predict the risk of drilling accident vertically and horizontally, and also overcome the shortage which occurs when the upper indices are processed only by using Markova chain. Moreover, it provides the theoretical basis for the risk diagnosing, monitoring and controlling. The case study shows that this new method is correct and feasible. The goodness of fit between the vertical prediction and the actual data of the integrated method is higher than that of Markova chain (82%) and Bayesian network (46%).