基于朴素贝叶斯算法的钻井溢流实时预警研究

高温高压井钻井过程中溢流未及时发现将引起严重后果，现有溢流监测手段依赖井下或地面工具，存在一定的时间滞后性，为此创建了基于朴素贝叶斯算法和钻井大数据的溢流实时预警方法。在具备一定已钻井规模的区域内，分析历史井溢流发生与地质资料、随钻测录井数据的概率联系，分别建立溢流的先验概率计算模型和包含了区域、地层、岩性、扭矩、泵压、机械钻速共6项属性的条件概率计算模型，基于贝叶斯理论计算溢流的后验概率，实现实时预警功能。研究表明，基于朴素贝叶斯的溢流预警方法在可靠性、传输时效性、资料可获得性等方面存在较大优势，结合实际算例，验证了方法的可行性。

Real-time early warning of drilling overflow based on naive Bayes algorithm

If the overflow in the drilling process of high temperature and high pressure (HTHP) well isn’t identified in time, serious consequence will be caused. Existing overflow monitoring technology depends on downhole or surface tools, so it is time lagged to some extent. To solve this problem, this paper developed the real-time overflow early warning method based on naive Bayes and drilling big data. As for an area with a certain scale of drilled wells, the probabilistic relationships between the historical drilling overflow and the geological data, mud logging while drilling and wireline logging while drilling are analyzed, the prior probability calculation model of overflow and the condition probability calculation model containing six attributes of region, stratigraphy, lithology, torque, pumping pressure and rate of penetration (ROP) are established. The posterior probability of overflow is calculated based on Bayes theory. Thus, the function of real-time early warning is realized. It is indicated the overflow early warning method based on naive Bayes is more advantageous in terms of reliability, transmission efficiency and data accessibility. And its feasibility is verified based on actual cases.