基于朴素贝叶斯算法评估页岩油藏产能

1.中国石油新疆油田分公司吉庆油田作业区,新疆吉木萨尔831700;2.长江大学石油工程学院,湖北武汉430100

油田开发;页岩油;压裂水平井;朴素贝叶斯算法;概率模型;产能评估;吉木萨尔

Productivity evaluation based on naive Bayesian algorithm in shale reservoir
LIANG Chenggang1,LI Juhua2,CHEN Yiwei1,QIN Shunli2,ZHANG Jinfeng1,and HU Ke1

1.Jiqing Oilfield Operation Area of Xinjiang Oilfield Company, CNPC, Karamay 831700, Xinjiang Uygur Autonomous Region, P. R. China;2.School of Petroleum Engineering, Yangtze University, Wuhan 430100, Hubei Province, P. R. China

oilfield development; shale oil; fractured horizontal well; naive Bayesian algorithm; probability model;productivity evaluation; Jimsar

DOI: 10.3724/SP.J.1249.2023.01066

备注

作为非常规油气资源,页岩油储量巨大,是重要的接替资源,高效开发页岩油藏具有十分重要的意义.针对页岩油藏“水平井+体积压裂”衰竭式开发产量递减快,现有产能评价方法预测采收率低的问题,基于朴素贝叶斯算法构建压裂水平井产能评估方法.以新疆油田吉木萨尔页岩油藏为研究对象,以3a累产作为分类评价指标,分别建立多分类的朴素贝叶斯先验概率模型以及包括地质参数和工程参数2类共5项属性的条件概率模型,得到基于贝叶斯理论计算的4类产能后验概率模型,实现页岩油藏压裂水平井产能评估.结果表明,基于朴素贝叶斯算法得到的页岩油藏产能分类方法适用性高,产能预测准确率分别为Ⅰ类井94%,Ⅱ类井为71%,Ⅲ类井为87%,Ⅳ类井为92%;绘制了吉木萨尔页岩油藏压裂水平井产能分级趋势分布图,评估高产潜力区主要分布在油藏东南部;构建的后验概率模型指出,优化压裂施工参数可提升高产概率,为后续的水平井大规模压裂改造开发页岩油提供指导依据.
Shale oil, as an unconventional oil and gas resource with huge reserves, has become an important replacement resource and great significance to develop. Aiming at the problems of rapid decline oil production in the depletion development of shale oil reservoirs by "depleted horizontal well+volume fracturing", and the lower recovery factor predicted by the existing productivity evaluation methods in shale reservoir, we construct a productivity evaluation method for fractured horizontal wells based on naive Bayes algorithm. Taking the Jimsar shale reservoir in Xinjiang oilfield as the target reservoir, we establish the multi-classification Naive Bayes prior probability model and the conditional probability model containing five attributes of geological parameters and engineering parameters by taking the three-year cumulative production as the classification evaluation index. Then we obtain the posterior-probability model of the four types of production capacity based on Bayesian theory to achieve the capacity assessment of fractured horizontal wells in shale oil reservoirs. The results show that the proposed shale reservoir capacity classification method based on the Naive Bayes model is applicable. The accuracy of productivity prediction is 94% for Class Ⅰ wells, 71% for Class Ⅱ wells, 87% for Class Ⅲ wells, and 92% for Class Ⅳ wells, respectively. The productivity classification trend distribution map of fracturing horizontal wells in Jimsar shale reservoir is drawn, and the high-production potential area is mainly distributed in the southeast of the reservoir. The analysis of the constructed posterior probability model shows that the optimization of fracturing construction parameters can improve the probability of high well production rate. The study provides a guiding basis for the subsequent large-scale fracturing reconstruction of horizontal wells for shale oil development.
·