In order to quantify the performance ranking of oil wells in high water cut reservoir, a quantitative evaluation method of oil wells is established from two perspectives of numerical simulation and machine learning respectively. The backward flight time of oil wells control area is obtained by numerical simulation. The flow heterogeneity of oil wells is evaluated based on Lorentz coefficient, and the concept of potential index is proposed to characterize the oil wells potential and flow capacity. A large number of dynamic data of oilfield development is collected to establish time series model. The production history of oil wells is fitted by using VAR algorithm of machine learning, the impact of oil wells production on the whole reservoir development is evaluated through impulse response analysis and the production capacity is quantified by the cumulative influence coefficient. The scores obtained by the two methods are compared, and the comprehensive performance ranking of oil wells is determined by the entropy weight method. The evaluation method is applied to Gangxi Oilfield. The results show that the two evaluation methods are based on different assumptions, but the scoring trend of each well is basically the same. Due to the comprehensive consideration of the influence of numerical simulation and development dynamic data, the final evaluation score objectively reflects the development status of oil wells. The new method provides a theoretical basis for the effective development of high water cut reservoir.

为量化高含水油藏油井性能等级，分别从数值模拟和机器学习两个角度建立了油井定量评价方法。通过数值模拟得到油井控制区的后飞时间。基于洛伦兹系数评价油井的流动非均质性，提出势指数的概念来表征油井势和流动能力。采集大量油田开发动态数据，建立时间序列模型。使用机器学习的VAR算法拟合油井的生产历史，通过脉冲响应分析评估油井产量对整个油藏开发的影响，并通过累积影响系数量化产能。将两种方法得到的分数进行比较，采用熵权法确定油井综合性能等级。该评价方法应用于港西油田。结果表明，两种评价方法基于不同的假设，但每口井的评分趋势基本相同。由于综合考虑了数值模拟和开发动态数据的影响，最终的评价分数客观地反映了油井的开发状况。新方法为高含水油藏的有效开发提供了理论依据。