协同过滤在油井增产预测中的应用

APPLICATION OF THE COLLABORATIVE FILTERING IN THE PREDICTION OF THE OIL WELL STIMULATIONS

预测油田增产措施效果是减少低效措施,提高措施有效率的重要前提。建立表征增油量的模型,描述了措施后增油量与时间的产量递减规律。并且建立一套集数据预处理、最近邻搜索和产生预测值的协同过滤算法流程。利用灰色关联分析、相关性分析筛选了措施的主要影响因素,优化筛选出了最佳邻域阀值,利用该邻域阀值筛选了目标潜力井的相似井,从而得到该目标潜力井的预测值。油田现场应用结果表明,表征增油量的模型与预测方法的可靠性较高。

The prediction of the oilfield stimulation effects is the essential prerequisite to reduce the low-efficiency well stimulations and improve the stimulation efficiency. The characterizing model of the incremental oil production was established,the production decline laws with the time after the operations were described. At the same time,a set of the collaborative filtering algorithm processes were built,which contains the data preprocessing,nearest neighbor searching,and generated predicted values. With the help of the grey correlation and coherent analyses,the main influencing factors of the stimulations were screened,the best neighborhood threshold value was optimized,and then the similar wells with the potential target wells were chosen,finally the predicted values of the potential wells were obtained. The oilfield application results showed that the reliabilities of the incremental charactering model and predicting method are pretty higher.