基于窗口样本相似因子分析的油井工况识别方法

针对传统工况识别方法在应对生产波动异常数据干扰时,容易发生工况误判的情形,提出了采用窗口样本相似因子分析的方法来合理表征不同工况下的数据特性,以窗口样本间的相似因子来衡量不同样本数据的相似性.采用改进的K-means聚类算法根据窗口样本相似因子对不同工况下的生产特征参数进行聚类分析,完成多工况的识别过程.根据辽河油田生产数据进行实验验证,结果表明,该方法能够有效消除异常数据对工况数据特性的影响,减少工况误判情况的发生.

Recognition method of working conditions of oil well based on similarity factor analysis for window samples

Aiming at the fact that traditional condition recognition method is liable to the misjudgment for working conditions when responding to the production fluctuation caused by abnormal data interference, a new method based on the similarity factor analysis for window samples was proposed to reasonably characterize the data characteristics under different conditions. In addition, the similarity of different sample data was measured with the similarity factors among window samples. With the improved K-means clustering algorithm, the clustering analysis for production characteristic parameters under different conditions was performed. According to the similarity factor of window samples, the recognition process of multi-conditions was accomplished. The experimental verification was performed, according to the production data of Liaohe Oil Field Group. The results show that the as-proposed method can effectively eliminate the effect of abnormal data on data characteristic and reduce the occurrence of misjudgment for working conditions.