用于油水界面测量的中值预处理聚类算法

Clustering algorithm with median pretreatment for oil-water interface detection

原油储罐油水界面测量过程是油田联合站信息化系统过程控制中的重要环节。油水界面测量过程中,为了解决油水界面数据存在伪数据及油水界面传统算法结果误差较大等问题,提出了用于油水界面测量的中值预处理聚类算法。该算法首先采用中值预处理算法消除油水界面数据中的伪数据,获得有利于聚类划分的优化数据;其次根据油罐内介质特性,利用改进的K-means聚类算法确定聚类中心数量和阈值大小,划分油水界面数据分段区域,并反复对比区域变化,不断修正初始阈值,确定一组最优阈值;最后按照分类统计的方法求得油水界面及液位高度,实现油水界面精确计算。实验结果表明,油水界面中值预处理聚类算法比传统算法更稳定、更准确、更智能,适合于为原油储罐油水界面监控和盘库系统提供精确的监测数据。

The oil-water interface measurement process of crude oil storage tank is very important for the information system process control of oil combination station. In order to solve the problems that pseudo data and improper threshold in oil-water interface calculation, this paper proposes a clustering algorithm with median pretreatment for oil-water interface detection. The proposed algorithm works as follows: Use the median pretreatment algorithm to eliminate the pseudo data to get optimized data for clustering division, then use the improved K-means clustering algorithm, determine the number of cluster centers and size, divide oil-water interface data area, compare the data area changes and fix the initial threshold many times until a set of optimal thresholds is determined, and finally calculate the oil-water interface and the liquid level according to the method of classifying statistics. Experimental results showed that the clustering algorithm with median pretreatment for oil-water interface detection is more stable, more accurate and more intelligent than the traditional algorithm. In other words, it is suitable for providing accurate detection data for oil-water interface monitoring and inventory system by the proposed algorithm.