基于K-means聚类分析和多元线性回归的相关流量数据处理方法

Data Processing Method of Cross-correlation Flow Based on K-means Cluster Analysis and Multiple Linear Regression

相关流量计在油井产出剖面测量中得到了成功的应用。但因传感器、调理电路以及流体本身噪声的影响,相关流量计所测量的渡越时间值会出现少量异常数据,使瞬时流速的计算结果与实际值相差很大,进而平均流量计算也出现较大的测量误差。对此提出基于K-means聚类算法对渡越时间样本数据聚类分析,并根据聚类结果建立多元线性回归预测模型,合理预测渡越时间值,以修正渡越时间的异常值。对预测值与实际值进行比较,最终获得准确的相关流量数据。采用多相流装置的实验数据对所建立的方法进行验证,结果表明,该方法可有效消除渡越时间的异常,优化流量测量的数据,对两相流流量测量有一定的实践意义。

Cross-correlation flowmeter have been successfully applied in the measurement of oil well production profiles.However,due to the influence of sensors,conditioning circuits,and the noise of the fluid itself,a small amount of abnormal data may appear in the transit time values measured by the cross-correlation flowmeter,resulting in a significant difference between the calculated instantaneous flow rate and the actual value,leading to significant measurement errors in the average flow rate.In this regard,K-means clustering algorithm is proposed to cluster the transit time sample data,and a multivariate linear regression prediction model is established according to the clustering results to reasonably predict the transit time value to correct the outlier of the transit time.Comparing the predicted value with the actual value,the accurate cross-correlation flow data was obtained.The established method was validated using experimental data from a multiphase flow loop,and the results show that the method can effectively eliminate anomalies in transit time,optimize flow measurement data,and have certain practical significance for two-phase flow measurement.