A Small Leak Detection and Localization Method for Oil Pipelines Based on Improved Robust Principle Component Analysis

Abstract

Abstract When there is a small leak in the oil pipeline, their susceptibility to the influence of external noise prevents their detection based on the robust principal component analysis (RPCA) method, which does not consider the matrix of dense noise. Thus, to solve this problem, leak detection and localization method based on an improved robust principal component analysis (IRPCA) for pipelines is proposed. By solving a convex optimization function, this method can remove the dense noise contained in the collected data and improve the efficiency of small leak detection. In addition, the collected leakage data is analyzed. Leak detection is monitored by the combined index D2, which is a combined indicator that is composed of Hotelling’s T2 statistic and the SPE statistic. The experimental results show that the missing and false leak detection accuracies of the combined index D2 are much higher than those of the Hotelling’s T2 statistic and the SPE statistic separately. Furthermore, it verifies that the small leak localization method proposed in this paper has a good effect.