Multi-leak diagnosis and isolation in oil pipelines based on Unscented Kalman filter

Abstract

This paper proposes a method that handle multi-leak detection problems by Unscented Kalman Filters. To solve this problem, firstly, oil flow in pipelines is considered as a system of non-linear equations. Secondly, since state estimation can provide valuable information, the model is updated and the number of the observer is changed at the same time when each new leak occurs. In this application, the multi-leak detection is solved by using only sensors of flow and pressure at the extremes of the duct. Using information of the leaks occurred before and the estimated state to realize the multi-leak diagnosis and isolation. At the end of this paper, simulation proves that Unscented Kalman Filter has a good result on the multi-leak problem.