This paper proposes a novel model-based leak location method for pipelines. We first formulate the leak location problem into an optimization problem under certain conditions. Then we develop a derivative-based search algorithm to solve that optimization problem. The derivative will be calculated from an approximate of the underlying mathematical model that describes the dynamics of pipeline systems, by means of discretization and linearization. Results of simulation and tests using real pipeline data show that this proposed method is accurate, timely, and computationally effective.
Pipelines are one of the major transportation means in oil and gas industry. Leakage remains one of the concerns for pipeline operators. Once a leak occurs, it is important to detect and locate it quickly and accurately in order to minimize potential economic loss and/or environmental damage. Thus leak detection and location is an indispensable technology for managing pipelines smoothly and safely. In this note we are only concerned about leak location, assuming that an existing leak has been detected.
Despite the fact that many techniques have been developed in order to address this issue, locating a leak timely and accurately remains a challenging problem both in industry and in academia, see and references therein. One of the conventional methods for leak location is the pressure-gradient method which utilizes the fact that the leak outflow changes the steady pressure-gradients for the sections before and after the leak, respectively. Thus the intersection of those two gradient lines represents the true leak location. This method assumes that the pipeline system is in steady state, which however is a quite stringent condition since in practice most pipelines contain rich transients all the time.