A novel data-driven leak detection and localization algorithm using the Kantorovich distance

•Developed a novel data-driven algorithm for pipeline leak detection and localization.•Pipeline operation status change is detected through Kantorovich distance calculation.•Developed pipeline simulation and leak simulation method for performance validation.•Successful application to real industrial oil pipeline revealed the efficacy of the proposed method.

A novel data-driven leak detection and localization algorithm is proposed based on the Kantorovich distance concept. Mass flowrates and pressure measurements are used to identify possible change in the pipeline status. Based on the pipeline leak signature, leak is detected and the location is further inferred. The proposed method was applied successfully to a simulated pipeline in transient condition. The efficacy of the proposed method was also proven by applying it to an industrial pipeline network with controlled leak tests in real time. The method successfully detected both small (as small as 1% of the nominal flow rate) and large test leaks in the realistic pipeline. The time required to detect and localize leak with the proposed algorithm was much lower than the available commercial leak detection system. The accuracy of the proposed leak localization method was demonstrated to be better especially for the small leaks.