Real-time gas pipelines leak detection and localization using an adaptive Kalman filter with adjusted process noise and iterative pseudo-leak point

Abstract

Pipeline leak issues have become increasingly prominent due to various disasters, making real-time detection and accurate localization of leak points critically important. The conventional Kalman filter (KF) method for leak localization offers real-time capabilities but suffers from slow convergence and low accuracy. To address this issue, this paper proposes an adaptive Kalman filter (AKF) that accelerates leak magnitude estimation through adjusted process noise and improves leak localization accuracy via iterative pseudo-leak point. State equations for a multi-node system incorporating pseudo-leak points are derived, enabling the distinction between bad data and fault data through nodal gas flow conservation principles and the transmission characteristics of fault data. Detection sensitivity is enhanced by using dual judgment criteria that integrate the mass balance principle with estimated leak magnitude, facilitating rapid verification of leak status via step adjustment of process noise during detection. Once a leak is confirmed, process noise is gradually adjusted to expedite the estimation of leak magnitude. After the estimated leak magnitude stabilizes, the position estimation accuracy is further refined through iterative optimization of the pseudo-leak point. For a seven-node system, where each pipeline is 5 km, under diverse scenarios encompassing complex flow, minor leaks, and major leaks, the relative error of estimated leak location remains below 0.5% within 5 minutes. Compared with conventional KF, the proposed method reduces the convergence time of leak magnitude from 2 hours 40 minutes to 18 seconds and improves leak localization accuracy from 8.8% to 0.29%.