Natural gas pipeline weak leakage detection based on negative pressure wave decomposition and feature enhancement

Abstract

Natural gas pipeline leakage detection (PLD) based on negative pressure wave (NPW) signals faces significant challenges, including external noise that obscures crucial information and inadequate feature extraction, which often result in low detection accuracy. To address these issues, a weak leakage detection model for natural gas pipelines, named MDDet, is proposed, which integrates variational mode decomposition (VMD)-based signal decomposition and feature enhancement. The MDDet consists of two main components. The first component is the mutual difference distance (MDD) algorithm, which processes NPW signals by integrating signal decomposition for denoising and selecting the optimal intrinsic mode function $I_o$ related to leakage information. The second component is the dual-stream enhanced feature (DEF) algorithm that uses data cropping and dimensionality enhancement to enhance feature for weak leakage detection. Field tests were conducted on natural gas supply systems in two cities to validate the model, with further evaluation of its efficiency in realistic urban pipeline environments in China. The results demonstrate that the MDD algorithm accurately extracts effective leakage information and the DEF algorithm effectively classifies multi-channel feature sample, reflecting the working conditions of the monitored pipelines.