A new leak recognition method for natural gas pipelines in the urban underground space based on probabilistic neural network

Abstract

Natural gas leakage is one of the sources of combustible gas in the underground spaces, which can easily lead to dangerous explosions and fatal accidents. Recognizing the natural gas leakage accurately from underground combustible gas can improve the efficiency of early warning and maintenance process. However, the problem of poor gas pipeline leak identification efficiency in combustible gas concentration detection has to date remained unsolved. This paper proposes a novel and effective method for leak recognition of gas pipelines based on probabilistic neural network (PNN). First, the characteristics of natural gas leakage are analyzed. Then time-domain feature quantities and correlation feature quantity are extracted from the methane concentration and temperature profiles, which are used as the input vectors. Finally, the feature vectors are input to the PNN for natural gas leak recognition. The results show that the proposed leak recognition method based on PNN can effectively identify the natural gas leak, and it exhibits better accuracy (as high as 92.5%) in comparison with the random forest and support vector machine methods. This work can achieve timely natural gas leak detecting with high accuracy and provide a guideline for early warning of gas leak in the urban underground space.