A hydrate blockage recognition technique for natural gas pipelines based on BP neural network

Abstract

Abstract In energy industry field, hydrate blockage in natural gas pipelines has become a critical issue with an increasing demand of natural gas in the world. In this paper, a hydrate blockage recognition technique has been proposed based on BP neural network, which can recognize leakage in natural gas pipelines as well. A monitoring system has been built for trials, which mainly includes a loudspeaker and a microphone for acoustic signals transmission and reception. This paper mainly introduces wavelet packet theory and BP neural network recognition algorithm for signal analysis and recognition purpose. The reflected signals are analyzed and extracted eigenvectors by using wavelet packet transform, then a three-layer BP neural network is built and trained. Experimental results show that this technique can recognize hydrate blockage and leakage effectively.