Oil and Gas Pipeline Leakage Detection using IoT and Deep Learning Algorithm

Abstract

Pipeline leaks are a frequent occurrence in oil and gas infrastructure worldwide. Though leak detection systems are expected to be installed on all pipelines in the near future, relying on human efforts to physically monitor these pipelines is and will continue to be challenging. Though today's leak detection techniques are not able to completely stop leaks from occurring or to detect most leaks, they are essential in lessening their effects. Despite recent developments toward solving this problem, the solution still falls short of expectations. This research presents an approach to pipeline leak detection by leveraging on the exceptional abilities of Convolutional Neural Network (CNN) and Internet of Things (IoT). A comprehensive dataset on oil and pipeline leakage is collected, and the CNN model is developed and trained with the collected dataset. Thereafter, the trained model is integrated into the monitoring system to provide notifications of leaks. The model is adaptable and scalable and its performance, as evaluated, shows an improvement over existing systems with an accuracy of 97% hence well suited for deployment in various pipeline networks for the overall improvement of safety environment in the oil and gas sector.

Keywords

• artificial intelligence, convolutional neural networks, computer vision, pipeline leak detection, machine learning