Leak Detection in Natural Gas Pipelines Using Intelligent Models

Abstract

Detection of small leaks in gas pipelines is an important and persistent problem in the oil and gas industry. However, the industry is beginning to investigate how tools of Machine Learning, Artificial Intelligence, Big Data, etc. can be used to improve current industry processes. This work aims to study the ability of intelligent models to detect small leaks in a natural gas pipeline using operational parameters such as pressure, temperature and flowrate through existing industry performance metrics (sensitivity, reliability, robustness and accuracy). Observer design technique was applied to detect leaks in a gas pipeline using a regresso-classification hierarchical model where an intelligent model acts as a regressor and a leak detection algorithm acts as a classifier. Five intelligent models (Gradient Boosting, Decision Trees, Random Forest, Support Vector Machine and Artificial Neural Network) were used in this present work. Results showed that the Random Forest and Decision Tree models are the most sensitive as they can detect a leak of 0.1% of nominal flow in about 2 hours. All the intelligent models had high reliability with zero false alarm rate in testing phase. However, due to this level of reliability, the models had low accuracy with the Artificial Neural Network and Support Vector Machine performing best and better regressors than the others. All the intelligent models are robust. The average time to leak detection for different leak sizes for all the intelligent models were compared to a real time transient model in literature. The intelligent models had a time savings of 25% to 48%. Results in this present work further suggest that intelligent models could be used alongside a real time transient model to improve leak detection. Also, that the tools of big data, data analytics, artificial intelligence can be harnessed to improving leak detection results.