Perpustakaan Digital ITB

Pipeline in the oil and gas industry is a network of pipe systems used to transport oil, gas and related products from one location to another. The urgency of detecting pipeline leaks is very important in the oil and gas industry because leaks can cause serious impacts, such as environmental damage, accidents, fires, and potential dangers to society. Early detection of leaks is crucial to protect public safety, maintain operational continuity, and minimize large financial losses. The Machine Learning (ML) method using Artificial Neural Network (ANN) has been used effectively in detecting leaks in the pipeline in this study. ML ANN can be trained using historical leak data and inlet-outlet parameters (mass flow, pressure, gas velocity, and temperature) to study patterns and characteristics that indicate a leak. Once trained, the ML ANN model can be used to analyze real-time data obtained from sensors and pipeline monitoring systems to detect leaks quickly and accurately. This approach enables early detection of leaks, reducing negative environmental, safety and financial impacts. This study will use historical data from one of the existing pipelines in Indonesia, specifically in the West Java region. This pipeline is 12” in diameter and connects Field A with Onshore Processing Facility (OPF) B for 34 km. It is known from the gas analysis data that the gas fluid flowing in the pipeline is dominated by the methane component (80%) with impurities CO2 (0.98%) and N2 (2.43%). For flowrate and pressure data recorded from 1 January 2017 to 31 July 2022. At the end of this study, a system was obtained that can predict the location and size of pipeline leaks using the ANN machine learning model. This model can predict the location of the leak with an r^2 score of 0.86 and predict the size of the leak with an r^2 score of 0.996. This study can be implemented if given real-time input data, so predictions will be made on an ongoing basis in order to anticipate leaks.