Pipeline networks play a key role in the efficient transportation of oil and the overall economy of the oil industry. However, pipelines are vulnerable to the risk of oil leakage, which can lead to spilling, loss of product, pollution, loss of lives, and loss of the source of livelihood. This research aims at detecting oil leakages. The Trans-Niger Pipeline (TNP), which is used to route crude oil from a Shell Petroleum Development Company (SPDC) facility in the Gbaran community of Bayelsa state, as a case study. In this study, an artificial neural network (ANN) approach was used to predict oil pipeline leakage based on the outlet pressure of the facility. The model was developed using a feed-forward propagation neural network (FFNN) based on historical outlet pressure data from oil fields and trained using the Bayesian regularisation algorithm using outlet pressure data from shell field locations in Bayelsa State to evaluate performance. The data was detrended to remove a downward trend in the measured or sensed pressure. The Bayesian algorithm converged before 1000 maximum iterations with a mean squared error (MSE) of 12.262. The long short-term memory (LSTM) used to predict leakge the pipeline, gave a root mean square error (RMSE) of 0.4 at the 30th iteration. This study can help the operators expedite pipleline maintenance and clean up actions in any event of pipe leakage.
Keywords: Oil leakage, Pipeline, Pressure, Bayesian regularisation, Trans-Niger Pipeline, Algorithm.
