Predictive AI Models for Detecting Pipeline Leaks in the Energy Industry

Abstract

Pipelines serve as critical infrastructure for transporting oil and gas, but any leaks in these systems can lead to severe outcomes, including fires, injuries, environmental pollution, and property destruction. Thus, maintaining the integrity of pipelines is paramount for ensuring a safe and sustainable energy supply. This thesis investigates the application of Artificial Intelligence (AI), Machine Learning (ML), and Deep Learn- ing (DL) in enhancing leak detection within oil and gas pipeline systems, key to ensuring environmental safety and economic stability. Through a comprehensive review and data-driven methodologies, the study demonstrates how ML algorithms, including neural networks and deep learning models, significantly out- perform traditional leak detection methods in accuracy and timeliness. Herein, the research introduces machine learning-based anomaly detection models proposed to solve the problem of oil and gas pipeline leakage. To address this, several machine learning and deep learning algorithms, namely, Random For- est, Support Vector Machine, K-Nearest Neighbor, Gradient Boosting, Decision Tree, Convolutional Neural Network, and Multi-Layer Perceptron, were employed to develop robust detection models for pipeline leaks. Among these, the Support Vector Machine algorithm, achieving an accuracy of 96.6%, notably out- performed other models, thereby confirming its efficacy as a highly accurate tool for detecting leakage in oil and gas pipelines.