Artificial Intelligence for Co2 Pipeline Monitoring: Cross-Domain Insights From Oil, Gas, Water, and Hydrogen Systems

Because carbon capture, utilization, and storage (CCUS) infrastructure is growing rapidly, it has now become more important than ever to monitor carbon dioxide (CO 2) transit pipelines. Nonetheless, investigations on AI-driven leak detection for CO 2 pipelines are scarce. This analysis combines successes from oil, gas, water, and hydrogen pipeline systems, pointing out methods that can be used in other fields and problems that are unique to CO 2 . Deep learning (DL) architectures, especially convolutional neural networks (CNNs), long short-term memory networks (LSTMs), and hybrid models, are always very good at handling sensor data that is nonlinear, high-frequency, and multivariate across several fields. Classical machine learning (ML) algorithms remain effective for structured pressure and flow datasets, especially when interpretability and limited training data are critical. The choice of sensor modality has a significant impact on algorithm appropriateness: acoustic methods provide high sensitivity, pressure/flow-based models enable scalability via existing SCADA systems, and non-destructive testing techniques such as magnetic flux leakage (MFL), distributed temperature sensing (DTS), and thermography aid in structural defect characterization. Multimodal data fusion improves robustness and lowers false alarm rates across systems. Although laboratory accuracies usually exceed 90%, most models require significant field validation. CO 2 pipelines face challenges such as supercritical thermodynamics, rapid depressurization, phase transitions, and impurity-induced corrosion, necessitating physics-based modeling and specific training datasets. This review suggests that while algorithms are well-established in similar disciplines, CO 2 transport requires special modifications rather than just imitating present approaches. Future improvement requires CO 2 -specific benchmark datasets, multimodal sensor integration, physics-informed learning frameworks, and extensive field-scale validation. This work provides a systematic methodology to improve AI-driven monitoring in future CO 2 transport infrastructure by incorporating cross-domain insights. • Focuses specifically on AI-based leak detection for CO 2 pipelines • AI models adapted to CO 2 ’s supercritical behavior and impurity-driven dynamics • Physics-informed and modular hybrid models improve interpretability • Evaluates transfer learning from hydrogen pipelines for CO 2 -specific leak detection • Provides structured guidance for deploying scalable, AI-driven CO 2 monitoring solutions