Development of a hybrid Digital Twin for the dispersion of transient CO2 clouds accompanying accidental pipeline leaks

This study introduces a conceptual Digital Twin framework for real-time modeling of CO 2 dispersion following accidental pipeline leaks or ruptures. Forecasts cloud pathways and concentration gradients both in real-time operations and within a simulated training environment. The system offers simulations on demand for operator training, strengthening preparedness, situational awareness, and rapid decision-making during critical incidents. By integrating high-frequency nondispersive infrared (NDIR) sensor data with a physics-informed long-short-term memory (LSTM) model, validated against synthetic data from DNV Phast software, the developed system generates accurate spatial forecasts of the 5 000 ppm CO 2 dispersion contour. Experimental validation in more than 300 000 simulated scenarios showed a mean squared error just below 5 %, indicating high reliability and operational effectiveness. The proposed Hybrid Gaussian Digital Twin (HyG-CO2DT) offers scalability, maintains robustness under varying sensor densities without requiring model retraining, and provides a practical framework for enhancing safety, regulatory compliance, and effective risk management in industrial carbon capture and storage (CCS) applications. • Real-time CO 2 dispersion forecasts driven by sensor data and physics models. • Cost-effective twin enables real-time monitoring and prediction, avoiding costly CFD. • Novel hybrid algorithm predicts evolving CO 2 zones and release consequences. • Scalable design ensures safety, compliance, and operator training in CCS sites.