Study on throttling risk assessment for high-pressure CO2 pipelines based on machine learning

To address the risks of low temperature and leakage during the throttling and venting process of high-pressure CO₂ pipelines, this paper proposes a risk assessment and prediction method with machine learning. By comprehensively considering venting parameters such as inlet pressure, inlet temperature, valve opening, as well as throttling pipe diameter and length, this study selects outlet temperature and mass flow rate as the risk assessment criteria. This study constructs and trains a risk assessment model, which achieves accurate risk level classification and enables the visualization of risk windows. The Random Forest algorithm was chosen for model construction and compared with Support Vector Machine. Results showed Random Forest demonstrated higher accuracy, leading to its adoption for throttling risk assessment. The results indicate that valve opening, inlet pressure, inlet temperature, and throttling pipe diameter are the primary factors influencing throttling risk, with their respective mechanisms of influence analysed. The valve opening has the most significant impact on throttling risk, accounting for 44.71%. For industrial venting operation, strictly controlling valve opening and prioritizing smaller throttling pipe diameter can reduce low-temperature and leakage risks. This study provides a reliable theoretical tool and engineering guidance for risk prevention and safe operation during throttling processes in high-pressure CO₂ pipelines.