Flow characteristic and blockage mechanism with hydrate formation in multiphase transmission pipelines: In-situ observation and machine learning predictions

Gas hydrate blockage is a major challenge in flow assurance. The mechanisms underlying hydrate formation and blockage are still not well understood, and appropriate forecasting methods are needed. In this study, using a high-pressure fully visual flow loop to simulate multiphase flow conditions, hydrate formation and blockage were evaluated in a pure water system. The effects of liquid loading and pump speeds on the pressure difference and water conversion rate were investigated. Three regions and two critical transition points (i.e., “action point” and “blockage point”) in the hydrate formation process were defined. Based on analysis, it is believed the wet agglomeration results in increased pressure difference and hydrate blockage. An underlying mechanism in the pure water system was proposed. By applying machine learning methods, two prediction models for hydrate blockage were established, with accuracies of 99% and 71%, respectively. Data-driven machine learning models with high flexibility could be an effective method for hydrate blockage management in future.