A Mechanism based Data-Driven Model for Prediction of Hydrate Formation

Abstract

Hydrate is one of the most common challenges in flow assurance. Mechanism model or empirical model is usually adopted to predict hydrate formation under a specific condition. However, the methods are difficult to operate in real-time change of actual situation. In this paper, a mechanism-based data-driven modeling method is built to predict hydrate formation. Based on the collected data, including temperature, pressure and components, a data-driven method is introduced to identify the unknown parameters in the mechanism model. 131 groups of experimental data were collected to make a correlation analysis to determine the main components affecting hydrate formation. Four different component systems were calculated using the mechanism model (P-P model), empirical model (Makogon model) and data-driven mechanism model for comparison. Results show that the average error of the data-driven model is as low as 0.0085 MPa, and this method can overcome the irrationality of prediction caused by only using historical data or mathematical formula.