| Summary: | Oil and gas transportation in cold regions, such as the Arctic and offshore, may encounter flow assurance issues that are caused by hydrates, wax, or even ice. To evaluate the effect of these solids on pipeline safety, this dissertation studied pipeline safety from three aspects, a mechanistic model to predict wax and ice deposition rate, a thermodynamic model to predict hydrate forming temperature based on machine learning algorithms, and a probabilistic approach to evaluate hydrate forming probability, hydrate blockage probability, and wax blockage probability. The mechanistic model is proposed to estimate the deposition rate for the systems containing only wax, only ice, or the combination of wax and ice. The model is established based on the mechanism of molecular diffusion. The model is validated using experimental data. The simulation results show that the wax deposition rate is faster than the ice deposition rate. In addition, the existence of ice can affect wax deposition rate. To develop the thermodynamic model, 702 experimental data points from 1951 to 2020 were collected to establish a database for methane hydrate formation temperatures in saltwater. Five machine learning algorithms, including Multiple Linear Regression, Support Vector Regression (SVR), k-Nearest Neighbor (k-NN), Random Forest (RF), and Gradient Boosting Regression (GBR), are compared. The model based on Gradient Boosting Regression has the best prediction result among the five machine learning algorithms, with R^2 = 0.998 and AARD% = 0.074, which is much lower than the AARD% of 0.7%-37.6% from the traditional thermodynamic models, such as van der Waal-Platteeuw or CSMGem for predicting hydrate formation temperature in saltwater. To link the deposition rate prediction model and thermodynamics model with risk assessment, three probabilistic models are developed to assess hydrate formation probability, hydrate blockage probability, and wax blockage probability. The proposed methods convert the thermo-hydraulic data into the probability ...
|
|---|