ANALYSIS OF HYDRATE FORMATION IN PIPELINE TRANSPORTATION USING A MACHINE LEARNING APPROACH IN THE “IAR” FIELD

• Main Author: Afif Ritonga, Ichwan
• Format: Theses
• Language: Indonesia
• Subjects: Pertambangan dan operasi berkaitan
• Online Access: https://digilib.itb.ac.id/gdl/view/84408
• Institution: Institut Teknologi Bandung

The formation of gas hydrates in hydrocarbon transport pipelines is a critical issue that can disrupt flow and cause blockages, potentially posing safety risks and economic losses. Hydrate are ice-like crystaline solids known as clathrates from a structural perspective. The basic structure of thee compounds consists of water molecules forming cage-like crystals that contain guest molecules, resulting in hydrate formation. Hydrate will form when there is enough water to create them, under conditions of low temperature and high pressure. This study examines the application of machine learning to predict hydrate formation in transport pipelines. The study uses three different cases, namely fluid case A, fluid case B dan fluid case C. these three fluid cases will be matched to the conditions of the IAR field using a dynamic multiphase flow simulator. The matching results for the three fluids will be used as the dataset for developing the machine learning model. The algorithms used to predict hydrate formation include Random Forest, Gradient Boosting, Extreme Gradient Boosting, CatBoost and Support Vector Machine. These models are trained and validated using datasets that are divided into training data, testing data and cross-validation. Next we will compare the constructed data with the blind test data. In this study, fluid case A and fluid case B are the datasets used for model development, while fluid case C serves as the blind test data.the evaluation results of the developed models show that Random Forest provides the best performance with a prediction accuracy 0f 82 %, followed by Extreme gradient boosting at 60%, Gradient boosting at 54%, Catboost at 53% and Support Vector Machine at 0,57%