WAX DEPOSITION RATE INSIDE PIPELINES PREDICTION USING ARTIFICIAL NEURAL NETWORK APPROACH: A CASE STUDY OF X FIELD, RIAU

Abstract

This final thesis research provides an AI-based solution approach for the wax deposition problem to predict the wax deposition rate inside pipelines. A deep learning model of the Artificial Neural Network (ANN) algorithm is exercised through machine learning processes, starting with data preprocessing, ANN model building, and ANN model evaluation. All the data used in this research are based on the real-time measurement of field data and laboratory tests from Station A and Station B at Field X, with a total of 375 data points. Two different analytical are being compared to opt for the best approach as the data basis for wax deposition rate. From the calculation, the wax deposition rate from the Matzain model gives more reasonable results. This study involves input parameters of pressure, temperature, flow rate, API, density, viscosity, % wax content, WAT, pipeline inner diameter, and pipeline length. Results show that the optimum ANN model gives the values of Coefficient of Determination (R2), Mean Squared Error (MSE), and Mean Absolute Percentage Error (MAPE) for the test set of 0.9881, 0.3211, and 0.0274, respectively. The error potential from the MSE score is only 0.5666 pm/s or 0.0044675 mm wax thickness accumulation if the assumption of three months period with a constant wax deposition rate is set. Therefore, this research proves the suitability of an ANN algorithm in predicting wax deposition problems.