Predicting Wax Formation Using Artificial Neural Network

Abstract

Abstract Wax formation is a problematic phenomenon in petroleum production operations especially in deep offshore fields. Wax begins to form when temperature falls below the cloud point of the oil as it travels from reservoir to surface lines. Wax deposition may lead to increased pumping power to maintain smooth flow, blockage of flow lines, production delay and even total abandonment of the entire flow line system. A reliable prediction of the onset conditions for wax formation is an important step towards prevention and mitigation of the resultant wax deposition. Prediction of WAT by most of the existing models deviate from the experimental results due to simplifying assumptions. In this work, Artificial Neural Network (ANN) approach was used to predict the WAT for various hydrocarbon mixtures. The model was trained using different combinations thermodynamic properties of twelve different hydrocarbon fluids subsequently validated with experimental data. Results show that the ANN approach predicts the WAT more accurately than the existing models. The Average Absolute Deviations (AAD%) for the ANN are less than those of the existing models. A combination of molecular weight, density and activation energy as input parameters provided the best prediction.