Predicting the Occurrence of Wax Precipitation in Crude Oil Pipelines Using Machine Learning

Abstract One major problem faced by the oil industry is wax formation in pipelines which is the high molecular weight paraffin fraction of crude oil. Deep water pipes, including onshore and offshore surface infrastructure like pipelines, separators, and crude oil processing plants, are susceptible to wax deposition. The production of oil and gas condensate may be interrupted, facilities may fail and require expensive intervention operations, operational inefficiencies may arise, maintenance costs may rise, pipelines may become blocked, or pipes may be completely abandoned as a result of this unwanted event. The main objective of this research is to provide a novel approach for predicting the occurrence of wax precipitation in crude oil pipelines. In this research three Python models which are Support Vector Classifier (SVC), Gradient Boosting Classifier (GBC) and Gaussian NB (GNB) were trained and validated on a dataset. The SVC, GBC and GNB models gave an impressive performance having high accuracy scores of 0.9824, 0.9298 and 0.7543 respectively. From the results predicted by the models, it shows that the Support Vector Classifier (SVC) model performed the best among the other models. Therefore, from the performance made by the models it is clear that Python based models can be used as a powerful tool for predicting the occurrence of wax formation in crude oil pipelines.