A Comprehensive Approach to Organic Precipitation Damage by CPA EoS from Monte Carlo, and Machine Learning Methods

Abstract

Summary In this study, an integrated machine learning (ML) model was proposed that allows to identify the risk of organic precipitation damage and estimate the asphaltene onset pressure (AOP). In addition, an estimation of the association parameters to estimate the AOP using a Cubic-Plus-Association (CPA) equation of state (EoS) using stochastics (Monte Carlo) and ML approach was carried out. To predict the asphaltene damage risk the asphaltene stability class index (ASCI) data and the in-situ live crude oil densities were used along with the support vector machines (SVM) method. To propose the AOP-ML model a dataset of 53 samples was considered, evaluating different ML methods. In both cases, 80 % of the dataset was used to train the model, whereas 20% was to validate it. In the Monte Carlo (MC) simulations, 6 fluids taken from literature were used. The ML classification model had a perfect accuracy (100 %), which was compared to conventional compositional asphaltene screening models, with a classification accuracy of 33% for the resin/asphaltene ratio, 29% for ASI, 67% for CII, and 88% for de Boer plot. The AOP-ML model described properly the 77% of the variation of the experimental AOP of the 6 fluids evaluated using a stepwise bidirectional linear regression with 9 input features. Finally, the MC results indicated that several combinations of association energies and volumes reproduce the experimental AOP, obtaining a linear model for estimating the cross-interaction energy with a coefficient of determination of 0.934. This study provides disruptive findings since it opens the possibility of formulating predictive EoS, obtaining the association parameters from a fluid's compositional and structural characteristics. This approach is an opportunity for a comprehensive understanding of asphaltene precipitation damage that allows to understand the mechanisms of formation damage and therefore look for promising solutions to restore the productivity of fields affected by asphaltene precipitation formation damage.