(721g) Prediction of Mineral Scales in Oil & Gas Production Systems Using a Machine Learning Approach | AIChE

(721g) Prediction of Mineral Scales in Oil & Gas Production Systems Using a Machine Learning Approach

Precipitation of mineral scales is one of the most sedulous and expensive operational problems in Oil & Gas Industry, because mineral scales may lead to severe permeability loss, reduces flow areas of productions and injections wells and breakdown of production facilities equipment generating a hydrocarbon production drop and high maintenance costs. In this context, the first practical step is to obtain an accurate estimation using a reliable method able to identify which mineral scales and maximum amount are going to precipitate under different reservoir operating conditions (temperature, pressure and composition). It is also known that thermodynamic ion interaction models and equations of state, used to determine behaviour of different ions and gas compounds leading to mineral scale formations, are limited by the ranges of temperature and pressure on which they can be used, making the calculation very complex and allows high error values. As a response, in recent years, many works have been focused on the development of a simple and reliable method that allows the prediction, with high accuracy, of the inorganic scales behaviour under wide ranges of operating conditions, preventing production and economical issues.

In this study a reliable model based on machine learning techniques (Least Squares Support Vector Machines and Coupled Simulated Annealing) was developed aiming to predict the type and the amount of mineral scales precipitated and compared with thermodynamic ion interaction models (Debye-Hückel, Pitzer, Extended UNIQUAC and e-NRTL), equation of state (Soave-Redlich-Kwong) and finally with commercial softwares (ScaleChem and DownHole SAT) predictions. Based on the results shown, the model proposed is adequate for mineral scales prediction with higher accuracy than thermodynamic models, equations of state and softwares prediction under a wide range of operating conditions, resulting in the reliable and correct prediction of mineral scales precipitation in Oil & Gas processes.