Experimental study and machine learning modeling of water removal efficiency from crude oil using demulsifier

Area of Science:

• Petroleum Engineering
• Chemical Engineering
• Process Optimization

Background:

• Crude oil processing requires efficient water removal to meet quality standards and prevent operational issues.
• Demulsifiers are crucial chemical additives used to break oil-water emulsions.
• Understanding the interplay of operational parameters is key to optimizing demulsification.

Purpose of the Study:

• To investigate the impact of time, demulsifier concentration, and temperature on water removal efficiency (WRE) from crude oil.
• To develop predictive models for WRE using soft-computing techniques.
• To perform a sensitivity analysis to identify critical operational factors.

Main Methods:

• Experimental study of WRE under varying conditions: time, demulsifier concentration (up to 40 ppm), and temperature.
• Development and validation of predictive models using Multilayer Perceptron (MLP) and Gaussian Process Regression (GPR).
• Contour diagram visualization and sensitivity analysis using the MLP model.

Main Results:

• Temperature directly correlates with and significantly enhances WRE.
• Optimal demulsifier concentration improves WRE; higher concentrations lead to overdose and increased consumption.
• MLP model demonstrated high accuracy (0.84% average absolute relative error) in predicting WRE, outperforming GPR.
• High temperature and concentration reduced the time needed for effective demulsification.

Conclusions:

• Temperature and demulsifier concentration are critical factors influencing WRE, with an optimal concentration range.
• Soft-computing models, particularly MLP, accurately predict WRE, aiding in process design.
• The study provides valuable insights for optimizing crude oil demulsification processes.