Application of Machine Learning and Dimensional Analysis to Evaluate the Performances of Various Downhole Centrifugal Separator Types

Abstract

Abstract Pumping artificial lift techniques, such as rod pumps and ESPs, are applied for gassy wells more than ever before. Gas largely affects pump performance, causing cavitation, gas locking, low production, and run-life shortening. These phenomena have made downhole separators a critical part of such installations. There are multiple categories of downhole separators, with various techniques developed to assess and improve their performances, but no general guidelines are established for their application. Through dimensional analysis and machine learning techniques, this paper aims to establish key parameters in the performance characterization of this equipment, and prediction models are attempted. A comprehensive literature review is conducted to understand factors (retention time, turbulence, and head) influencing the efficiency of this equipment and collect the available downhole separator performance data. This information is collected to identify the optimum conditions for each separator type, considering the effects of liquid and gas rates and other flow parameters. The data collected from various research projects over the last 20 years are combined to make a comprehensive centrifugal downhole separation databank. The available data show that most separators provide separation efficiencies higher than 79% if the downstream pressure-controlling systems and liquid levels are adequately monitored. The separation efficiencies decline as the liquid and gas rates increase past an upper limit of around 800 BPD. Furthermore, the separator's control system is a critical factor that adds to the uncertainty, significantly affecting the separation efficiency calculation. Data analysis of the Buckingham Pi Theorem for dimensionless numbers is used to compare the performances of different downhole separator classes, and machine learning is applied to identify a robust prediction model. Two dimensionless numbers, the Weber number, and the gas Reynolds number, were found to largely dominate the behavior of the separators’ performance. Results provide a fundamental source and a valuable guideline for downhole liquid-gas separation, particularly in artificial lift applications.