Predictive Electrical Submersible Pumps ESPs Performance for better production sustainability using Machine Learning

Abstract

The electrical submersible pump (ESP) is considered one of the most important and rapidly growing types of artificial-lift pumping technologies. Utilized in 15–20 percent of oil wells globally, ESPs offer an effective solution at high production volumes and great depths. The performance of ESPs often gradually declines and can eventually lead to service interruptions due to factors such as high gas volumes, elevated temperatures, and corrosion. The failure of an ESP results in a significant financial impact due to lost production and the costs of replacement. Electrical Submersible Pump (ESP) failures are unexpected. To avoid excessive downtime, early failures identification is essential. This study suggests an innovative strategy that use a comprehensive dataset and various machine learning algorithms to achieve this goal. The Machine Learning models are based on the data collected from surface and downhole ESP monitoring equipment from four (04) wells, Several ML models are tested and evaluated using the K Nearest Neighbour (KNN), Random Forest (RF), Decision Tree (DT), etc. This approach is essential for helping operators to move from reactive to proactive and predictive maintenance.