Machine Learning Model for Efficient Surveillance of Electrostatic Pump Operations & Predictive Analytics for Early Alarming System

Abstract

Abstract In today's fast-paced technological world, Oil and Gas E&P industry must embrace innovative solutions for boosting efficiency and minimizing expenses. One promising direction involves investigating smart systems that can supervise, notify, and foresee the malfunction of electrical submersible pumps (ESPs). By leveraging artificial intelligence (AI) and machine learning (ML), we can transform how we handle the upkeep and administration of these vital elements. Adopting AI and ML-driven solutions offers significant advantages, such as preventing expensive failures and promoting eco-friendly resource management. By detecting and resolving potential issues early, we can extend equipment life and reduce our operations’ environmental footprint. The widespread application of Electrical Submersible Pumps (ESPs) in Kuwait Oil Company (KOC) has recently experienced a surge in pump failures due to increased water production from maturing fields. This issue prompted the exploration for adopting intelligent systems, and integrating AI and IoT, to improve ESP monitoring and predictive maintenance, ultimately enhancing efficiency and reducing costs. The study employed a novel workflow connecting field, well, and detailed ESP operational data, such as intake and discharge pressure, wellhead and flowline pressure, motor current and voltage, motor temperature, reservoir temperature, and vibration for failure diagnosis. Predictive analytics were utilized to proactively identify potential issues before they manifested, thereby mitigating the impact of these failures. The study results show that utilizing advanced technologies such as Machine Learning and Automatic Surveillance Systems can significantly reduce ESP failures. Key conclusions from the analysis include, A streamlined solution is provided to identify and predict ESP failure causes. ESP real-time data is digitized and converted into a computationally friendly form. Early anomaly detection enhances well uptime and minimizes ESP failures. Model parameters can be fine-tuned for better results with a successful proof of concept (POC). The proposed workflow aids in cost reduction for workovers and maintains production. Decision-making is improved through the monitoring system. This approach offers an accessible and user-friendly platform for non-specialists within organizations. Oil companies can adopt this workflow to boost well performance and pinpoint areas for improvement. Ultimately, implementing digital transformation will elevate organizations’ competitiveness in the global industry.