Intelligent Fault Detection in Electrical Submersible Pumps Systems Using Real-Time Oilfield Data

Abstract

Electrical Submersible Pumps (ESPs) are among the most extensively utilized artificial lift systems in crude oil production, particularly in high-rate wells. Their capability to efficiently extract large volumes of fluid from medium to deep reservoirs plays a critical role in enhancing hydrocarbon recovery and sustaining production. Despite their advantages, ESPs are exposed to harsh downhole conditions and are subjected to continuous mechanical, thermal, and electrical stresses. These operating challenges often lead to a range of faults, including shaft breakage, motor overheating, insulation degradation, seal failure, gas locking, and pump wear. Undetected faults can result in unplanned shutdowns, reduced equipment lifespan, and significant production losses. This paper presents a fault detection approach using Principal Component Analysis (PCA), an unsupervised machine learning method well-suited for multivariate process monitoring. The methodology was applied to operational data from three ESP wells located in the Meleiha field, operated by Agiba Petroleum Company in Egypt. The results confirm that PCA effectively detects deviations from normal behavior, enabling timely fault diagnosis and supporting predictive maintenance decisions.