Real-time Electrical Submersible Pump Smart Alarms Suite Enabled Through Data Analytics and Edge-based Virtual Flowmeter

Abstract

Abstract Continuous monitoring of electrical submersible pumps (ESPs) ensures optimal working operating conditions and avoids deferred oil production. With the increased population of ESPs deployed worldwide, a comprehensive alarm triggering system is at the center of modern oilfield production surveillance systems. In this paper, a data-driven ESP smart alarm suite, integrated with a virtual flowmeter (VFM) deployed on the Edge, is presented with testing results and field applications for effectiveness demonstration. The overall Smart Alarm Suite consists of eight different alarms, each targeting a specific potential suboptimal pump working condition. For three alarms, a data-driven approach is adopted with the application of multiple classical machine learning models such as logistics regression, K-Means clustering, continuous linear regression, etc. The workflow also uses hybrid pipelines when manual labeling is not sufficient for model training. For the other five alarms, the workflow uses an Edge-enabled physics-based flowrate calculation to flag the status of different alarm categories. The presented workflow innovatively combines machine-learning algorithms with rule-based criteria and a robust VFM to raise awareness of the ESP performance conditions. The injection of domain expertise and physical modeling into AI-based workflows improves the robustness of the algorithm and reduces false alarms with limited data exposure. The completeness of the alarms promotes more comprehensive monitoring of the assets and reduces the risk of lost production due to ESP failure or downtime.