Predicting Non-Newtonian Fluid Electric Submersible Pump failure using Deep Learning and Artificial Neural Network

Abstract

Abstract The monitoring of electric submersible pumps (ESPs) is essential for optimal petroleum artificial lifting operations. Most ESP research are aimed at operation improvement and optimization of the centrifuge multi-stage pump motor and the load that the pump has to discharge which is a function of the pumps mechanical properties and characteristics, liquid compositions, pressure and temperature. ESPs failure often lead to oil production losses or “oil deferment” which affects revenue for all the parties involved. Also, pulling the ESP out of the wellbore of interest, requires mobilization of a rig because it is installed several hundred meters down the wellbore. To prevent these loses, a predictive approach is needed to avert these scenarios. In the current decade, machine learning algorithms studies have spurred real- time technologies research interest due to their abilities to predict future outcomes using already existing data sets. This study presents a predictive approach for Electric Submersible Pump failure during artificial lift operations. The study creates an “algorithm” that helps to predict via Machine learning, the failure of an ESP with the assumption that failure is usually caused by pressure build-ups. A deep learning model for predicting ESP failure was proposed and artificial neural network was used in developing the suggested model. Based on the outcomes of this study, it can be concluded that the selected AI algorithm and its characteristics, are suitable for applications in detecting ESP failure before it happens using upstream-data.