Identifícaction of Donwnhole Dynamometer Charts Using Neural Networks as a Tool for Help in Mechanical Pumping

Abstract

Successful identification of downhole problems is essential in mechanical pumping to achieve optimum production and to minimize operating and maintenance costs. It is necessary to develop and apply methodologies that allow fast identification of problems which affect production. The mechanical condition and performance of the downhole equipment (rod string, pump, valves, etc.) and physical properties of well like pump submergence, gas interference, pump leaks, etc, may be evaluated using dynamometer charts. In this work a fault analysis and operation conditions methodology for pumping system is presented; this includes the development of analysis software based on neural networks to identify system problems using downhole dynamometer charts. The developed system allowed to identify a set of the most common problems with a high precision and is a tool that could assist engineers and operations personnel in day to day oilfield works in rod pumping. Additionally, a basic functionality was developed to identify nearest charts according to statistical and geometric features, which can be used as the starting point to develop a smart system to predict potential faults in the future.