Operational Control of an Electric Submersible Pump Working with Gas-Liquid Flow using Artificial Neural Network

Luis R. Pineda, Alberto Luiz Serpa, Jorge Luiz Biazussi, and Natache A. Sassim

Keywords

Electric Submersible Pump, Surging Point, Gas-Liquid Flow, Artificial Neural Network, Direct Inverse Control

Abstract

The performance of centrifugal pumps depends mainly on the intake pressure and the rotational speed. In the case of electric submersible pumps working with both gas and liquid, the performance also depends on the gas amount that can cause operation instabilities and a pronounced drop in the pump pressure increment for lower flows than the surging point (point of maximum pressure in pressure increment versus liquid flow rate curve). So, it is important to find the condition that surging occurs and keep the pump operating in a considered safe region. In this paper, a control technique to keep the pump operating in the desired point is proposed considering different intake conditions. For this, two artificial neural networks are used: the first to model the ESP behavior and another to control the pump operation. To control, the Direct Inverse ControlMethod is used considering the rotational speed as the control input. Data used to train and verify the artificial neural networks behavior were obtained from experimental tests of a pump operating with water-air flow.

Important Links:

Go Back