Eventos Anais de eventos
COBEM 2017
24th ABCM International Congress of Mechanical Engineering
USE OF ARTIFICIAL NEURAL NETWORKS FOR IDENTIFICATION OF THE ELECTRIC SUBMERSIBLE PUMP PERFORMANCE WORKING WITH A VISCOUS FLUID
Submission Author:
Luis Felipe Barrera Salamanca , SP
Co-Authors:
Alberto Luiz Serpa, Natache Sassim, Jorge Luiz Biazussi, William Monte Verde, Luis Felipe Barrera Salamanca
Presenter: Luis Felipe Barrera Salamanca
doi://10.26678/ABCM.COBEM2017.COB17-0304
Abstract
The methods of artificial lifting are fundamental in the petroleum industry for initiating or incrementing the production of wells that have insufficient reservoir energy to raise the fluids to the surface. In this case, the use of Electric Submersible Pump (ESP) is very important. For the petroleum industry, it is important to determine the ESP’s performance working with viscous fluids. Therefore, in this paper, an Artificial Neural Network is proposed to identify the behavior of pump working with glycerin. Two different ANN structures were tested to identify the ESP’s performance, the Multilayer Perceptron, and the Neural Network Finite Impulse Response, and the best structure was chosen through various validation tests. This paper also proposes to compare the response of the best ANN structure with a correlation proposed by Biazussi on ESP working with viscous fluid and low viscosity in single-phase flow.
Keywords
Artificial neural networks, Electric Submersible Pump, dimensionless numbers, Neural Networks Finite Impulse Response, Multilayer Perceptron
