Artificial Neural Network Based Identification of the Gas Volume Fraction in an Electrical Submersible Pump

Carlos Uriel Cortés Rodriguez, Alberto Luiz Serpa, and Jorge Luiz Biazussi


Electrical Submersible Pump, Artificial Neural Network, Volume Gas Fraction, Back Propagation, Identification


The electrical submersible pump (ESP) under multiphase flow is very common in the oil industry. These pumps present frequent premature failures when the gas flow is high. In addition to this, a further increase of the gas may fill most of the pump impeller, making the flow rate to decrease down to zero, known as \textit{gas locking}. Due to lack of information and mathematical models that can be used in real time for this type of pumps, experimental studies are usual in this area. This paper applies artificial neural network (ANN) modeling for the volume gas fraction identification in ESP. The algorithm uses experimental data collected directly from the system for different gas fractions, such as pressure, flow rate, mechanical torque, elevation, etc. This model uses a back propagation learning algorithm and multi-layer perceptron neural network, where different structures are analyzed to find the optimal number of hidden layers. Results show that the system is able to identify the volume gas fraction in the pump with a very good accuracy.

Important Links:

Go Back