NEURAL NETWORK FOR THE DIAGNOSIS OF ELECTRICAL SUBMERSIBLE PUMP FAULTS
Electrical Submersible Pump. Neural Network. Pressure on the Head. Fault Detection. Industry 4.0.
In artificial elevation, automation methods are being used in order to increase the efficiency and production of oil wells. In the Electrical Submersible Pump (ESP) this becomes essential, since the analysis of the available data is still insufficient to monitor, diagnose, interpret and analyze the reservoir performance, well integrity, ESP operation and efficiency in real time. In addition, traditional amperometric chart diagnostics cannot identify pipeline leakage, poor pump lift and low pump operating efficiency. Thus, this work aims to provide a proposal for detecting problems related to gas interference and pump gas block from the study of the current based on the ameprimetric charts. This will be done with neural networks using the BackPropagation (BP) algorithm in the Python programming language in the Google Collaboratory environment. In addition, a pressure analysis at the wellhead will be proposed in order to confirm what was obtained by the neural network and to identify where the possible failure occurred and, therefore, to propose a more adequate solution. This model is a reliable complement to the traditional method of diagnosing amperometric charts, as well as providing additional flexibility to cooperate with field operations in real time. The early identification and resolution of ESP problems can lead to great cost savings and less maintenance requirements due to this intelligent system.