Method and system for predicting the lifespan of electric submersible pumps using random-forest machine-learning

A method comprising:

• disposing an electric submersible pump (ESP) inside a wellbore,
• wherein the ESP comprises an electric motor that provides an artificial lift to hydrocarbon production from a hydrocarbon reservoir to a well surface;
• obtaining wellhead data from a wellhead associated with the ESP using a well monitoring and control system connected to the wellbore and a plurality of sensors,
• wherein the wellhead data comprises wellhead temperature data and wellhead pressure data;
• obtaining ESP operational data from the ESP,
• wherein the ESP operational data comprises voltage data for the ESP, current data for the ESP, power data for the ESP, motor temperature data for the ESP, discharge pressure data for the ESP, vibration data for the ESP, and frequency data for the ESP;
• predicting, by a computer system, a remaining expected life of the ESP using a machine-learning model, the wellhead data, and the ESP operational data,
• wherein the machine-learning model is a random forest model that performs a regression using the wellhead data and the ESP operational data to predict a timeline of a failure event for the ESP, and
• wherein the random forest model comprises a plurality of decision trees that are trained using bootstrap aggregating;
• and
• repairing the ESP based on the remaining expected life of the ESP.