Estimation and prediction of motor load torque applied to electrical submersible pumps

Abstract

This work presents a methodology to model a time series of electrical current using state space approach that aims at estimating and predicting the load torque of electrical submersible pump motors, in addition to classifying the oil well dynamic behavior. For this purpose, the historical values of the stator current of 9 motors from subsea well production were used. To perform the validate the modeling and prediction approach test, one motor is used to identify the model parameters. In sequence, the model is compared with others 8 real motors. The model is also compared with established MATLAB/Simulink motor model. Finally, experimental test using three equipment in operation demonstrated the viability of the method. The torque estimation reached an average absolute percentage error of 2.73% and 5 min forecast presented the absolute percentage error below than 1%. With these results, it was possible to predict variations in the load of an oil well.