A Bayesian approach to the dynamic modeling of ESP-lifted oil well systems: An experimental validation on an ESP prototype

Abstract

This article presents an integrated method for estimating parameters for an electric submersible pump system with process variables data. Validation of a phenomenological model is also performed. The parameters and the associated probability density function are obtained through Bayesian inference, and the model validation is achieved in two stages. The first one is the validation of the dynamic response in which the model is compared with the experimental data. The second is achieved by comparing the regions covered by the experimental data and the model, both in steady-state. The experimental data's uncertainty is assessed using the Guide for the Expression of Uncertainty in Measurement. In turn, the uncertainty of the model's prediction is obtained by propagating the probability density function parameters. The results indicate that the method can provide a model to represent the system behavior within the existing uncertainties. Additionally, the procedure can be applied in oil production fields to provide substitute models for general purposes, such as production control, optimization, and assistance.