Dynamic parameter estimation and uncertainty analysis of electrical submersible pumps-lifted oil field using Markov chain Monte Carlo approaches

Parameter estimation is an important part of model fitting in the design and validation of a first-principle oil well model, which describes pressure and flow dynamics in an oil well system. In particular, the estimation of dynamic parameters remains technically challenging. Meanwhile, the estimation of uncertainty in the prediction of oil production has long been a question of great interest in the field of oil operations. This study set out to estimate the parameters and uncertainty of oil wells lifted by electrical submersible pumps (ESP) using Bayesian analysis. A surrogate model was utilized as a simulator to generate measurements containing white noise. The study began with a case study where the parameters of the oil well were constant. Based on some results of the constant parameter estimation, we then tested the estimation of dynamic parameters. Multiple Markov chain Monte Carlo (MCMC) algorithms were compared for the estimation of constant parameters. One of the algorithms was chosen and combined with the moving window technique for dynamic parameter estimation. The estimation result is presented using trace plots, density plots, mean plots, and corner plots. The estimation method is analyzed and compared according to the diagnostic criteria and autocorrelation plots. This proposed scheme offers a fresh perspective on dynamic parameter estimation for oil wells in real time and potentially benefits the control and operation of oil and gas fields.