PARTICLE FILTER-MODEL PREDICTIVE CONTROL FOR OIL RESERVOIR MANAGEMENT
Résumé
In this work, the application of a Nonlinear Model Predictive Control (NMPC) with state estimation in the waterflooding process as supervisory control was studied. The methodology employed the Particle Filter (PF) in step of the model-based optimization of the NMPC with Particle Filter Optimization (PFO) and step of the filtering through the Auxiliary Particle Filter (APF). This way, it is presented a new Bayesian approach to NMPC for reservoir management. The proposed methodology aimed at being robust to overwhelm nonlinearities and non-Gaussian uncertainties inherent in oil reservoir systems. The results showed the potential of APF in the estimation step not presenting degeneration or impoverishment of the sample, in view of that the dimension of the problem is considerably high, and, in the model-based optimization step with PFO which handled well the nonlinearity of the model and maintained control of production. The results also supported the contribution of this work with advances in the application of PF as a tool in the context of reservoir management and especially in its use as an optimization tool.