Controle preditivo com aprendizado por reforço para produção de óleo em poços inteligentes

This work presents the modeling and development of a methodology based on Model Predictive Control with (MPC) applied to the control of oil production in an oil reservoir with existing production and injection wells. The MPC strategy is based on a machine learning model - Reinforcement Learning (Reinforcement Learning) - as the method of searching the optimal control policy. The experiments were carried out in an oil reservoir with synthetic valve actuators that are 3 water injections. Therefore, the action is performed by injecting water rates for certain time intervals. The output variables of the field are: average pressure of the reservoir, the daily production of oil, gas, water and water cut. The forecast of these variables is accomplished by a proxy, which is a model identification og the plant implemented by neural networks. The results indicate that the proposed model is capable of controlling oil production even with disturbances in the producing well, for different reference values for oil production.