Abstract
Effective oil reservoir management aims to maximize profit by navigating operational, technological, economic, and regulatory constraints. This management is inherently linked with production infrastructure and spans the entire reservoir life cycle — from exploration and production to revitalization. To streamline reservoir studies, the infrastructure is mapped into project variables (PV), control variables (CV), and revitalization variables (RV). Currently, short-term decisions in the management of oil and gas capture valves, or fluid injection, are often made without a formal procedure, which can lead to suboptimal production. Leveraging historical production data for future decision-making can support more effective management. This study introduces a data mining-based methodology to derive control rules for managing CV. Data simulated from a reservoir model, representing actual reservoir conditions, was utilized for development. Experiments were conducted both to validate the control rules obtained by the proposed methodology, using a numerical simulator, and to evaluate the performance of the rules when applied to the management of control variables. The simulation results indicated that the control rules derived from the proposed methodology enhanced the economic production indicator for one-third of the evaluated wells. It is noteworthy that the initial set of rules, generated by the simulator, was already the result of a search and optimization process. Furthermore, the application of these rules to well control management proved effective across all tested wells. Comparisons with classical optimization methods further indicate that the proposed approach delivers the most consistent performance across wells. The results reported in this work provide evidence in favor of the proposed methodology and can be considered an initial step towards generating computationally efficient and interpretable control rules applicable in practice for short-term decision-making.