The purpose of oil production strategy in the context of production optimization is to find the best configuration of wells that contributes to maximizing the Net Present Value. This value is calculated mainly from the amount of oil, gas, and water recovered from the field, which can be obtained by running the reservoir simulator. However, many parameters and variables must be prefixed and inserted into the simulation system in order to generate these production values. This process involves a high computational cost for modeling the transfer of fluids within the simulated reservoir. Thus, the use of simulators may be substituted by approximators. In this thesis, we aim to develop these approximators using Quantum-Inspired Linear Genetic Programming, a technique of Evolutionary Computation. These approximators were used to replace the reservoir simulation in the process of optimizing the location and type of wells to be drilled in a field. For the reservoir proxies construction, samples obtained from the technique of Latin Hypercube were simulated to create the database. The model for creating approximators was tested on a synthetic reservoir. Two types of optimization were performed to validate the model. The first was a deterministic optimization and the second an optimization under uncertainty considering three different geological settings, a situation in which the number of simulations becomes extremely high. Our results indicated that the model for the creation of proxies achieves a satisfactory performance in the replacement of simulators due to low levels of errors and a considerable reduction of the computational cost.