Estimating geomechanical parameters from hydraulic fracturing tests using a soft computing-based methodology

Resumo

The discovery of naturally fractured reservoirs in the Brazilian pre-salt has attracted considerable attention for a better understanding of reservoir characterization and fluid flow inside fracture channels. Predicting the hydromechanical behavior of these reservoirs is a cumbersome task, which requires the identification of their geomechanical parameters. In this scenario, a soft computing-based methodology is implemented to estimate geomechanical parameters from borehole injection pressure in hydraulic fracturing tests. Based on artificial intelligence techniques, this approach integrates a proxy model and an optimization algorithm to match the field measurements and the borehole pressure curve predicted by a finite element model. Considering a multistep-ahead strategy to predict time series, a multilayer perceptron-based proxy model computes the borehole pressure curves, substituting the numerical simulation of a minifrac test. The adoption of a proxy model substantially reduces the computational effort of the parameter identification task. Therefore, a genetic algorithm can efficiently estimate the reservoir geomechanical parameters by solving a nonlinear least squares problem. The application to field-measured data from a minifrac test confirms the capability of the proposed methodology to estimate geomechanical parameters from hydraulic fracturing tests.