Accurate and low-cost information on in-situ stresses and fracture properties is critical in reducing well costs and increasing well recoveries. In this paper, a hybrid model based on the displacement back analysis is proposed for determining the in-situ stress magnitudes and fracture properties at the wellbore scale using wellbore displacements. The new methodology is an integration of artificial neural network (ANN), genetic algorithm (GA), and numerical analysis. An ANN is used to map the non-linear relationship between the maximum and minimum horizontal in-situ stresses (
) and natural fracture properties (e.g., joint angle,
, and spacing,
) and the wellbore displacements. A forward modelling (UDEC) is used to compute wellbore displacements as a function of horizontal in-situ stresses and natural fracture properties, and to create the necessary training and testing samples for ANN. The set of unknown horizontal in-situ stresses and natural fracture properties at wellbore scale are searched in a global space using GA based on the objective function. Results of the numerical experiment show that the hybrid ANN-GA model based on the displacement back analysis can effectively recognise the horizontal in-situ stresses and natural fracture properties from wellbore deformation during drilling.
Keywords: petroleum geomechanics, in-situ stress, natural fracture properties, wellbore displacements, back analysis, artificial neural network, ANN, genetic algorithm