Characterization of In Situ Stress State and Joint Properties from Extended Leak-Off Tests in Fractured Reservoirs

Abstract

Characterization of geomechanical parameters in naturally fractured reservoirs remains one of the most challenging tasks in civil, mining, and petroleum engineering. Extended leak-off tests (XLOTs) are generally carried out in new wells to obtain in situ stresses for hydraulic fracture-treatment design and well-trajectory optimization in petroleum engineering. The largest and smallest principal in situ stresses can be calculated by shut-in/closure pressure and breakdown/reopening pressure of XLOTs. However, in situ stresses obtained from XLOTs in the traditional theoretical framework are not completely correct because XLOTs still keep the same test collocations as leak-off tests. In addition, the traditional method cannot be used to simultaneously calculate other parameters beyond in situ stresses. Given these challenges, a hybrid artificial neural network (ANN)–genetic algorithm (GA) method was tested for identification of the principal in situ stresses and joint parameters. First, XLOTs were performed to generate samples for an ANN. The ANN model was then applied to map the nonlinear correlation between geomechanical properties and pressures. Finally, a GA was used to identify geomechanical properties on the basis of the fitness function established using pressures of XLOTs. The results indicate that the inverse-analysis model of pressure established by the ANN–GA provides a powerful and effective tool for multiparameter identification, and it is also a cost-saving and time-saving method.