DEVIATION OF HYDRAULIC FRACTURES THROUGH POROELASTIC STRESS CHANGES INDUCED BY FLUID INJECTION AND PUMPING

This paper presents an analysis of the deviation of hydraulic fracture s associated with the poroelastic change of the in situ stress field c aused by fluid injection and pumping in the reservoir. This mechanism is studied within the confines of a simple model involving one injecti on and one pumping well, and a hydraulic fracture propagating along th e path initially equidistant from the two wells, Analysis of the fract ure deviation from its straight-ahead path and determination of the co nditions leading to attraction of the fracture by the injection well a re both based on a theoretical study of the stress trajectories. Compa rison of the analytical prediction of the fracture path with the compu ted path using a numerical technique shows excellent agreement between the two methods, provided that a certain dimensionless toughness is s mall. The principal results of this study can be summarized as follows . First, a fracture propagating along a path, initially midway between an injection and a pumping well, will always be deviated by the injec tion well due to the shear stress induced by fluid injection and pumpi ng along the initial path. Second, the injection well acts as an attra ctor of hydraulic fractures propagating within its ''attraction basin' '. Then, the fracture will propagate toward the injection well rather than simply be deviated by it, One of the features of this attraction basin is the existence of a ''fracture barrier'' characterized by a 90 degrees rotation of the principal stress directions, with respect to far-field principal directions. Third, fracture deviation and attracti on towards the injection well appears to be primarily controlled by on ly two dimensionless quantities: Pi = S-0/sigma, the ratio of the stre ss deviator at infinity over the characteristic stress sigma associat ed with injection and pumping of fluid; and tau = 4ct/L-2, a dimension less time (where c is the diffusivity and L is the half-distance betwe en the two wells). The number of significant parameters is thus remark ably less than expected from a dimensional consideration. Only in the region close to the injection well is there an influence of an additio nal number Pi(1) = (P-0 - p(0))/sigma, the ratio of the difference be tween the mean pressure at infinity and initial pore pressure to the c haracteristic stress sigma. (C) 1997 Elsevier Science Ltd.