Perímetro urbano de Pueblo Nuevo y Muchilena

Davies et al. (1998) investigated the relationship between net effective stress and in-situ permeability and noted that a significant decline in matrix permeability of tight gas sands can be seen as a result of an associated effective stress increase and depletion. They concluded that the rate of change of permeability with increasing values of stress is a function of the pore geometry.

Davies et al. (1999) investigated the decline in productivity during early production of idealized wells and illustrated that it is directly related to the impact of stress-sensitive matrix permeability. Their results concluded that, in unconsolidated reservoirs, a significant reduction of permeability occurs in the highest initial reservoir quality sands; sands with relatively low initial reservoir quality are less stress sensitive. Whereas in consolidated reservoirs, the loss of permeability occurs with increasing stress in sands with slot pores (tight-sand pore geometry consists of ultrafine micro porosity in an all-pervading matrix, long narrow tube-like). This stress sensitivity highlights the need to test tight gas samples at reservoir stress conditions in

order to obtain representative porosity-permeability relationships for correct reservoir modelling.

Holditch (2006) and Abass et al. (2009) noted that the small pore and pore throats in tight gas sands make these formations more stress sensitive compared to conventional, high- permeability sandstone reservoirs. The change in permeability associated with depletion has been explained through the unfavourable aspect ratio of the pores in these sandstones, which would usually be lenticular-shaped, with the smallest axis in the vertical direction. An increase in effective stress causes flattening and closure of these pores.

Walsh (1981) defined that a fracture is two rough surfaces in contact, and that flow can only be achieved when the viscous drag of the fluid between the two narrow surfaces overcomes the resistance to flow and the tortuosity of the flow path. He then derived an expression for fluid flow through fractures as a function of the external confining pressure and the pore fluid pressure and concluded that the effect of aperture and tortuosity are the two factors that control flow rate.

Ostensen (1983) extended the work of Walsh and introduced a theory of permeability based on flow through cracks. He presented an extensive comparison of data and theory that provided support to the Gaussian crack model and indicates that the permeability of tight sand cores is a result of micro cracks. The theory predicted that the square root of permeability should decrease linearly with the log of net confining stress. El Rabaa (1989) conducted an experimental study of hydraulic fracture geometry initiated in horizontal wells and identified that in vertical wells a single planar fracture may occur, whereas in horizontal wells non-planar fracture geometry patterns occur as multiple fractures away from the well, and their shapes and size will be dictated by the well inclination and azimuth relative to σmax, the maximum compressive stress.

Field experiments were carried out on tight sandstone (surface mining) to measure hydraulic fracture growth in naturally fractured rock (Jeffrey et al. 2009). Tilt meters and micro seismic arrays were installed in the zone adjacent to where a hydraulic fracture was initiated. The layer in which the fracture was initiated was then revealed by mine-through mapping (i.e., mining a tunnel across the fractured zone). The purpose was to determine the fracture geometry; visual results indicated that the induced hydraulic fracture intersected with pre-existing natural fractures, shear zones, and veins that were adjacent to the hydraulic fractures. The hydraulic fracture trace was seen to propagate through solid rock and step along shear zones.

2.5 Summary

Whereas each of the three simulation tools has unique advantages and applications, the choice for this research was to use finite-difference simulation as a primary proof of concept to investigate the areas and regions close to the wellbore in tight sandstone reservoirs with induced hydraulic fractures. Secondly, the finite-element reservoir simulator was chosen as a means to effectively simulate the natural fracture network that is pre-existing in the reservoir. This will enable implementation of the dynamic fracture network (DFN) concept and synthesis of the properties of the fractures and the effect of the induced hydraulic fracture on the natural fracture network in an attempt to investigate and measure the effect the induced fracture may have on well test simulation results.

Modelling of tight, naturally fractured reservoirs is a relatively complex endeavour due to the existence of uncertainties, mainly due to the lack of knowledge of the distant extent of the DFN and of the matrix as well. The issue becomes even more complex when it comes to modelling the tight, naturally fractured reservoir post induced hydraulic fracture. The complexity is further increased because the system has essentially gone past the elastic state and crossed through to the plastic state of deformation.

In essence, there are at least three main factors that need to be taken into consideration, namely the mechanical deformation that is caused by the fluid pressure and the fracture propagation into virgin matrix as well as into the DFN, the flow, and proppant distribution within the fracture itself. Given that the entire process is actually induced in dynamic and in- situ conditions, attempting to model the process is not easy. The simulation typically will require the coupling of implicit fracture network modelling, the geomechanical in-situ stresses, and the rock properties as well as the dynamic induction of the hydraulic fracture. The irregular shape of the natural fractures and their complex intersection with the matrix material and each other creates a complex network that is best modelled with the use of the mesh approach. However, the wide range of fracture size and geometry still makes it difficult to physically model their existence in static conditions, let alone to predict their behaviour in dynamic conditions.

In this research study, we are modelling the process and evaluating the modelling results with well test data that is a measure of the in-situ state of the reservoir. And the well test process itself is a reflection of both a dynamic flow process that is followed by a static shut in process. Thus this is considered to be the best way to evaluate the modelling technique.