**5. Discussion**

#### **5.1. Fluid-driven fracture nucleation, growth and connection**

In the model, we only consider hydraulic fracture growth through a finite set of fractures, some of which are initially very small, but potentially provide a conduit with the help of high fluid pressure. The fracture seeds are pre-assumed in this paper and are represented by these small fractures. Therefore, fracture nucleation in a highly stressed area is not dealt with in this paper. This treatment of fracture nucleation can underestimate the fracture number and the fracture connectivity. For the case shown in Figure 4, the lower fluid flow path cannot move to the right outlet and a higher stress level might create new fractures near the entry zone. It would be interesting to study the impact of crack nucleation based on stress conditions rather than just from these pre-existing fractures.

direction of the maximum stress. This orientation reduces the viscous dissipation and injection pressure. However, an increase in fluid pressure because of increased rate or viscosity can produce higher pressures upstream of a local offset or restriction which can then lead to

Development of Fracture Networks Through Hydraulic Fracture Growth in Naturally Fractured Reservoirs

http://dx.doi.org/10.5772/56405

153

In addition, this paper only considers a limited number of specific initial fracture geometries, the results may be different if the starting geometries are changed and more cases are being considered as a way of making our conclusions stronger and more general. Although a method to deal with network development is presented here, there is a need to work on different geometries to extract some useful general responses for rational simplifications of the expected

Our numerical results quantify the overall path of discrete hydraulic fractures growing through a network of pre-existing natural fractures, as shown in Figs. 3 and 4. These results give insight to the behaviour of fracture-controlled flow systems, where the fluid flow and the rock deformation and fracturing are strongly coupled. It is clear that the conductivity depends on the stress-dependent fracture aperture through a strong coupling to fluid flow, as opposed to fixed aperture fractures in conventional percolation models. Local areas can exhibit higher effective permeabilities or strong growth barriers. Such enhanced or restricted opening occur at intersections and offsets, and their existence can affect the total system conductivity, producing a higher pressure level as shown in Figure 5. Early time rapid hydraulic fracture propagation and intersection of small natural fractures establishes a path for the fracture through the natural fracture network, and a single fracture connection event can cause a strong

The model has particular application for understanding the hydraulic fracture connection process through a network. By varying parameters, one finds the transition of fracturecontrolled flow pattern from more uniform to more localized and from multi-directional to unidirectional. The hydraulic fractures tend to develop wider and more connective localized fracture channels in establishing a preferred path through the network of natural fractures. In contrast, low rate injection processes that do not involve significant fluid viscous dissipation effects, tend to result in flow occurring along all already connected conductive paths. How to better characterise this difference is still open and to find meaningful parameters in connecting the intricate topological fracture network with diffusion flow patterns requires prediction of propped and unpropped fracture permeability that remains after the hydraulic fracture

treatment. The model used here may provide a tool for such parametric studies.

The authors thank CSIRO for supporting this work and granting permission to publish.

opening of cross-cutting natural fractures and branching.

response for hydraulic fracture network growth.

**5.2. Implications for fracture-controlled flow system**

change in the hydraulic fracture channel system that develops.

**Acknowledgements**

Without considering fluid loss into the rock matrix, hydraulic fracture growth as the main driving force in connecting fractures can create more new segments at the upstream end of the fracture system than at the downstream. In terms of fracture length, both pre-existing and newly created, we can define fracture density as the fracture number per unit area. Although fracture density is a significant measure for fracture connectivity, the longer fractures includ‐ ing the newly created parts would be more important contributors because they are more compliant and will open wider under the same internal fluid pressure [9]. Predicting the early growth of these hydraulic fractures through a pre-exising network, as modeled in this paper, must account for the effect of viscous fluid or incorrect fracture behaviour is predicted.

In this model, fracture growth occurs when the failure criterion is satisfied at any fracture tip, with the failure condition defined within the framework of linear elastic fracture mechanics. Fracture curving is the natural result of the local stress field around the tip if the growth follows the maximum tensile stress criterion. Normally, fractures will reorient themselves to the maximum compressive stress direction to increase fracture opening, resulting in local con‐ ductivity enhancement as indicated in above results. However, the fracture curving can sometimes lead to intersection of two fractures at a small acute angle, which will make it difficult for the subsequent flow to enter some segments. Sometimes, the subsequently developed sliding on one fracture can seal the fracture channels near the junctions. The development of geometric networks, produced by growing hydraulic fractures, are illustrated by the results obtained above. The results imply that not all connected fractures can contribute to overall conductivity of the system which is contrary to conventional percolation model predictions. These geometric factors affecting fracture growth and fluid flow have been mentioned in early studies [7]. Some fracture growth can occur in the wake of the fracture and flow fronts near the higher pressure entry zone. Local reversed flow has also been observed in the results due to the pressure changes.

Actually, in addition to injection conditions, many other factors such as injection rate and in situ stress can affect the crack growth and coalescence. At the elevated pressure and based on the assumed fracture geometries, one can find that, even through a network of natural fractures, the hydraulic fracture average direction tends to align as much as possible with the direction of the maximum stress. This orientation reduces the viscous dissipation and injection pressure. However, an increase in fluid pressure because of increased rate or viscosity can produce higher pressures upstream of a local offset or restriction which can then lead to opening of cross-cutting natural fractures and branching.

In addition, this paper only considers a limited number of specific initial fracture geometries, the results may be different if the starting geometries are changed and more cases are being considered as a way of making our conclusions stronger and more general. Although a method to deal with network development is presented here, there is a need to work on different geometries to extract some useful general responses for rational simplifications of the expected response for hydraulic fracture network growth.

#### **5.2. Implications for fracture-controlled flow system**

**5. Discussion**

152 Effective and Sustainable Hydraulic Fracturing

from these pre-existing fractures.

in the results due to the pressure changes.

**5.1. Fluid-driven fracture nucleation, growth and connection**

In the model, we only consider hydraulic fracture growth through a finite set of fractures, some of which are initially very small, but potentially provide a conduit with the help of high fluid pressure. The fracture seeds are pre-assumed in this paper and are represented by these small fractures. Therefore, fracture nucleation in a highly stressed area is not dealt with in this paper. This treatment of fracture nucleation can underestimate the fracture number and the fracture connectivity. For the case shown in Figure 4, the lower fluid flow path cannot move to the right outlet and a higher stress level might create new fractures near the entry zone. It would be interesting to study the impact of crack nucleation based on stress conditions rather than just

Without considering fluid loss into the rock matrix, hydraulic fracture growth as the main driving force in connecting fractures can create more new segments at the upstream end of the fracture system than at the downstream. In terms of fracture length, both pre-existing and newly created, we can define fracture density as the fracture number per unit area. Although fracture density is a significant measure for fracture connectivity, the longer fractures includ‐ ing the newly created parts would be more important contributors because they are more compliant and will open wider under the same internal fluid pressure [9]. Predicting the early growth of these hydraulic fractures through a pre-exising network, as modeled in this paper, must account for the effect of viscous fluid or incorrect fracture behaviour is predicted.

In this model, fracture growth occurs when the failure criterion is satisfied at any fracture tip, with the failure condition defined within the framework of linear elastic fracture mechanics. Fracture curving is the natural result of the local stress field around the tip if the growth follows the maximum tensile stress criterion. Normally, fractures will reorient themselves to the maximum compressive stress direction to increase fracture opening, resulting in local con‐ ductivity enhancement as indicated in above results. However, the fracture curving can sometimes lead to intersection of two fractures at a small acute angle, which will make it difficult for the subsequent flow to enter some segments. Sometimes, the subsequently developed sliding on one fracture can seal the fracture channels near the junctions. The development of geometric networks, produced by growing hydraulic fractures, are illustrated by the results obtained above. The results imply that not all connected fractures can contribute to overall conductivity of the system which is contrary to conventional percolation model predictions. These geometric factors affecting fracture growth and fluid flow have been mentioned in early studies [7]. Some fracture growth can occur in the wake of the fracture and flow fronts near the higher pressure entry zone. Local reversed flow has also been observed

Actually, in addition to injection conditions, many other factors such as injection rate and in situ stress can affect the crack growth and coalescence. At the elevated pressure and based on the assumed fracture geometries, one can find that, even through a network of natural fractures, the hydraulic fracture average direction tends to align as much as possible with the

Our numerical results quantify the overall path of discrete hydraulic fractures growing through a network of pre-existing natural fractures, as shown in Figs. 3 and 4. These results give insight to the behaviour of fracture-controlled flow systems, where the fluid flow and the rock deformation and fracturing are strongly coupled. It is clear that the conductivity depends on the stress-dependent fracture aperture through a strong coupling to fluid flow, as opposed to fixed aperture fractures in conventional percolation models. Local areas can exhibit higher effective permeabilities or strong growth barriers. Such enhanced or restricted opening occur at intersections and offsets, and their existence can affect the total system conductivity, producing a higher pressure level as shown in Figure 5. Early time rapid hydraulic fracture propagation and intersection of small natural fractures establishes a path for the fracture through the natural fracture network, and a single fracture connection event can cause a strong change in the hydraulic fracture channel system that develops.

The model has particular application for understanding the hydraulic fracture connection process through a network. By varying parameters, one finds the transition of fracturecontrolled flow pattern from more uniform to more localized and from multi-directional to unidirectional. The hydraulic fractures tend to develop wider and more connective localized fracture channels in establishing a preferred path through the network of natural fractures. In contrast, low rate injection processes that do not involve significant fluid viscous dissipation effects, tend to result in flow occurring along all already connected conductive paths. How to better characterise this difference is still open and to find meaningful parameters in connecting the intricate topological fracture network with diffusion flow patterns requires prediction of propped and unpropped fracture permeability that remains after the hydraulic fracture treatment. The model used here may provide a tool for such parametric studies.
