A stochastic model for the fracture network in the Habanero enhanced geothermal system

Abstract

Fracture Network Modelling (FNM) plays an important role in many areas where the characterization of discontinuities in deep ground is required. Applications of the FNM include, but not limited, hydrocarbon reservoir production, mineral extraction, tunnelling, underground storage or disposal of hazardous wastes and geothermal systems. One important step in FNM is to estimate the density of fractures and geometries and properties of individual fractures such as the size and orientation. Due to the lack of data, the tortuous nature of fractures and the great uncertainty involved in practice, the only feasible approach is via a stochastic modelling. This paper describes a general optimization approach to modelling the fracture network in a geothermal reservoir, conditioned on the seismic events several kilometres beneath the surface detected during the fracture stimulation process. Two key aspects of our method are the construction of an appropriate objective function and the derivation of an efficient updating scheme, which still remain to be the two challenging issues of most global optimization techniques. In our application, the objective function consists of two important components: the minimisation of squared distances of the seismic points to the fracture model and the minimisation of number of fractures or the amount of fracturing, which corresponds to the least consumption of fracturing energy. The model updating process includes several proposals for perturbing the parameters of individual fractures and also to alter the size of the fracture network in order to get a global optimal solution. As a case study, the model is applied to Geodynamics’ Habanero reservoir in the Cooper Basin of South Australia.