Fracture networks within hot dry rock (HDR) geothermal reservoirs are complex, and heat extraction via water injection is thus a coupled process of heat-fluid-solid multifield. In this paper, utilizing the theory of normally distributed random functions, we develop a corresponding pre-processing subprogram to establish a discrete network model of complex fracture distribution in HDR reservoirs; then construct a heat-fluid-solid finite element model for heat extraction via water injection and compare the numerical solution with the analytical solution of the one-dimensional non-isothermal consolidation problem for verification. The numerical simulation results show that the main factors affecting the heat extraction efficiency of HDR reservoirs include fracture width, fracture density, fracture permeability, and matrix permeability. When a HDR reservoir is injected with water for heat extraction, there is a certain threshold value of these influential parameters, beyond which the outlet temperature drops significantly, resulting in an obvious thermal breakthrough. When injecting water for heat extraction, the values of these parameters should be controlled and kept at a reasonable level, otherwise, the HDR reservoir may enter a thermal breakthrough stage in advance, which is not conducive for long-period heat extraction. Influenced by the random distribution of complex fractures, the leading edge of the cold front may present an irregular distribution. During the process of heat extraction, the stress gradually changes from a compressional state to a tensile state, which induces further damage to the HDR reservoir.