Although the reservoir simulation is widely utilized to predict geothermal reservoir performances, the results of the simulation are sometimes different from those actually observed in field operations due to non-equilibrium conditions. For example, the recharge water sometimes reaches producing wells much earlier than predicted by reservoir simulation. Therefore, in this research, we attempted to develop a numerical simulator that can deal with the non-equilibrium vaporization of water and condensation of steam for predicting geothermal reservoir performances more accurately. First, we developed a three-dimensional simulator that can predict the flow behavior of geothermal fluids in a non-equilibrium state. Conventional geothermal simulators solve the only material balance equation for all the water molecules regardless of the phase condition. On the other hand, in the simulator developed in this research, water molecules in the liquid phase are distinguished from those in vapor phase, and the two material balance equations are derived for water and steam separately. These equations have the terms to express the molecular transportation from steam to water and vice versa. Non-equilibrium vaporization and condensation of water molecules are expressed by adjusting the kinetic rate of transportation of water molecules across phases. Next, we expanded the functions of the above simulator, incorporating two types of double porosity models, Kazemi and MINC, to reproduce the fluid flow preferentially through fractures and faults. After verifying the simulator functions, we investigated how the speed of the transportation of water molecules across phases affected the geothermal reservoir performances, especially those with recharging water. Case studies revealed that the non-equilibrium condition hastened the movement of the water injected as recharge water through fractures, which resulted in the water breakthrough earlier than predicted by conventional (equilibrium type) simulators.