Natural gas hydrates are considered as one of the most important potential alternatives for energy shortage, and depressurization is believed to be the most economic effectiveness for hydrate exploitation. The exploitation induces solid hydrate dissociation, liquid / gas generation and migration, heat transfer and skeleton deformation, showing strong thermo-hydro-mechanical (THM) coupling behavior of sediments. These processes will trigger geotechnical hazards, such as borehole inclination and reservoir collapse. A THM coupling numerical model for hydrate exploitation was developed based on the open-source FEM platform OpenGeoSys. In detail, the kinetic reaction equation was implemented to represent the liquid/gas generation during gas hydrate dissociation. Air and water mass balance equations were introduced to illustrate the phase change between liquid and gas phases. Besides, the nonlinear complementarity problem (NCP) combined with the choice of special primary variables was adopted to strictly constrain the liquid / gas saturation, and kinetic reaction rate was connected to the source/sink terms of governing equations. The model was validated and verified through a test and a large-scale numerical model, and the effects of pore water compressibility on the THM coupling response of hydrate-bearing sediments were discussed. Results show that the model was numerically stable in dealing with the nonlinear problems of phase change among solid, liquid and gas phases, and of phase appearance / disappearance of pore fluids caused by hydrate dissociation; the hydrates dissociate from near to far regions, and the produced gas / liquid gradually migrated to the driving well and reached a steady state, where the gas / liquid saturation tended to be stable; due to the influence of air dissolution, the evolution of gas phase saturation lagged behind that of the liquid saturation; the pore water compressibility had a non-negligible effect on the dissipation of pore pressure and skeleton deformation under large pore pressure.