The distribution of ice and water within frozen soil is closely correlated with electrical parameters and is a key factor for frozen soil detection by ground penetrating radar (GPR). However, the relationship between the evolution of freeze-thaw layering and electromagnetic response fails to be characterized by traditional GPR forward modeling methods from a thermodynamic perspective. To address this issue, a comprehensive GPR forward modeling framework based on frozen soil hydro-thermal coupling theory and effective medium theory was proposed. The electrical structure of frozen soil under different thermal states and its stepped frequency continuous wave (SFCW) GPR response characteristics were revealed. First, a coupled model of temperature and moisture fields was established using COMSOL to obtain the distribution of temperature, saturation, and ice content of the active layer of frozen soil at different freeze-thaw stages. Subsequently, the equivalent dielectric constant and electrical conductivity of the frozen soil model were calculated based on effective medium theory. Finally, the electrical parameters were imported into gprMax software, and forward modeling comparative analysis was conducted using SFCW signals and Ricker wavelets. It is indicated by the results that the dynamic evolution process of freezing and thawing fronts within frozen soil can be reconstructed by the proposed modeling method. It is demonstrated that SFCW signals and Ricker wavelets remain consistent in interface position and trend without obvious waveform distortion. It is confirmed that complex dielectric layered structures in frozen soil caused by drastic changes in temperature and moisture can be effectively identified by SFCW. A forward modeling reference is provided for the detection of active layers in permafrost regions using stepped frequency GPR.