Abstract: Ground Penetrating Radar (GPR) technology can be integrated into coal mining machinery to detect and identify coal-rock interfaces, providing technical support for intelligent and unmanned coal mining. The GPR response of coal-rock interfaces is easily affected by factors such as interface morphology and geological structures. Existing studies mostly focus on single factors and lack systematic analysis of typical geological influencing factors under a unified modeling framework. To address this issue, the Finite-Difference Time-Domain (FDTD) method was adopted. Based on unified electromagnetic parameters and a consistent simulation framework, forward simulations of GPR detection at coal-rock interfaces were conducted. The GPR response characteristics under typical geological conditions—such as different interface morphologies, gangue distributions, and gas-bearing/water-bearing fractures—were analyzed. The results showed that interface morphology had a significant influence on GPR response characteristics: the reflection signals of horizontal and inclined interfaces were prominent, while those of undulating planar and curved interfaces exhibited phase differences due to disturbances, and the interface position could be effectively identified after processing. Coal seam gangue caused superposition interference in GPR reflection signals, increasing the difficulty of image interpretation. Thus, the coal-rock interface could be determined by combining the amplitude and phase characteristics of reflection signals with their temporal relationships. Gas-bearing fractures exhibited high similarity to the GPR reflection signals of coal-rock interfaces, requiring integrated interpretation using radar profile image features and nearby borehole geological information. Water-bearing fractures caused significant interference to GPR reflection signals and required methods such as waveform deconvolution to suppress interference and improve the distinguishability of coal-rock interface signals. Laboratory GPR experiments were conducted on coal-rock structural models with different morphologies, and the results were consistent with the forward simulations, verifying the accuracy of the forward modeling.