Abstract: Existing underground wireless channel modeling mainly focuses on specific scenarios with relatively regular geometric structures such as roadways, shafts, and room-and-pillar structures, which is insufficient for goaf scenarios with complex and variable geometric shapes and non-uniform media distribution. For the porous heterogeneous media environment of coal mine goaf, the bouncing ray method is adopted to establish the wireless channel model. First, a coal mine goaf environment model was established, and the bouncing ray method was used to simulate the propagation mechanism of electromagnetic waves in the porous heterogeneous media of the goaf. Direct and reflected paths between transmitter-receiver, transmitter-repeater, and repeater-receiver were obtained. Then, for each propagation path, reflection coefficients were introduced to establish the channel impulse response model and the path loss model, and a thermal noise model was constructed by comprehensively considering the goaf environment temperature and the equipment noise coefficient. Finally, through coherent vector superposition of multipath components, the composite received signal was generated. Based on the established porous heterogeneous media wireless channel model of the goaf, the signal-to-noise ratios at frequencies of 0.7, 0.8, 0.9, 2.6, 3.5, and 5.9 GHz were simulated. The results showed that SNR was significantly negatively correlated with frequency, which was consistent with the theoretical principle of underground coal mine wireless communication, indicating that the established wireless channel model accurately reflected the propagation characteristics of electromagnetic waves at different frequencies in the goaf environment. In addition, under the simulated goaf environment, ZigBee and LoRa communication modules were used to conduct received signal strength tests at frequencies of 2.4 GHz, 868 MHz, and 915 MHz with different transmission distances. The test results showed good consistency with the simulation results, verifying the accuracy of the established wireless channel model.