Abstract: With the development of electronic devices towards integration and miniaturization, the heat flux density of the heat dissipation system of the electronic device keeps increasing, so the micro-channel radiator is more and more widely applied. Aiming at the problems that the simulation speed of micro-channel radiator is slow and difficult to converge, an establishment method of a fast thermal model based on neural network is proposed in this paper. First, a thermal network model that can accurately characterize the heat transfer process is established through the theoretical derivation and analysis of the heat transfer process of the heat dissipation system; then, a fast thermal model is established based on the neural network algorithm and the model is customized and packaged; finally, the fast thermal model is tested and adjusted through a thermal model test system and the verification prototype. Through comparative verification, the simulation results of the fast thermal model highly agree with the experimental test results, with the maximum error of 7.82%，which is within the acceptable error range of engineering design. The simulation based on the fast thermal model only spends 30 seconds, which demonstrates that this fast thermal model can significantly improve the simulation speed.