In order to improve the accuracy of fault phase selection of transmission lines, a new method of fault type identification of transmission lines based on the characteristics of current fault components and random forest is proposed. This paper calculates the energy relative entropy of the three-phase current fault component and the sum of energy of the zero-sequence current fault component under multiple short-time sequences, and combines the results calculated under each short-time sequence into a characteristic sample vector in a specific order to characterize the fault type characteristics of transmission lines. It establishes a random forest intelligent fault type identification model, and uses the fault characteristic sample set to train and test it to identify specific fault types. Simulation results show that the proposed algorithm can accurately identify specific fault types under different initial fault angles, different transitional resistances and different fault distances, and can accurately identify fault types even with data loss and noise interference. This work is supported by the Project of Sichuan Provincial Science and Technology Department (No. 2017JY0338, No. 2019YJ0477, No. 2018GZDZX0043 and No. 2018JY0386) and the Project of Artificial Intelligence Key Laboratory of Sichuan Province (No. 2019RYY01).