The radio interference generation mechanism of AC/DC parallel transmission lines is different from that of conventional lines, so that the prediction models are either too rough and too large in error, or too complicated to be mastered by engineers. In order to take into account the accuracy of radio interference solution and engineering practicability, an error back-propagation feedforward neural network model combined with gray correlation theory is proposed. First, grey theory screening are performed of 17 impact indicators that affect radio interference, and then the obtained indicators that have a real impact on radio interference are used as input variables, and radio interference is the output variable, and the radio interference prediction is performed through the BP neural network fitting algorithm. It is verified by combining the actual measured radio interference samples of the AC/DC parallel transmission lines that have been put into operation, and the results show that the prediction accuracy of the BP neural network prediction model is improved by 13.58% compared with traditional algorithms.