<p>This paper presents a transmission equipment condition prediction and maintenance decision - making support system based on digital twin technology and convolutional neural networks. The model incorporates a dual - branch CNN structure and an attention mechanism. It combines a time - series branch and an image branch to extract the time - domain features of vibration signals and the time - frequency features of two - dimensional images respectively. The optical time - domain reflectometry system is utilized to collect the vibration signals of the transmission line under different operating conditions for predicting the state of the transmission equipment. This model has been preliminarily applied to a certain transmission line in Guangzhou. The experimental results show that the dual - branch model proposed in this paper is significantly superior to other comparative models in all evaluation indicators, achieving an accuracy rate of 94.09%. The digital twin platform enables the fusion of multiple parameters, which significantly improves the robustness of condition prediction.</p>