To address the high synchronization accuracy and high reliability requirements of the double-mast lifting system of a nuclear island aerial work platform, an attention-based encoder-decoder neural model predictive control (AED-MPC) method is proposed. First, based on the structural composition and working principle of the mast lifting system, a dynamic model of the servo motor-driven rope-pulley system is established, and a simulation analysis model of the double-mast synchronous lifting system is built in AMESim. Meanwhile, using the training data collected from the AMESim simulation platform, an attention-based encoder-decoder network is designed in MATLAB as the internal prediction model of the MPC, enabling rolling optimization and hard constraint handling of the synchronization error. Then, a co-simulation model for mast lifting is constructed using the AMESim S-Function module and MATLAB, and the synchronization control performance of the system under different operating conditions is simulated and analyzed. The results show that under a rotational speed step change, the settling time of AED-MPC is 0.007 s, which is 68.1% less than that of the traditional deep neural network based MPC (DNN-MPC), with no overshoot. Under a sudden 8 N·m load disturbance, the maximum synchronization error is reduced to 2.6 r·min?1 (a reduction of 68.6%), and the settling time is reduced by 47.1%. Under parameter mismatch, the system still maintains high synchronization accuracy and robustness. Finally, experiments verify the effectiveness and strong robustness of AED-MPC in real physical environments.