Spacecraft assembly task planning system and sequence allocation planning method are proposed to address the problem of spacecraft idling and low assembly efficiency due to the assembly constraints during the multibranch spacecraft cluster execution of on-orbit assembly task. First, multi-spacecraft assembly task planning model is established based on the assembly constraint relationship graph considering non-order-conserving assembly assumption scenario. Second, the assembly order forward index and homologation inertia index are proposed to constrain the problems of fragmentation and explosion of the number of subassemblies under the non-order-conserving assumption, as well as the problem of switching between different spacecrafts for successive assembly tasks. Finally, the topological sorting algorithm is improved based on depth-first search to solve the sequence allocation planning problem under the non-order-conserving constraints; the genetic algorithm is adopted for assembly sequence optimization, which effectively reduces the number of fragmented sub-assemblies and the number of consecutive task switches while improving the assembly efficiency; and the Gumbel-Sinkhorn network is used as the sequence allocation solution set for the assembly tasks, which realizes the mapping of a single assembly sequence to a feasible assembly solution set. The simulation calculation of on-orbit assembly task sequence allocation planning verifies the effectiveness of the model and method.