Given the characteristics of decommissioning product diversity, wear difference, and structural complexity, we analyzed the disassembly line balance problems of minimum disassembly workstation, station load balancing, priority disassembly of high-demand parts and the lowest cost of invalid operations, and further considered the influence of random factors on actual disassembly operations. The purpose of this paper is to establish a model of sequence-dependent stochastic mixed-flowed partial disassembly line balancing problem and propose an adaptive hybrid particle swarm genetic algorithm to solve the model. The algorithm replaces the fixed value genetic operator with an adaptive crossover and mutation operator to improve the global optimization ability. The introduction of adaptive weighted particles leads to improved local optimization ability of the algorithm so that the algorithm has strong global and regional optimization ability to improve algorithm accuracy. The effectiveness of the proposed algorithm is verified by solving the classic disassembly line balancing problem with different scales and comparing it with the solution results of underlying genetic and particle swarm optimization algorithm. Meanwhile, the proposed model and algorithm are applied to the mixed-flow disassembly engineering project of a 25-task reducer. The results indicate that the proposed model is superior to the control group in terms of minimal disassembly workstation, station load balancing, invalid operation cost, and overall performance of the disassembly line, which verifies the validity of the model.