Disassembly line balancing problems directly affect the reverse/remanufacturing cost. A multi-objective sequence-dependent disassembly line balancing problem(SDDLBP) optimization model is established for minimizing the number of opened workstations, minimizing the total disassembly time, distributing the idle times across the opened workstations evenly, and processing the hazardous components and high-demand parts early. Then a hybrid artificial bee colony(HABC) algorithm is proposed to solve this problem. In the onlooker bee phase, a multi-stage evaluation method is presented to distinguish food sources better. In the scout bee phase, a search strategy based on the global learning mechanism is constructed to improve the exploitation ability of the scouts. When the bees search for new food sources, the reduced variable neighborhood search(RVNS) strategy is used to promote the search efficiency. Finally, computational results demonstrate the effectiveness of the proposed model and algorithm.