In this study, we propose a multi-objective heuristic wolf pack algorithm to tackle the issues presented by several workpiece type switches and machine start-stop evaluation functions as the optimization objectives for the unrelated parallel machine batch scheduling. During the generation of the initial population, the algorithm incorporate list-based backward learning and heuristic strategies using machine processing efficiency and design an irregular real matrix encoding method for achieving task binning. The proposed model used a combination of local and global neighborhood search to implement intelligent behavioral search in the wolf pack algorithm. Furthermore, a neighborhood search is performed using batched adjustment learning mechanism for the current results, and the improve integer solution Pareto non-dominated sorting method is used for circular iteration. Finally, the effectiveness and superiority of the algorithm are verified by testing practical arithmetic cases of different scales and comparing related algorithms.