Abstract: To overcome the limitation of traditional flexible job shop scheduling in which processing route constraints are ignored, a flexible job shop scheduling problem with alternative process plans (FJSP-APP) is formulated, simultaneously considering process route flexibility and machine flexibility, with the objective of minimizing the makespan. A self-learning grey wolf optimization (SLGWO) algorithm is developed to solve the proposed model. A three-stage encoding strategy is adopted to uniformly represent process route selection, operation sequencing, and machine assignment. During the initialization stage, hybrid heuristic rules combined with a reverse mapping mechanism are introduced to improve the quality and diversity of the initial population. In the evolutionary search process, a deep Q-network (DQN) is employed to adaptively adjust key parameters of the grey wolf optimization algorithm, thereby achieving a dynamic balance between global exploration and local exploitation. Experimental results obtained from benchmark instances of different scales and an engineering case show that the proposed SLGWO outperforms comparative algorithms in terms of solution quality, stability, and convergence performance, demonstrating its effectiveness and engineering applicability for complex flexible job shop scheduling problems.