Aiming at the problems that traditional genetic algorithm is prone to fall into local optimal solution, parameters cannot be adjusted intelligently, and local search ability is poor when solving flexible job-shop scheduling problems, a flexible job-shop scheduling model with the goal of minimizing the maximum completion time was established. A reinforcement learning improved genetic algorithm (RLIGA) based on reinforcement learning was proposed to solve the model. Firstly, in the iterative process of genetic algorithm, reinforcement learning was used to dynamically adjust key parameters. Secondly, the discrete Lévy flight mechanism based on process coding distance was introduced to improve the solution space. Finally, the variable neighborhood search mechanism was introduced to improve the local development ability of the algorithm. PyCharm was used to run Brandimarte examples to verify the solving performance of the proposed algorithm. The experiment proves that the proposed algorithm has higher solving efficiency, stronger ability to jump out of the local optimal, and better solving results.