The distributed heterogeneous hybrid flow shop scheduling problem (DHFSP-HF) under Industrial Internet of Things (IIoT)-enabled smart grids poses NP-hard challenges due to its multi-objective coupling of makespan and energy costs. This study proposes a two-stage Q-learning hyper-heuristic evolutionary algorithm (TQHHEA) that combines reinforcement learning with adaptive heuristic selection for IIoT-driven manufacturing systems. TQHHEA integrates Q-learning with hyper-heuristic evolutionary strategies to explore and exploit the solution space. Comparative experiments across diverse factory configurations validate TQHHEA’s advantages over existing methods in maintaining solution diversity and Pareto front optimality. The algorithm demonstrates strong adaptability in complex scenarios, achieving efficient coordination between production scheduling and energy consumption management. Its edge-cloud architecture supports distributed industrial computing requirements, offering practical value for sustainable manufacturing optimization.