<p>With the advancement of manufacturing, distributed manufacturing models have been applied to hybrid flow shop scheduling problems (HFSP). The variability among different factories renders traditional scheduling methods based on single-factory models no longer applicable. Furthermore, real-world production processes are inevitably influenced by changes in factors such as equipment conditions, worker skills, and production environments, leading to uncertainties in the processing times of machine operations. To address this problem, a dual evolutionary algorithm based on Dyna-Q is proposed. Trapezoidal fuzzy numbers are used to represent uncertain processing times, aiming to solve the heterogeneous hybrid flow shop scheduling problem (HHFSP) with the objectives of minimizing the makespan and total energy consumption (TEC). First, three efficient initialization methods are proposed to generate high-quality initial populations. Second, two evolutionary algorithms with different search directions are designed to enhance population diversity and convergence, respectively. Furthermore, a Dyna-Q-based selector is introduced to adaptively choose the optimal solver under the current state, achieving a balance between exploration and exploitation. Finally, extensive experiments are conducted to validate the effectiveness and efficiency of the proposed algorithm.</p>