Paddy drying is an energy-intensive process that involves complex interaction such as inertia, nonlinearity, and random disturbances. Real-time energy consumption regulation is challenging due to the interplay of these factors. This study proposes a two-level hierarchical model predictive control (MPC) strategy for industrial-scale circulation counter-flow paddy drying process. The first-level optimizer encompasses an energetic optimizer, engineered to minimize energy consumption. This optimizer integrates drying mathematical and energetic models, as well as drying and ambient data. It operates at a low frequency of once every 180 s to handle computational complexity and slow-changing ambient conditions. To handle high-frequency disturbances, a second-level MPC operates at 2.25 s intervals, relying exclusively on drying mathematical model and tracking ideal trajectory established by first-level optimizer. Experiments show that first-level optimizer reduces total energy consumption by 12.8 % compared to previous proposed static ventilation strategy. Hierarchical MPC strategy consistently achieves lower relative average deviations (0.70 %, 0.79 %, and 0.81 %) from ideal trajectory under varying disturbance fluctuation rates (±60 %, ±80 %, and ±100 % respectively). These deviations are markedly lower (by 1.58 %, 15.67 %, and 19.52 % respectively) than those observed when applying first-level optimizer under noisy conditions. These findings underscore the enhanced energy-saving and disturbance-suppression capabilities of proposed hierarchical MPC strategy.