Abstract: Abstract: The traditional load distribution model of hydropower units has been confined to the "dimension explosion" in the calculation, due to the redundant constraints during operation, leading to the slow solution speed and low efficiency of the model. There are many times of units crossing the vibration zone, and a large fluctuation of unit output in the load distribution. In this study, a multi-objective multi-constraint double-layer intelligent optimization model was established for the hydropower units using the coupled dynamic programming (DP) and improved biogeography-based optimization (IBBO). The minimum water consumption in the process of load distribution was taken as the optimization objective, and the 15-minute duration was taken as the minimum load distribution period to calculate the short-term load distribution of hydropower units. A multi-constraint optimization model of the hydropower unit was constructed using the DP outer layer. The average fitness of the population was introduced into the inner layer using the IBBO. A dynamic migration model was then established from the relationship between the average fitness of the population and the individual fitness. The mixed crossover was introduced into the adaptive updating strategy of the migration operator. The unit load distribution optimization model was constructed using the unit output fluctuation constraints, according to the unit combination. Taking the data of demand load and generating head of a hydropower station on a certain day as an example, firstly, the iterative convergence times of the improved IBBO increased from 45-50 to 20 generations, and the traffic consumption was reduced by 1.28% and 1.82%, respectively, compared with the traditional IBBO and BBO, when the minimum traffic consumption was taken as the fitness function of the model. The convergence and optimization of the improved IBBO were outstandingly improved by the migration model and operator of the algorithm. Secondly, the IBBO-DP model saved 10.56% of water consumption, compared with the double-layer dynamic programming (DDP) model. The zero crossing of the vibration zone of the unit was achieved in 50 calculations on average, leading to the reduced 24 crossing of the vibration zone, compared with the DDP model. Finally, the IBBO-DP model effectively stabilized 60.24% of unit average output fluctuation amplitude, and 47.28% of unit output fluctuation ratio, compared with the DDP model, considering the constraint of unit output fluctuation. There was a great increase in the unit vibration avoidance, operation stability, and reliability in the process of load distribution of hydropower units. Consequently, the double-layer intelligent optimization model was constructed to consider the fluctuation constraint of unit output, and then a case study was used to verify the load distribution optimization of hydropower units, indicating a wide range of engineering applications for the actual operation of hydropower units. This finding can also provide a strong reference for the subsequent operation of hydro-photovoltaic complementary power generation.