<p>Reservoirs are vital to the rational allocation, exploitation, and utilization of water resources. Studies on staged drought operation strategies for reservoirs attempt to increase the storage capacity of water resources and improve the benefits of reservoirs in terms of prosperity. In this study, the minimum-risk Bayesian decision theory is introduced to determine the multiyear water-supply combination of a reservoir drought-resistance operation strategy. The staged drought-resistance operation optimization model for reservoirs is modeled based on the Pareto-improved artificial bee colony algorithm. Furthermore, the Whale Optimization Algorithm (WOA)-BP neural network is used to forecast the wetness–dryness year of the Jinpen Reservoir. The results reveal that the prediction model, which provides forecasting information for the dynamic control of the drought-resistance operation strategy, is highly reasonable for watershed prediction. Monte Carlo simulation indicates that the most unfavorable intra-annual distribution of a normal year is preferred. Additionally, the drought-resistance operation strategy is dynamically controlled in a normal year based on the drought-resistance optimal operation model. The simulated operation of the historical runoff process from 1955 to 2022 reveals that the dynamic-staged drought-resistance operation strategies improve the ability of the reservoir to resist extreme drought and reduce the surplus water volume in the wet season and wet year of the reservoir. Furthermore, it increases the utilization rate of water resources and reduces drought loss, thus providing a theoretical basis for the drought-resistance operation of the reservoir and the optimal allocation of water resources.</p>