In the water sector, issues concerning the aquatic environment have been extensively discussed due to climate change. Particularly, water quality problems such as harmful cyanobacterial blooms (CyanoHABs) can be threats to the water environment while harming human health and aquatic ecosystems. This study focused on establishing a practical framework for the optimal operation of upstream reservoirs to address the problem of CyanoHABs in a downstream river. Furthermore, the applicability of this framework was demonstrated using observational data related to the quantity and quality of the upstream reservoirs in the study area, the Nakdong River of South Korea. Methodologically, three models were incorporated: a machine learning model to predict the occurrence of CyanoHABs, a river water quality model to simulate a water quality parameter influencing CyanoHABs, and an optimization model for the joint operation of upstream reservoirs. The research findings can support the decision-making of reservoir operation to create a favorable aquatic environment in a downstream river by reducing the frequency of CyanoHABs downstream. In particular, the framework to be established in this research can be a novelty in terms of efficiency since it can be a solution to the problem of CyanoHABs without using an additional amount of water from an upstream reservoir.