Abstract: Flood early warning and flood control operation are two non-engineering measures for flood defense, both playing a key role in reservoir flood control. This paper develops a flood warning system for cascade reservoirs combining these two measures and a flood control operation model considering early warning in the upper Yellow River for flood control and disaster reduction. To calculate the early warning indicators of the system, a fuzzy set theory is used and the early warning indexes and levels are verified using the indicators. This model is solved using the cuckoo search method, and optimal early warning countermeasures and optimal operation schemes are obtained. Results show that flood control operation models considering early warning are more reliable than the flood control scheduling simulation model, providing a new approach to the flood control of cascade reservoirs.