Reservoir Flood-Season Staging Method Based on Improved Fuzzy Sets and K-Means Clustering

To address the issue of strong subjectivity in traditional flood season staging methods, this paper proposes a comprehensive staging method that integrates the coefficient of variation method and K-means clustering. Firstly, the maximum flood volumes of 1 day, 3 days and 5 days are selected as key indicators. The coefficient of variation method is adopted for objective weight assignment to improve the fuzzy set analysis and preliminarily define the flood season range. Furthermore, K-means clustering is applied to achieve the accurate division of flood season stages. When this method is applied to the flood season staging of a certain reservoir, the flood season can be clearly divided into the pre-flood season (7/07-8/06), the main flood season (8/07-9/04), and the post-flood season (9/05-10/01). Through comparative analysis with other classic methods, the rationality and reliability of its division results were verified. This method can achieve more accurate and objective identification of flood season phases, providing methodological support for guiding the flood control and dispatching operation of reservoirs during the flood season.