Abstract: Aimed at the risk problem in reservoir operation caused by uncertainty in water demand process, this study examines the risk of reservoir water supply operation. Based on the initial dispatch mode of the Sanhekou Reservoir, we develop a risk indicator system and a dispatch model for annual average water shortage minimization, and solve it using an improved flower pollination algorithm (IFPA). A four-dimensional coordinate system is constructed and used in analysis of the risks of different reservoir operation schemes. Results indicate IFPA has strong optimization ability and stability in solving the long-term scheduling model. And for water supply risk, the most unfavorable scheme is the water demand process scheme 3 based on the frequency of high-low water encountering, with annual reliability as low as 0.64 and a water shortage index of 2.35, showing a water shortage of several successive years. The values of risk indicators significantly depend on time scale, the smaller the time scale, the greater the risk, the more unfavorable to the scheduling results. The results demonstrate new ideas for accessing reservoir water supply operation risk, and also provide important reference for reservoir management and planning departments.

Key words: reservoir operation, flower pollination algorithm, risk analysis, water demand process, water shortage index, Hanjiang-to-Weihe River water diversion project