Seite - 266 - in Book of Full Papers - Symposium Hydro Engineering

conduct flood related risk analysis for dam safety. 2. METHOD 2.1. STATISTICAL PARAMETERS OF OBSERVATION According to the drainage characteristic of study area, the representative hydrological stations are selected. Annual maximum discharge in several-year time series are collected and following a P-III frequency distribution curve. The average of annual flood peak (EX), coefficient of variation (Cv) and skewness (Cs) are derived from the fitted P-III frequency curve. 2.2. STOCHASTIC SIMULATION FOR FLOOD PEAKS IN REPRESENTATIVE HYDROLOGICAL STATION Considering changes in landscape and climate which may lead to nonstationary hydrological regimes, GAMLSS (Generalized Additive Models for Location, Scale, and Shape) model[10] is introduced to provide a flexible modeling framework for flood peaks and to reveal the trend of flood peaks observed in hydrological stations. For a stationary hydrological series, according to the EX, Cv and Cs deduced in 2.1, followed the sampling method of P-III distribution [9], a great number of flood peaks for representative hydrological station are generated. 2.3. RESULTS FOR CASCADE RESERVOIRS According to hydrological analogy method and typical hydrograph, the inflow flood hydrographs for each reservoir are acquired. 2.4. FLOOD REGULATION Based on water balance principle and regulation requirements, adopted storage-capacity relation of reservoir and stage-discharge relation of spillway, applied trial and error method, the outflow and maximum stage corresponding to each inflow flood are calculated. Therefore, considering the uncertainty of flood inflow, the probability of different stages for each reservoir can be identified. 266