Page - 265 - in Book of Full Papers - Symposium Hydro Engineering

acceptable risk level frequently serves for decision-making after balancing the cost against benefits from that risk. ALAPR (As low as reasonably practicable) is a widely used acceptable risk criterion which involves unacceptably high risk, generally acceptable risk and negligible risk [1]. The acceptable risk level lies between unacceptably high risk region and generally acceptable risk region, which determines the feasibility. HSE (The Health and Safety Executive) proposed that risk of existing infrastructure is annual fatality risk of 10-4 [2]. Concerned with the acceptable risk of dam safety, USBR (United States Bureau of Reclamation) and ANCOLD (Australian National Committee on Large Dams) suggested 10-2 and 10-3 fatalities/person/year for each dam failure event respectively, while BC Hydro presented the acceptable life and economic risk associated with each dam failure is of 10-3 fatalities/person/year and US$7120/year respectively [3]. Risk depends on probability of failure and loss of hazard. The loss varies with time and place, is closely related to the development of local economy and society, while failure probability can be approached by statistics and simulations. In this article, we focus on dam failure probability. According to analysis on different dam failure modes, flood related problems including flood exceeding control criterion and spillway failure both leading to erosion damage on dams and bank joint, which accounts for one third of overall dam failure events. The dam in United States which were built after 1950s with a height more than 30m is of a failure probability in the range of 1.3×10-3～3.0×10-3. It indicates [4] the dam failure caused by flood related problems is in the range of 0.43×10-3～1.0×10-3. Yang et al[4] considered 10-3～10-4 on the lower side of criterion for dam safety and flood control. Regardless of flood regulation, Xu [5] provided 10-5 as risk probability of spillway for high dam. As the technology of design, construction and management for dams and reservoirs improved, dam failure events reduced. In China, from 1990 to 2011, there were 278 dam failed in the overall approximately 98000 dams [6]. It offered a dam failure probability of 2.84×10-4 and deduced that probability of flood related risk is 9.47×10-5. The probability of flood related problems could be attained by numerical simulation as well as statistics. Xu [7] attributed flood-related risk to large flood and failure of spillway, and applied JC method to risk calculation. The P-III and normal distribution were used to describe the uncertainty of flood inflow and initial stage respectively, while triangle and normal distribution were applied to describe the uncertainty of spillways capacity. Based on stochastic simulation approach, Mei et al [8] took both hydrological and hydraulic uncertainty into account, used seasonal AR(1) models to generate a great number of reservoir inflow. The results revealed that flood-related hydrological uncertainty plays the major role in risk analysis compared with any other factors. Based on stochastic simulation, we researched in the view of: (1) dam safety itself, (2) flood control capability of dam and, (3) flood control for downstream regions provided by dam. The case study is 4 built cascade reservoirs in the Jinsha River, China. Based on statistical parameters obtained from observed 49-year flood peak series of 3 hydrological stations in the study area, climate change is studied and stochastic simulation is adopted to reflect the uncertainty of flood inflow. The calculation and procedure will contribute to 265