Abstract
Ecofriendly reservoir operation is an important tool for sustainable water resource management in regulated rivers. Optimization of reservoir operation is potentially affected by the stochastic characteristics of inflows. However, inflow stochastics are not widely incorporated in ecofriendly reservoir operation optimization. The reasons might be that computational cost and unsatisfactory performance are two key issues for reservoir operation under uncertainty inflows, since traditional simulation methods are usually needed to evaluate over many realizations and the results vary between different realizations. To solve this problem, a noisy genetic algorithm (NGA) is adopted in this study. The NGA uses an improved type of fitness function called sampling fitness function to reduce the noise of fitness assessment. Meanwhile, the Monte Carlo method, which is a commonly used approach to handle the stochastic problem, is also adopted here to compare the effectiveness of the NGA. Degree of hydrologic alteration and water supply reliability, are used to indicate satisfaction of environmental flow requirements and human needs. Using the Tanghe Reservoir in China as an example, the results of this study showed that the NGA can be a useful tool for ecofriendly reservoir operation under stochastic inflow conditions. Compared with the Monte Carlo method, the NGA reduces ~90% of the computational time and obtains higher water supply reliability in the optimization.