A basic analysis for flood control of small rivers and securing water resources under post disaster condition by using optimized control of reservoirs

在本研究中，我们开发了方法、算法和计算机系统，通过优化较小水库的泄洪来减少城市地区的洪水。具体地说，我们开发了基于分布式水文模型的水文模拟器，在该模型中，来自雷达的有一定误差的降雨信息被用于定量地给出概率洪水预报。在此基础上，应用随机洪水预报的优化方法进行水库防洪。首先，利用雷达回波奇异向量外推方法进行概率集成，得到了一种短期降水预报的概率方法。然后，建立了基于广义似然不确定性估计(GLUE)的概率分布水文模型。其次，将动态规划优化控制算法应用于水库防洪。在本研究中，我们使用随机动态规划(SDP)来处理日本气象厅(JMA)发布的每周集合天气预报。在SDP中，目标函数是使下游流域某点的水位最小。SDP与周集合天气预报的结果使水位下降，达到了本研究的目的。

In the present study, we developed methods, algorithms and computer systems which reduce floods in urban area by optimizing the releasing discharge from smaller reservoir. Specifically, we developed hydrological simulators based on distributed hydrological model in which rainfall information from radar with some extent of error which was quantitatively utilized for giving probabilistic flood forecasting. Then, an optimized method was applied for flood control using reservoir with stochastic flood forecast. First, a probabilistic method for short term rainfall forecast was achieved by using radar echo extrapolation method with singular vector for making probabilistic ensembles. Then, a hydrological model which also deals with probabilistic spread based on generalized likelihood uncertainty estimation (GLUE). Second, an optimized control algorithm using Dynamic Programming has been applied for flood control by using reservoir. In the present study, we use Stochastic Dynamic Programming (SDP) for dealing with weekly ensemble weather forecasting issued from Japan Meteorological Agency (JMA). In the SDP, objective function was minimizing the water level at a point in lower river basin. The result of the SDP with weekly ensemble weather forecasting made water level reduced and the objectives of the present study has been achieved.