确定水库分期汛限水位的神经网络调洪演算方法

水库汛限水位动态控制是目前协调防洪与兴利矛盾实现洪水资源安全利用的关键问题. 现行汛限水位是通过对不同频率设计入库洪水进行调洪演算确定的. 由于汛期不同分期对应不同的设计入库洪水，于是在保证水库大坝工程安全的前提下，从水库实际调度方案和下游防洪标准角度出发，应用BP神经网络对各分期不同频率设计入库洪水进行逆时序调洪演算，充分挖掘水库汛期兴利效益，最大限度抬高汛限水位，由此提出确定水库分期汛限水位的神经网络调洪演算方法. 应用结果表明：该方法直观、计算量小，精度可满足实际工程需要，在水库汛限水位计算中具有推广应用价值.

Application of Neural Networks to Determine Limited Water Level of Reservoir Flood Routing

The dynamic control of reservoir limited water level is the key point to resolve conflict between flood control and benefits promotion, and realize safe utilization of flood resources. At present, the limited water level is determined by using flood routing to design flood of different frequencies. Different periods in a flood season have different designed flood. Based on safety of the dam and in view of practical operations and flood control standards,BP neural networks are used in the reservoir flood routing process to fully realize the benefits promotion and raise the limited water level as far as possible in a flood season. A neural network method for determining various limited water levels of reservoir flood routing is proposed. Applications show that the method is easy for visualization and simple for implementation. The precision satisfies practical needs, and it can be used in the computation of limited water level.