基于改进BP神经网络的水库调度函数研究  

Study on reservoir operation function based on improved BP neural network

作　　者：吴佰杰[1] 李承军[1] 查大伟[1] 

机构地区：[1]华中科技大学水电与数字化工程学院,湖北武汉430074

出　　处：《人民长江》2010年第10期59-62,74,共5页Yangtze River

摘　　要：介绍了BP神经网络的基本原理及其改进方法。运用改进BP神经网络对水库调度函数进行非线性拟合,并将所得的调度函数应用于水库群联合调度。模拟调度中以出流和时段末水位做决策,两者相互印证和补充。实例表明,神经网络拟合精度较高,水库群的模拟联合调度取得了满意的效果。对于金沙江下游梯级及三峡葛洲坝水库组成的6库水电系统,对比按常规调度和线性调度函数方式,梯级保证出力分别增加了380 MW和160 MW,年均电量分别提高了73.79亿kW.h和12.45亿kW.h,年均弃水量分别减少了512.5亿m3和87.7亿m3,效果显著。The basic theories of BP neural network and its improved methods are introduced.The nonlinear fitting for the reservoir operation function is performed by using improved BP neural network,and the obtained result is applied to the joint operation of multi-reservoirs system.The outflow and end water level in the simulated operation are determined as the decision-making variables which can be mutually complementary.The result shows a high fitting precision of the neural network.The method is applied in the joint operation of 6-reservoirs system,consisting of the cascade stations on lower reaches of Jinsha River,Gezhouba and TGP stations,in comparison with adopting the methods of conventional operation and linear dispatching function,firm power and annual power generation are increased and annual surplus water is decreased.

关 键 词：BP神经网络 非线性 水库调度 决策 调度函数 

分 类 号：TV697.11[水利工程—水利水电工程]