基于ANN的抽水蓄能电站建模与过渡过程优化

对大型带长引水管道的抽水蓄能电站机组调节系统进行了建模与过渡过程调节参数优化的研究.依据某抽水蓄能电站的具体参数,利用Matlab/Simulink对其机组调节系统建模,其中长引水管道系统采用弹性水击模型以较准确地反映较长压力管道内的水力振动特性,然后利用BP和RBF神经网络训练水泵水轮机特性参数,以获取较精确的水轮机模型,从而避免用插值法处理密集、交叉现象严重的水泵水轮机“S”特性曲线带来的误差.基于所建立的仿真模型,引入改进正交试验法对其空载频率扰动、负荷扰动及甩负荷过渡过程进行调节参数优化.结果表明,所建立的调节系统仿真模型可以在一定程度上反映机组的实际特性,并且改进正交试验法在机组调节系统过渡过程仿真试验参数优化中取得了很好的效果.本研究对大型抽水蓄能机组调节系统建模仿真具有一定的指导意义,为实际运行中机组过渡过程优化控制提供了理论依据和技术参考.

ANN based modeling and transient process optimization of pumped storage power station

Modeling and parameters optimization in transient processes of governing system of large a pumped storage power station with long pipeline were studied. According to the actual parameters of a certain pumped storage power station, Matlab/Simulink toolbox was utilized for modeling and simulation of these subsystems. More specifically, friction resistance coefficient and water elasticity were taken into consideration in modeling of pressure diversion system with long water pipeline to accurately simulate hydraulic vibration characters in the pipeline. BP neural network and RBF neural network were adopted and compared in modeling of pump turbine to avoid the errors caused by dealing complex "S" characteristic curve of pump turbine. Based on the established pumped storage governing system simulation model, improved orthogonal experiment method was applied in parameter optimization and simulation of no-load frequency disturbance transient, load disturbance transition and load shedding transient of pumped storage governing system. According to the results, the established model can reflect the actual characteristics of pumped storage units to some extent, and improved orthogonal experiment method is efficient and effective in figuring out the best parameter group within the given scope. The results can provide guidance for modeling of governing system of large pumped storage units, and can set references and theoretical basis for actual optimal control of transient processes in pumped sto-rage units.