基于Elman神经网络的NSG水位特性辨识方法研究

针对核动力蒸汽发生器在瞬态、启动和低功率下的“收缩”与“膨胀”现象引起的逆动力学效应使核动力蒸汽发生器水位特性难以辨识的问题,提出了基于Elman神经网络的NSG水位特性辨识的新方法.采用串-并联型辨识结构,以保证辨识的收敛性和稳定性.网络训练采用Levenberg-Marququardt BP学习算法.仿真结果表明,所提出的方法能够正确地辨识核动力蒸汽发生器的水位特性,且具有较高的辨识精度.

On identification method for NSG water level based on Elman neural network

The false water level,caused by the reverse thermal-dynamic effects which are known as "shrink" and "swell" effects and occur during plant transients and are more prominent at start-up and low power operation,makes the water level characteristicof the nuclear steam generator difficult to be identified.In order to solvethis problem,anovel identification method based on the Elman neural networkis put forward.The series-parallel model is applied to the identification to assurethe convergenceand stabilityand the back propagation algorithm of Levenberg-Marququardt type is employed totrain the network.The proposed method thatcan identify the characteristic of NSG water level correctly and has enough precision of identification is demonstrated by the computer simulation.