Identification of Dynamics for Nuclear Steam Generator Water Level Process Using RBF Neural Networks

Abstract

In the operation of nuclear steam generator (SG), the reverse thermal-dynamic effects make SG water level process dynamics characteristic difficult to identify. In order to improve the effect of identification, a new method based on radial basis function (RBF) neural networks (ANN) is proposed and investigated in this paper. The identification model employs series-parallel model to assure the convergence and stability of identification process. The train algorithm for the RBF neural network (RBFN) adopts the orthogonal least square (OLS) method. The mathematical model of the SG in Qinshan Nuclear Power Plant (NPP) in China is used for simulation demonstration. The identification on SG typical operation modes, which the steam flow rate and feed water flow rate are step change respectively, were implemented to demonstrate the feasibility of modeling SG process dynamics employing RBFN. The identification results show that employing RBFN can identify SG process dynamics correctly and has adequate precision and fast convergence.