DOI: 10.1615/ICHMT.1992.IntForumExpSysCompSimEE
ISBN Print: 978-1-56700-486-1
PREDICTION OF NUCLEAR REACTOR PARAMETERS USING ARTIFICIAL NEURAL NETWORK MODELS
摘要
Artificial neural networks (ANNs) are applied to the prediction of nuclear reactor parameters in load following (LF) operation. The system consists of four parameter processing banks that should be monitored in the load following operation. In each bank, two types of networks using either a general multi-layer network (GMLN) or a hybrid functional-link network (HFLN) are constructed and attempted to learn or to infer signal behaviors. The overall prediction results are agreed well with actual plant data except for the minor discrepancies at the nearly upper or lower bound of trained band. It also indicates that the HFLN approach performs better than the GMLN approach in the way that it uses less number of processing elements and weights, trains faster, and gives better results except for the minor difficulty of input data expansion.