Identification of Measurable Dynamics of a Nuclear Research Reactor Using Differential Neural Networks

Complete modeling of a nuclear reactor is a difficult task because dynamic behavior of this system depends on many factors. So, a complete description of the reactor dynamics implies necessarily the employment of high order nonlinear models. To overcome this problem, in this paper, we propose to use a low order differential neural network for the identification on-line of the uncertain measurable dynamics of a nuclear research reactor. As in real situations many variables associated with the nuclear process are not available for measurement, the identification is performed based on only the input and two states: the fuel temperature and the neutron power. In spite of that, the obtained low order model still shows a good behavior.