Using Point Reactor Models and Genetic Algorithms for On-Line Global Xenon Estimation in Nuclear Reactors

Estimation of xenon concentration at a given time instant is usually a difficult problem since the initial conditions are often unknown as well as a number of the model parameters. The feasibility of obtaining the model parameters of a point reactor xenon evolution model with genetic algorithms (GAs) has been investigated earlier using data obtained from a point reactor model under assumed conditions. Actual operational data from The Ohio State University Research Reactor (OSURR) and simulated operational data from the Oconee plant are used to extend this earlier work. It is shown that the point reactor model, joined with an efficient GA parameter estimation procedure, can be used for accurate prediction of global xenon evolution in small reactors (e.g., OSURR). It is also shown that this approach yields just qualitatively correct results in large reactors (e.g., Oconee) where spatial effects become significant. By continuously updating the model parameters obtained by GAs, xenon induced reactivity during transients can be estimated purely from the past reactivity and power data without a knowledge of initial conditions for 135Xe and 135I