Neutron flux estimation using Particle-H∞H∞ Filter for fast transient tracking

• Derivation for large core loosely coupled point kinetic modeling.
• Introduction of the concept of hybrid filtering for neutron flux estimation.
• Derivation of particle filtering and hybrid Particle-H∞H∞ Filtering.
• Simulation of hybrid particle filter algorithm for neutron flux estimation.
• Comparison of flux estimation accuracy with conventional verses hybrid particle filter.

The non-linear target tracking for fast transients using particle filtering or Sequential Monte Carlo based technique usually suffers from the common problem of sample impoverishment. The sample impoverishment has been observed to become much severe, for compact measurement densities. In this work, we propose a hybrid estimator for nonlinear target tracking which utilizes the strengths of both Sequential Monte Carlo approach and minimax estimator. The robust H∞H∞ filtering minimizes the maximum gain from disturbance to the estimation error. The estimation is performed in two phases; the first phase enables the state transitions of samples through Monte Carlo method; the second phase updates the samples through H∞H∞ correction using Gaussian observation density function. We validate the proposed method for fast reactivity transient in large core Pressurized Heavy Water Reactor model which occurs in a postulated accident of primary heat transport system integrity loss. The neutron flux variation with respect to the reactivity in large core reactor is approximately described by a set of point kinetic equations (PKEs). A number of especially distributed Self Powered Neutron Detectors are used as in-core devices to sense the point flux at their respective locations and together represents the overall core status. We also present the experimental work performed in SPNDs to develop the uncertainty model.