Large LOCA BEAU Analysis of a Generic 900MW CANDU Plant

Abstract

<p>A RELAP5 input model representative of a 900 MW CANDU plant was created. Due to the one dimensional nature of the code, complex plant components such as heat transport system pumps, fuel channels, and the reactor core were modeled using thermal hydraulically equivalent geometries. The model consists of 301 thermal hydraulic components and heat structures with five boundary conditions specifying the pressure, temperature, and in some cases the mass flow. Spacial convergence of the model was ensured through the implementation of various channel nodalizations (6, 12 and 24 axial nodes) and core representations (single channel group and 6 channel groups per core pass). A control subroutine representative of the pressure control system and steam generator control system was implemented in the model, mimicking the controllers that exist in operational CANDU Reactors. A point kinetics model was used to simulate the reactor power.</p> <p>Uncertainties for parameters related to the modeling of specific phenomenons were obtained from literature sources related to the models in which the parameters belong to. Where sources of open literature were unavailable (error in the void reactivity coefficient, and trip delay times), an estimate was made for the parameter uncertainties. Uncertainties for plant specific parameters were derived from plant operating data obtained either from plant instrument measurements (RIH pressure, RIH temperature, etc) or through code simulation (channel powers, bundle powers, etc). Breaks in the NW reactor inlet header were considered. Due to the additional flow from the pressurizer present in the east loop, it was found that breaks in this loop would lead to slightly lower results. The bulk net energy deposition (NED) related to expected maximum sheath temperatures, and the reactor peak power related to the fuel heat up rate were chosen as the figures of merit. Breaks with a size of 48% of the reactor inlet header area were found to lead to the highest NED. From parametric sensitivity studies performed at the best estimate and non best estimate points for the 48% break, a Phenomena and Key Parameter Identification and Ranking Table (PKPIRT) was established with the void reactivity feed back identified as the highest ranked parameter. It was found that the ranking based on best estimate local studies does not reject parameters that may be of importance at their 95th percentile (non best estimate point).</p> <p>Two best estimate methods along with a conservative approach were implemented. A brute force monte-carlo method consisting of 10,000 simulations provided a probability distribution function for the figures of merits. First and second order GRS methods were found to adequately predict an upper bound for the 95th percentile of the FOMs. Both best estimate methods demonstrated that the consequences of a hypothetical large break LOCA are much more benign than those predicted by the LOE approach.</p>